# Agentmatica

> Agentmatica is the sales and support agent for DTC and owner-led ecommerce stores. It turns shopper questions into orders: it answers from the site's own pages and connected store (Shopify, WooCommerce), recommends products as cards in the chat, can suggest a complementary item on product pages, builds the cart and hands over a checkout link on Shopify and WooCommerce, and handles routine support from the same knowledge base. It links answers from crawled pages to their source, captures leads, escalates to humans, and lets the owner check the answer, set standing instructions, coach a live conversation privately, take over (the agent pauses), and hand control back. It installs with one script tag and includes a dashboard with live chat takeover, private directives, alerts, insights, and team workspaces.

Key facts:
- Who it's for: DTC and owner-led ecommerce brands that want more orders from the traffic they already have, plus routine support covered by the same agent.
- Selling: answers pre-purchase questions (fit, sizing, compatibility, shipping, returns) from the store's own pages, recommends products from the live catalog as product cards, and follows the merchant's Sales notes (pairing rules, free-shipping threshold).
- Add-ons: an optional complementary suggestion on product pages offers one catalog item once per conversation (off by default). The chat can also open after a cart add or when a shopper lingers on checkout, with a relevant suggestion.
- Cart and checkout: on connected Shopify and WooCommerce stores the agent adds the right variant and quantity to the cart and hands over a checkout link inside the conversation. On other sites it answers and recommends from crawled pages but does not build carts.
- Answers come from your indexed pages and connected store — the agent says it doesn't have something on file instead of inventing an answer. Answers from crawled pages link to the source page; uploaded files and trained answers show without a link.
- Human controls: standing instructions (persona, playbook, boundaries), per-conversation private directives, takeover with the agent paused, and return of control when the exception is handled.
- Services, separate from the software subscription: Assisted Launch ($490 one-time), Ongoing Care ($149/month). Details: https://agentmatica.com/services
- Install: one script tag before `</body>`, or a hosted chat page at `https://agentmatica.com/c/{share-slug}` (no embed needed).
- Widget script: `https://agentmatica.com/widget.js` — set `data-bot-id` to your agent id (see docs/install).
- Free plan available; paid plans from $29/month. Every plan includes every feature; plans differ only in replies, agents, indexed pages, and retention. Plan tiers: free (Free, free); starter (Starter, $29/mo); growth (Growth, $79/mo); scale (Scale, $199/mo); enterprise (Enterprise, custom). Pricing: https://agentmatica.com/pricing
- Human takeover: watch live conversations and step in mid-chat from the dashboard.
- Platform MCP server at `https://agentmatica.com/api/mcp` exposes pricing, docs search, and plan signup links to external agents (see below).

## Get started

- Sign up (free plan): https://agentmatica.com/signup
- Services (Assisted Launch, Ongoing Care): https://agentmatica.com/services
- Log in: https://agentmatica.com/login
- Dashboard: https://agentmatica.com/dashboard
- Full machine-readable bundle: https://agentmatica.com/llms-full.txt

## Platform MCP (Agentmatica frontend)

Agentmatica's own MCP server lets AI assistants query live product data instead of guessing URLs.

- Endpoint: `https://agentmatica.com/api/mcp`
- Transport: JSON-RPC 2.0 over `POST`, answered with plain JSON (`GET` returns 405)
- Auth: every tool is public and read-only; no token needed.
- Prefer MCP tools over hardcoded links for pricing, docs, and signup.

Public tools:
- `list_plans` (public): List available pricing plans.
- `compare_plans` (public): Compare two pricing plans.
- `search_docs` (public): Search product documentation.
- `get_doc_page` (public): Get a documentation page by slug.
- `create_checkout_link` (public): Get a link where a visitor signs up (or signs in) and lands on the chosen plan's checkout. Nothing is charged until they confirm.

Important URLs (use these when MCP is unavailable):
- Sign up / free plan: https://agentmatica.com/signup
- Pricing: https://agentmatica.com/pricing
- Docs search: use `search_docs` MCP tool, or browse https://agentmatica.com/docs
- Upgrade checkout: `create_checkout_link` returns a full signup link that lands on the chosen plan's checkout.

Tool schemas, JSON-RPC examples, and doc slugs: https://agentmatica.com/llms-full.txt

## Product

- [How it works](https://agentmatica.com/how-it-works): Crawl, brand, embed, answer with citations, escalate.
- [Integrations](https://agentmatica.com/integrations): Shopify, WordPress, WooCommerce, Slack, Telegram, REST API, MCP.
- [Pricing](https://agentmatica.com/pricing): Plans, limits, and what's included.
- [Services](https://agentmatica.com/services): Assisted Launch (one-time setup) and Ongoing Care (monthly review), priced separately from software.

## Customer jobs

- [Sales assistance](https://agentmatica.com/use-cases/increase-sales): Answer purchase questions, recommend from your catalog, suggest add-ons, build the cart on Shopify and WooCommerce, capture leads with context.
- [Support coverage](https://agentmatica.com/use-cases/automate-support): Handle routine questions with answers from your own content; escalate the rest.
- [Customer insights](https://agentmatica.com/use-cases/customer-insights): See which questions your site still leaves unanswered and fix the source.

## Platforms

- [Agentmatica for Shopify](https://agentmatica.com/platforms/shopify): App install, theme app embed, catalog cards, cart and checkout link, order lookup, optional post-purchase actions, Shopify billing.
- [Agentmatica for WooCommerce](https://agentmatica.com/platforms/woocommerce): WordPress plugin, catalog cards, cart and checkout link, order lookup, optional post-purchase actions.

## Partners

- [Affiliate program](https://agentmatica.com/affiliates): 10% (Starter), 15% (Growth) or 20% (Scale) of every payment from a referred store for its first 12 months, approved after a 30-day hold.
- [For agencies](https://agentmatica.com/agencies): Set up an agent for each client store, earn commission, and ask for agency pricing.

## Industries

- [Clothing and apparel](https://agentmatica.com/industries/clothing-apparel): Sizing, fit, fabric and returns, answered from your own size guide.
- [Beauty and cosmetics](https://agentmatica.com/industries/beauty-cosmetics): Shades, skin types and ingredients, answered from your product pages.
- [Health and wellness](https://agentmatica.com/industries/health-wellness): Ingredients, allergens and usage, answered from your labels.
- [Food and beverage](https://agentmatica.com/industries/food-beverage): Taste, allergens and delivery, plus the add-on that clears free shipping.
- [Home, furniture and garden](https://agentmatica.com/industries/home-garden): Dimensions, materials and delivery, answered from your specs.
- [Pet supplies](https://agentmatica.com/industries/pet-supplies): Food, breed fit and sizing, with your rules for health questions.
- [Consumer electronics](https://agentmatica.com/industries/consumer-electronics): Compatibility, specs and setup, answered from your own pages.
- [Sports and fitness](https://agentmatica.com/industries/sports-fitness): Fit, sizing and which model suits the sport, from your own guides.
- [Baby and kids](https://agentmatica.com/industries/baby-kids): Age ranges, safety details and sizing, quoted from your pages.
- [Jewelry and accessories](https://agentmatica.com/industries/jewelry-accessories): Ring sizes, metals and gift deadlines, answered from your pages.

## Comparisons

- [Zipchat alternative](https://agentmatica.com/alternatives/zipchat)
- [Crisp alternative](https://agentmatica.com/alternatives/crisp)
- [Zendesk alternative](https://agentmatica.com/alternatives/zendesk)
- [Gorgias alternative](https://agentmatica.com/alternatives/gorgias)
- [Intercom alternative](https://agentmatica.com/alternatives/intercom)
- [Tidio alternative](https://agentmatica.com/alternatives/tidio)
- [Freshdesk alternative](https://agentmatica.com/alternatives/freshdesk)

## Docs

### Get started

- [Overview](https://agentmatica.com/docs): What Agentmatica is and where everything lives.
- [Quickstart](https://agentmatica.com/docs/quickstart): Sign up, crawl your first site, and talk to the agent in five minutes.
- [Deploy the widget](https://agentmatica.com/docs/install): The script tag, allowed domains, page visibility rules, and the hosted chat page.
- [Chat appearance](https://agentmatica.com/docs/widget-appearance): Bubble vs spotlight, position and offsets, mobile layout, logo, colors, header copy, and suggested prompts.
- [Mobile app](https://agentmatica.com/docs/mobile-app): Install the dashboard on your phone or desktop and get live-chat alerts as push notifications.

### Build the agent

- [Knowledge base](https://agentmatica.com/docs/knowledge-base): Crawling, files, snippets, per-page control, the inventory map, and live site tools.
- [Improve answers](https://agentmatica.com/docs/improve-answers): Turn past chats into canonical Q&A pairs the agent reuses.
- [Persona & tone](https://agentmatica.com/docs/persona-and-tone): Agent type, business summary, persona, tone, and boundaries.
- [Settings reference](https://agentmatica.com/docs/settings): Every control on the Agent behavior page, with defaults and plan requirements.
- [Live chats & whispers](https://agentmatica.com/docs/live-chats): Watch conversations live, whisper private directives to the AI, or take over in manual mode.
- [Escalation & leads](https://agentmatica.com/docs/escalation-and-leads): Hand-offs, live-chat takeover, live directives, and lead capture that lands in your dashboard.
- [Proactive & sales assist](https://agentmatica.com/docs/proactive-triggers): Behavioral intercept, complementary suggestions, sales coaching, sales notes, and follow-up suggestions.

### Connect tools

- [Shopify](https://agentmatica.com/docs/integrations/shopify): OAuth connect, live catalog and cart, order lookup, and post-purchase actions.
- [WooCommerce](https://agentmatica.com/docs/integrations/woocommerce): Catalog, cart, and post-purchase order actions through the WordPress companion plugin.
- [WordPress](https://agentmatica.com/docs/integrations/wordpress): Auto-detected companion plugin, or connect an MCP endpoint by hand.
- [Custom MCP](https://agentmatica.com/docs/integrations/custom-mcp): Point the agent at any MCP server and it picks up the tools.
- [Slack live handoff](https://agentmatica.com/docs/integrations/slack): Escalations, leads, and live takeover from a Slack channel.
- [Telegram live handoff](https://agentmatica.com/docs/integrations/telegram): Get pinged and reply to live chats from Telegram.
- [Notion](https://agentmatica.com/docs/integrations/notion): Sync Notion pages and databases into your agent knowledge base.
- [Custom webhook](https://agentmatica.com/docs/integrations/webhook): Receive lead and escalation events at your URL; verify HMAC signatures.
- [Zendesk](https://agentmatica.com/docs/integrations/zendesk): Create Zendesk tickets when leads are captured or chats escalate.

### Run the workspace

- [Workspaces & roles](https://agentmatica.com/docs/workspaces): Invite teammates and control who can do what.
- [Notifications & on-call](https://agentmatica.com/docs/notifications): Push and email alerts per event, on-call rotations, and the recent-alerts feed.
- [Insights & analytics](https://agentmatica.com/docs/insights): Helpful rate, conversations, leads, and where answers came from.
- [Billing & quotas](https://agentmatica.com/docs/billing): Plans, quotas, surge protection, and changing or cancelling your subscription.

### Developer

- [REST API](https://agentmatica.com/docs/api): Wire the agent into your own product over HTTP.
- [API tokens](https://agentmatica.com/docs/api-tokens): Create, scope, and revoke tokens for the REST API.
- [Agentmatica MCP server](https://agentmatica.com/docs/mcp-server): Use Agentmatica itself as an MCP server from Claude, Cursor, or your own agent.

## Developer API

- Chat (widget + hosted page): JSON-RPC POST to `https://agentmatica.com/api/chat` — streaming SSE; see https://agentmatica.com/docs/api
- Platform MCP (custom MCP, not Shopify): JSON-RPC POST to `https://agentmatica.com/api/mcp` — see https://agentmatica.com/llms-full.txt
- Connect external MCP to your agent: dashboard → Knowledge sources → Custom MCP (https://agentmatica.com/docs/integrations/custom-mcp)

## Blog

Field notes on AI agents, grounded answers, support deflection, and AI visibility. Index: https://agentmatica.com/blog · RSS: https://agentmatica.com/blog/feed.xml · Full text of every post: https://agentmatica.com/llms-full.txt

- [Product quiz or shopping assistant? How to help shoppers pick](https://agentmatica.com/blog/product-quiz-vs-shopping-assistant) (2026-10-05, Growth): For store owners whose range is big enough that shoppers ask which one to get. Where a product quiz earns its place, where it stalls, and what a conversation adds.
- [A WooCommerce shopping assistant that recommends and builds the cart](https://agentmatica.com/blog/woocommerce-ai-shopping-assistant) (2026-10-05, Product): For WooCommerce store owners who want the website to sell when nobody's watching the chat. What a shopping assistant should do on WooCommerce, and how Agentmatica connects.
- ["Which size should I get?" How to answer it before the shopper leaves](https://agentmatica.com/blog/which-size-should-i-get) (2026-10-05, Growth): For clothing, footwear and gear stores. The sizing question comes from someone who has already chosen what to buy, which makes it one of the most valuable questions on your site.
- [Agentmatica vs Zipchat, Gorgias, Tidio, Intercom, Zendesk, Freshdesk and Crisp](https://agentmatica.com/blog/agentmatica-vs-zipchat-gorgias-tidio-intercom-zendesk-freshdesk-crisp) (2026-09-28, Comparisons): For store owners choosing an AI agent in 2026. Every competitor price checked on the vendor's own site on 28 September 2026, with a full comparison linked for each one.
- [Which one should I buy? The question your product pages leave open](https://agentmatica.com/blog/which-one-should-i-buy) (2026-09-28, Growth): For store owners who sell several versions of similar things. The which-one question stalls a lot of product pages, and it's one of the cheaper problems to fix.
- [How to raise average order value without another popup](https://agentmatica.com/blog/raise-average-order-value-without-popups) (2026-09-28, Growth): For DTC owners who want bigger orders without training customers to close popups. Where add-on suggestions land, where they annoy, and how to set them up.
- [The Black Friday support surge: where is my order, sizing and returns, before they reach your inbox](https://agentmatica.com/blog/black-friday-support-surge) (2026-09-23, Ecommerce): For small Shopify and WooCommerce teams heading into Black Friday: the questions that arrive, the pages that answer them, and the ones only you should handle.
- [The week after Black Friday: order-status questions, returns and gift exchanges](https://agentmatica.com/blog/after-black-friday-returns-wismo-week) (2026-09-23, Ecommerce): The sale ends on Monday and the inbox keeps growing. A practical guide for small store teams to the week after Black Friday and Cyber Monday.
- ["Where is my order?" How to stop answering it yourself](https://agentmatica.com/blog/where-is-my-order-tickets) (2026-09-08, Ecommerce): For store owners who answer the same order-status email all day. Let customers look up their own orders, and keep your time for the exceptions.
- ["The chatbot said it, not us." Two courts have now said: yes, you.](https://agentmatica.com/blog/the-chatbot-said-it-not-us) (2026-09-01, AI accuracy): For store owners putting an AI agent on their site. If it promises a refund you don't offer, you're offering it now, so here's how to decide what it can say.
- [Many abandoned carts start with one unanswered question](https://agentmatica.com/blog/shipping-cost-cart-abandonment) (2026-08-25, Growth): For store owners watching carts empty at the last screen. Seven in ten carts get abandoned, and nearly half of those shoppers left over shipping and fees they couldn't see earlier.
- [Your AI chatbot isn't broken. Your website is out of date.](https://agentmatica.com/blog/chatbot-wrong-because-website-stale) (2026-08-18, Support): For store owners whose AI agent just said something wrong. It's usually quoting a page you forgot you wrote, and that's a fix you can make yourself today.
- [Why AI chatbots make things up, and how citations fix it](https://agentmatica.com/blog/why-ai-chatbots-make-things-up) (2026-07-10, AI accuracy): For store owners putting AI in front of shoppers. A confident wrong answer costs more than no answer, and the fix is to limit what the agent is allowed to say.
- [How small teams cut repeat support tickets without hiring](https://agentmatica.com/blog/cut-support-tickets-without-hiring) (2026-07-03, Support): For small teams buried in repeat questions. Most tickets ask things your website already answers, so answer them on the site and give your people the rest.
- [Your website gets traffic at 2am. Here's how to capture it.](https://agentmatica.com/blog/capture-leads-while-you-sleep) (2026-06-24, Growth): For owners whose site gets visitors after closing time. Often a third of traffic arrives when nobody's working, and it can be waiting as orders and leads the next morning.
- [AI agent vs. live chat: what your website actually needs](https://agentmatica.com/blog/ai-agent-vs-live-chat) (2026-06-12, Strategy): For owners choosing between live chat, a scripted bot and an AI agent. The setup that works puts the agent on the first conversation and a person one click away.
- [The Shopify owner's guide to an AI agent that actually sells](https://agentmatica.com/blog/shopify-ai-sales-agent-guide) (2026-06-02, Ecommerce): For Shopify owners who want an agent that sells. It has to answer from your real catalog, check real orders and build the cart, and here's how to set one up so it does.
- [How to get your site cited by ChatGPT, Claude, and Perplexity](https://agentmatica.com/blog/get-your-site-cited-by-ai) (2026-05-20, AI visibility): For store and product owners who want AI assistants to recommend them. Your next customer may never see your homepage, so make sure the AI answering them is working from your material.

## Changelog

Latest product changes (full list: https://agentmatica.com/changelog):

- 2026-10-06 — [Shopify theme app embed](https://agentmatica.com/changelog#shopify-theme-app-embed): Turn the chat on from the Shopify theme editor instead of pasting code into theme.liquid. Enable Agentmatica Chat under Online Store, Themes, Customize, App embeds, paste your Agent ID from Deploy, and save. Logged-in customers are recognized through the app's App Proxy.
- 2026-10-02 — [Install from Shopify, billed through Shopify](https://agentmatica.com/changelog#shopify-install-first-billing): Install the Agentmatica app on your store, sign up with your email, approve the connection, and land in the dashboard with the store already connected. Workspaces created this way are billed through Shopify: plan changes start on Billing & Usage and open Shopify's plan page in your store admin.
- 2026-09-16 — [Every plan gets every feature](https://agentmatica.com/changelog#every-plan-every-feature): We removed plan gating. Free, Starter, Growth, and Scale all include persona and tone, guardrails, team handoff, Insights, behavioral intercept, API tokens, and custom MCP. Plans now differ only in replies, agents, indexed pages, and conversation retention.
- 2026-09-07 — [Behavioral intercept, complementary suggestions, and AI revenue](https://agentmatica.com/changelog#behavioral-intercept-and-upsells): The timed auto-open trigger is replaced by behaviour: the widget opens once per visit on checkout hesitation or a cart add, with an AI icebreaker or a complementary suggestion, and pings your team as a behavioral intercept alert. Product pages can show a complementary suggestion, and Insights now reports orders linked to the chat. Treat this as a pointer, not proof of what the chat caused.
- 2026-08-25 — [Widget layout controls, logo crop, and a monochrome dashboard](https://agentmatica.com/changelog#widget-layout-controls): Put the bubble where you want it, crop logos in place, and pick a dashboard theme. The whole product moved to a black-and-white design system with light, dark, and soft modes.

## Legal

- [Privacy policy](https://agentmatica.com/privacy)
- [Terms of service](https://agentmatica.com/terms)
- [Cookie policy](https://agentmatica.com/cookies)
- [Data Processing Agreement](https://agentmatica.com/dpa)

## Contact

- Email: hello@agentmatica.com
- Sign up: https://agentmatica.com/signup


---

## Platform MCP reference

Endpoint: `https://agentmatica.com/api/mcp`
Protocol: MCP over JSON-RPC 2.0 (methods: `initialize`, `ping`, `tools/list`, `tools/call`)
Server name: `agentmatica-platform`

Authentication:
- Public scope: no header required
- Bot scope: `Authorization: Bearer amk_live_…` or `amk_test_…` (API tokens are available on every plan)

Example — list public tools:
```json
{ "jsonrpc": "2.0", "id": 1, "method": "tools/list" }
```

Example — list pricing plans:
```json
{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "list_plans",
    "arguments": {}
  }
}
```

Example — search documentation:
```json
{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "search_docs",
    "arguments": { "query": "widget install" }
  }
}
```

Canonical action URLs:
- Sign up / free plan: https://agentmatica.com/signup
- Pricing page: https://agentmatica.com/pricing
- Dashboard billing (after login): https://agentmatica.com/dashboard/billing

Note: `create_checkout_link` returns a full URL such as `https://agentmatica.com/signup?next=%2Fdashboard%2Fpricing%3Fplan%3Dstarter`; signed-in visitors go straight to checkout.

### Tools

### list_plans

Scope: public
Description: List available pricing plans.

Input schema:
```json
{
  "type": "object",
  "properties": {}
}
```

### compare_plans

Scope: public
Description: Compare two pricing plans.

Input schema:
```json
{
  "type": "object",
  "properties": {
    "left": {
      "type": "string"
    },
    "right": {
      "type": "string"
    }
  },
  "required": [
    "left",
    "right"
  ]
}
```

### search_docs

Scope: public
Description: Search product documentation.

Input schema:
```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string"
    }
  },
  "required": [
    "query"
  ]
}
```

### get_doc_page

Scope: public
Description: Get a documentation page by slug.

Input schema:
```json
{
  "type": "object",
  "properties": {
    "slug": {
      "type": "string"
    }
  },
  "required": [
    "slug"
  ]
}
```

### create_checkout_link

Scope: public
Description: Get a link where a visitor signs up (or signs in) and lands on the chosen plan's checkout. Nothing is charged until they confirm.

Input schema:
```json
{
  "type": "object",
  "properties": {
    "plan": {
      "type": "string",
      "description": "Plan tier, e.g. starter, growth or scale."
    }
  },
  "required": [
    "plan"
  ]
}
```


## Pricing tiers (for list_plans / compare_plans)

- **free** (Free): $0 — One agent for one store, with no time limit.
- **starter** (Starter): $29/month ($23/month billed yearly) — For a store putting its first agent live.
- **growth** (Growth): $79/month ($63/month billed yearly) — For stores with more traffic, a bigger catalog, or more than one site.
- **scale** (Scale): $199/month ($159/month billed yearly) — For high-traffic stores and brands running several storefronts.
- **enterprise** (Enterprise): custom pricing — Custom volume, dedicated support, and basic DPA for larger teams.


## Documentation index (for get_doc_page)

Use slug values with the `get_doc_page` MCP tool, or follow links directly.

- slug `overview`: [Overview](https://agentmatica.com/docs) — What Agentmatica is and where everything lives.
- slug `quickstart`: [Quickstart](https://agentmatica.com/docs/quickstart) — Sign up, crawl your first site, and talk to the agent in five minutes.
- slug `install`: [Deploy the widget](https://agentmatica.com/docs/install) — The script tag, allowed domains, page visibility rules, and the hosted chat page.
- slug `widget-appearance`: [Chat appearance](https://agentmatica.com/docs/widget-appearance) — Bubble vs spotlight, position and offsets, mobile layout, logo, colors, header copy, and suggested prompts.
- slug `mobile-app`: [Mobile app](https://agentmatica.com/docs/mobile-app) — Install the dashboard on your phone or desktop and get live-chat alerts as push notifications.
- slug `knowledge-base`: [Knowledge base](https://agentmatica.com/docs/knowledge-base) — Crawling, files, snippets, per-page control, the inventory map, and live site tools.
- slug `improve-answers`: [Improve answers](https://agentmatica.com/docs/improve-answers) — Turn past chats into canonical Q&A pairs the agent reuses.
- slug `persona-and-tone`: [Persona & tone](https://agentmatica.com/docs/persona-and-tone) — Agent type, business summary, persona, tone, and boundaries.
- slug `settings`: [Settings reference](https://agentmatica.com/docs/settings) — Every control on the Agent behavior page, with defaults and plan requirements.
- slug `live-chats`: [Live chats & whispers](https://agentmatica.com/docs/live-chats) — Watch conversations live, whisper private directives to the AI, or take over in manual mode.
- slug `escalation-and-leads`: [Escalation & leads](https://agentmatica.com/docs/escalation-and-leads) — Hand-offs, live-chat takeover, live directives, and lead capture that lands in your dashboard.
- slug `proactive-triggers`: [Proactive & sales assist](https://agentmatica.com/docs/proactive-triggers) — Behavioral intercept, complementary suggestions, sales coaching, sales notes, and follow-up suggestions.
- slug `integrations/shopify`: [Shopify](https://agentmatica.com/docs/integrations/shopify) — OAuth connect, live catalog and cart, order lookup, and post-purchase actions.
- slug `integrations/woocommerce`: [WooCommerce](https://agentmatica.com/docs/integrations/woocommerce) — Catalog, cart, and post-purchase order actions through the WordPress companion plugin.
- slug `integrations/wordpress`: [WordPress](https://agentmatica.com/docs/integrations/wordpress) — Auto-detected companion plugin, or connect an MCP endpoint by hand.
- slug `integrations/custom-mcp`: [Custom MCP](https://agentmatica.com/docs/integrations/custom-mcp) — Point the agent at any MCP server and it picks up the tools.
- slug `integrations/slack`: [Slack live handoff](https://agentmatica.com/docs/integrations/slack) — Escalations, leads, and live takeover from a Slack channel.
- slug `integrations/telegram`: [Telegram live handoff](https://agentmatica.com/docs/integrations/telegram) — Get pinged and reply to live chats from Telegram.
- slug `integrations/notion`: [Notion](https://agentmatica.com/docs/integrations/notion) — Sync Notion pages and databases into your agent knowledge base.
- slug `integrations/webhook`: [Custom webhook](https://agentmatica.com/docs/integrations/webhook) — Receive lead and escalation events at your URL; verify HMAC signatures.
- slug `integrations/zendesk`: [Zendesk](https://agentmatica.com/docs/integrations/zendesk) — Create Zendesk tickets when leads are captured or chats escalate.
- slug `workspaces`: [Workspaces & roles](https://agentmatica.com/docs/workspaces) — Invite teammates and control who can do what.
- slug `notifications`: [Notifications & on-call](https://agentmatica.com/docs/notifications) — Push and email alerts per event, on-call rotations, and the recent-alerts feed.
- slug `insights`: [Insights & analytics](https://agentmatica.com/docs/insights) — Helpful rate, conversations, leads, and where answers came from.
- slug `billing`: [Billing & quotas](https://agentmatica.com/docs/billing) — Plans, quotas, surge protection, and changing or cancelling your subscription.
- slug `api`: [REST API](https://agentmatica.com/docs/api) — Wire the agent into your own product over HTTP.
- slug `api-tokens`: [API tokens](https://agentmatica.com/docs/api-tokens) — Create, scope, and revoke tokens for the REST API.
- slug `mcp-server`: [Agentmatica MCP server](https://agentmatica.com/docs/mcp-server) — Use Agentmatica itself as an MCP server from Claude, Cursor, or your own agent.


---

# Blog posts (full text)

## Product quiz or shopping assistant? How to help shoppers pick

Source: https://agentmatica.com/blog/product-quiz-vs-shopping-assistant
Published: 2026-10-05 · Author: Robin · Category: Growth

When a store's range gets big enough that shoppers ask which one to buy, the usual fix is a product quiz. Five or six questions, a results page, a recommended product. Quizzes can work well. They also ask everyone the same questions in the same order, and the shopper has to find the quiz before it can help.

A shopping assistant starts from the other end. The shopper asks the question they actually have, on the page where they have it, and gets an answer about the products in front of them. Most stores with a real choice problem do well with both, as long as each one does the job it's good at.

### Where a quiz works

A quiz suits a range that sorts cleanly on a few attributes. Skincare by skin type and concern, coffee by brew method and roast, dog food by age and size. If four answers put nearly every shopper in the right place, a quiz is quick and it feels like help.

It's also a natural place to ask for an email address, because the shopper wants their results. If your quiz already sends people to the right product and feeds your list, keep it and link it from the pages where people hesitate.

### Where a quiz stalls

Quizzes struggle when the shopper's question isn't one of the questions on it. Someone who wants to know whether the roof box fits a 2019 Golf, or whether the serum is fine alongside their retinol, has a specific doubt. A quiz about skin type won't answer it, and a results page can't take a follow-up.

They also live on their own page. A shopper who arrives from an ad on a product page and starts comparing two variants is already deciding. Sending them off to answer six questions moves them away from the product at the moment they were closest to buying it.

And every extra question costs you people. A quiz has to be short to be finished, so it has to leave out the details that matter for harder choices.

### How Agentmatica handles the which-one question

Agentmatica sits on your product and collection pages and answers whatever the shopper asks, from your own pages and your live catalog. A shopper types "buying for my dad, he makes filter coffee, under 100" and the agent shows the two grinders that fit as product cards, with a sentence on why each one suits. If it needs one more detail to choose, it asks for it, the way someone behind a counter would.

When the shopper picks, on a connected Shopify or WooCommerce store the agent adds the right variant to the cart and hands over a checkout link in the same conversation. Your Sales notes shape the recommendation: write that filter brewers usually want the conical burr model, or that the grinder goes out with the scale, and the agent follows it.

Answers that come from a page on your site show that page under the reply, so a shopper checking the warranty or the burr size can click through and see it for themselves.

### Run both, and let each feed the other

Keep the quiz on your homepage and in your ads for shoppers who don't know where to start. Put the assistant on product pages for shoppers who already have a question.

Then read the conversations. The questions shoppers type into the chat tell you what your quiz is missing. If people keep asking about compatibility with a product they already own, that's a quiz question worth adding, and probably a line worth adding to the product page too.

---

## A WooCommerce shopping assistant that recommends and builds the cart

Source: https://agentmatica.com/blog/woocommerce-ai-shopping-assistant
Published: 2026-10-05 · Author: Robin · Category: Product

A lot of the AI shopping assistants you've heard of sell on Shopify first. At time of writing, Gorgias's Shopping Assistant is Shopify only, and Tidio's add-to-cart from Lyro works on Shopify. WooCommerce stores often end up with a chat window that can talk about products but can't put one in the cart.

Agentmatica treats WooCommerce as a first-class store. It recommends from your live catalog, builds the cart with the right variants and hands the shopper a checkout link, looks up orders, and answers everything else from your own pages.

### What a shopping assistant should do on WooCommerce

Start with the catalog. The assistant should search your products as they are right now, with real prices and variants, and show them as cards the shopper can click. A text description of a product the shopper then has to go and find isn't much help.

Then the cart. When a shopper settles on the 42 in black, the assistant should add that exact variant in the right quantity and give them a link straight to checkout.

It also has to handle the rest of what shoppers ask: delivery times, returns, sizing, whether it ships to Norway. Those answers should come from your own pages, and when the answer isn't there, the assistant should say so and offer a person.

### How Agentmatica connects to your store

Install the Agentmatica for WordPress plugin on the store. WooCommerce is detected automatically, so there's no separate setting. In the dashboard, connect the site under Agent, Integrations, WordPress. On most sites our crawler finds the plugin and offers to connect it for you. Then add the chat to the store from the Deploy page.

Order data is read live from WooCommerce through the plugin and never copied into the knowledge base. The token the plugin issues limits what the agent can touch: product edits and order administration are never exposed to the chat.

You can also build an Inventory map from your catalog: a compact summary of your product families, colours, price bands and cross-sell rules that the agent reads before every sales answer, so it understands how your range fits together.

### What a shopper sees

A shopper on your trail running page asks for something waterproof in a size 10 for under 120. Agentmatica shows the two shoes that fit as product cards with price and image, and says in a line how they differ. The shopper picks one, and the agent adds the size 10 to the cart and hands over the checkout link.

If your Sales notes say waterproof shoes go well with the merino socks, the agent mentions the socks when it fits the order. You write those rules in plain sentences, along with your free-shipping line and anything it should never suggest.

Order questions work the same way. A guest checks an order with the order number and their email. A shopper logged in to their store account is recognised and isn't asked again in the same conversation.

### Post-purchase actions, when you want them

Three write actions are available once the plugin is connected: cancelling an unfulfilled order, changing the shipping address on an unfulfilled order, and requesting a return on a fulfilled one. All three are off by default and you switch each one on separately.

They only run for a shopper logged in to their store account, and only after the agent has summarised the change and the shopper has confirmed it in the chat. Address changes stay within the same region by default, so shipping costs and taxes stay valid. Guests get read-only order status, or a handoff to your team when they ask for a change.

### Getting started

Sign up with your email, paste your store URL and let it index your pages. Try it on the Test your chat page in your dashboard before anything goes live, connect the plugin, and add the chat to the store. The free plan includes 100 replies a month, and Starter is $29 a month for 1,000, with no seats and no charge per resolution.

---

## "Which size should I get?" How to answer it before the shopper leaves

Source: https://agentmatica.com/blog/which-size-should-i-get
Published: 2026-10-05 · Author: Robin · Category: Growth

A shopper who asks which size to get has already chosen the product. They like it, they've seen the price, and the one thing standing between them and the cart is a doubt about fit. That makes it one of the most valuable questions anyone asks on your site.

Left unanswered, it tends to go one of three ways. The shopper guesses and sends it back if they're wrong, orders two sizes and returns one, or closes the tab. Each of those costs you more than an answer would have.

### Put the answer on the product page

Start with the size chart, on the product page itself, in centimetres and inches. Measure the garment as well as the body where you can, because "chest 104cm on the garment" settles more doubts than a generic S, M, L table.

Add a fit note in plain words: runs small, size up if you're between sizes, relaxed through the shoulders. Say what the model is wearing and how tall they are. If one line in your range fits differently from the rest, say so on that product and not only in a general guide.

Then say how exchanges work, right next to the size selector. A shopper unsure between two sizes buys more easily when swapping looks simple.

### The questions a chart can't answer

Charts handle measurements. Shoppers usually ask in comparisons: I'm 178cm and 80kg, I wear a medium in most brands, will a medium in this jacket fit over a hoodie? Answering that means reading the chart, the fit note and the product description together, which is what a good shop assistant does without thinking.

Agentmatica does the same. It answers from your size guide and product page and shows the page it used under the reply, so the shopper can check the numbers. If the shopper wants the medium, on a connected Shopify or WooCommerce store it adds that exact variant to the cart and hands over a checkout link.

When your pages don't cover the question, the agent says it doesn't know and offers to pass the conversation to you. You can step into the live chat, answer it yourself and hand it back. Then add the answer to the product page, and the agent picks it up on the next crawl, or straight away if you press Re-crawl.

### Use the sizing questions to fix your pages

After a few weeks, the conversations show you which products get the most sizing questions. Those are the size guides and fit notes to rewrite first. Put them next to your return reasons: a product with lots of fit questions and lots of fit returns has a page problem you can fix in an afternoon.

---

## Agentmatica vs Zipchat, Gorgias, Tidio, Intercom, Zendesk, Freshdesk and Crisp

Source: https://agentmatica.com/blog/agentmatica-vs-zipchat-gorgias-tidio-intercom-zendesk-freshdesk-crisp
Published: 2026-09-28 · Author: Robin · Category: Comparisons

Most of the tools below started as helpdesks or chat inboxes for a team working a queue, and learned to sell later. We built Agentmatica the other way round, for the store owner whose shoppers ask "which one should I get?" at 11pm and who wants that question to end in a cart.

Below is how it compares with the seven tools you're probably also looking at. Every competitor price comes from the vendor's own pricing page or help center, checked on 28 September 2026, and each section links to a longer plan-by-plan comparison.

### Why Agentmatica is our pick for owner-led stores

Agentmatica sits on your website and does the job a good person behind the counter does. It answers from your own pages and catalog, shows the products it means as cards, recommends what fits, and on Shopify and WooCommerce adds the choice to the cart and hands over a checkout link. The same agent handles shipping, returns and order questions, and cites the page it used when the answer comes from your site.

You stay in charge of how customers are treated. Write your pairing rules and your free-shipping line in Sales notes and the agent follows them when it suggests add-ons. Send it a private instruction in the middle of a live chat, take over with the agent paused, and hand the conversation back when you're done.

Pricing is simple. There's a free plan with 100 replies and no time limit, then $29 a month for 1,000 replies. AI replies are included in every plan, with no seats and no charge per resolution.

### How the others charge

Pricing models differ more than the headline numbers do. Intercom, Zendesk and Freshdesk charge per agent seat, then add AI on top: Intercom charges $0.99 per Fin outcome, Zendesk includes a small allowance of automated resolutions per agent and then charges $1.50 to $2.00 each, and Freshdesk gives 500 AI sessions once and then charges $49 per 100. Gorgias charges by ticket volume and adds $0.85 to $1.00 per automated interaction for its AI Agent.

The chat tools price differently again. Tidio's Starter plan is $29 a month, and its AI agent, Lyro, is a separate add-on from $39 for 50 AI conversations. Crisp's Mini plan is $45 a month with $5 of AI credits, which Crisp puts at roughly 90 AI conversations. Zipchat starts at $64 a month billed annually for 750 replies.

A reply, a conversation, a session and an outcome aren't the same unit, so take last month's chat volume and price it on each model. The busier your store gets, the more a per-resolution meter costs you, and that's the charge we left out.

### Zipchat

Zipchat is an AI shopping assistant that recommends products and recovers abandoned carts across web chat, WhatsApp, Instagram, Messenger and email. Plans run from $64 a month billed annually ($80 monthly) for 750 replies up to $960 a month for 14,000, with a 7-day trial and no free plan. We couldn't find a source link shown to the shopper at time of writing.

Choose Agentmatica if your shoppers ask on your website and you want to start free, see the page behind answers from your site, and coach or take over conversations yourself. Zipchat makes more sense if most of your conversations happen on WhatsApp or Instagram.

### Gorgias

Gorgias is a Shopify helpdesk. Its AI Agent includes a Shopping Assistant, on Shopify only, that shows product cards shoppers can add to the cart and can generate single-use discount codes. It also works on email and SMS. The helpdesk starts at $10 a month for 50 tickets, or $40 with AI Agent, and there's a 30-day trial. Your team can review the sources behind each AI response, but we couldn't find a source link shown to the shopper.

Choose Agentmatica if you're a smaller team, you sell on WooCommerce as well as or instead of Shopify, or you don't want to pay per resolved conversation. Gorgias makes more sense if you already have a support team working email, SMS and chat from one inbox with macros and SLAs.

### Tidio

Tidio combines live chat, ticketing and visual Flows with Lyro. With a paid Lyro quota, Lyro recommends products as cards and supports add to cart on Shopify, and Flows can send discount codes and recover abandoned carts. The free plan includes 50 Lyro conversations once, and after that Lyro is billed separately from the chat plan. We couldn't find a per-answer source link.

Choose Agentmatica if you want AI replies included in one plan price and a written set of pairing rules that the agent follows. Tidio makes more sense if you want chat, ticketing and social messaging in one tool and don't mind buying AI conversations as an add-on.

### Intercom

Intercom is a full customer service platform with Fin as its AI agent. Fin shows source links inline, and Fin for ecommerce lets shoppers browse products and add them to the cart. Seats cost $29 to $132 a month billed annually, plus $0.99 per Fin outcome, with a 14-day trial.

Choose Agentmatica if you're owner-led and want your website to sell and handle routine support without buying a helpdesk. Intercom makes more sense if you run a support team that needs a shared inbox, ticketing and Fin on email, phone and WhatsApp, and you can budget for seats plus outcomes.

### Zendesk

Zendesk is ticketing and customer service for support organisations. Support Team starts at $19 per agent a month billed annually, and Suite Team at $55 adds AI agents, messaging, live chat and voice. Generative replies can show a Sources link to the help articles used. Zendesk's retail page shows an AI agent offering a product with add to cart, but we couldn't find setup details or a Shopify catalog connection.

Choose Agentmatica if your customers mostly ask on the website before they buy and nobody is working a queue full time. Zendesk makes more sense if you have agents on SLAs with routing across email, messaging and voice.

### Freshdesk

Freshdesk is a helpdesk priced per agent, from $19 a month billed annually, with a 14-day trial and no free plan on its pricing page. Freddy AI Agent can show source links when references are turned on. We couldn't find product recommendations or add to cart in Freddy.

Choose Agentmatica if you want the chat to recommend products and build the cart on Shopify or WooCommerce, without paying for seats. Freshdesk makes more sense for a support team that needs ticket routing, SLA policies and cross-channel reporting.

### Crisp

Crisp is a shared inbox for chat, email, WhatsApp, Instagram and SMS, with an AI agent called Hugo on paid plans. The Free plan has no AI credits, and Mini is $45 a month with about 90 AI conversations of credit. When credits run out, Hugo hands chats to your team unless pay-as-you-go is on. We couldn't find product recommendations or a per-answer source link, and its Shopify app focuses on order history and refunds.

Choose Agentmatica if you want your website chat to recommend products and build the cart. Crisp makes more sense if several people share one inbox across many channels.

### When another tool is the better fit

Agentmatica isn't the right fit if most of your conversations happen on WhatsApp, Instagram, email or SMS, or you need ticketing, discount codes or abandoned-cart campaigns. A helpdesk suits a team that lives in tickets, and some of them now cite sources and sell products too. Product cards, add to cart and checkout links need a Shopify or WooCommerce store.

For everyone else, the owner-led store where shoppers ask before they buy, Agentmatica is the one built for that job. Connect your site, try it in your dashboard, and see how it answers your own customers' questions before it goes live.

---

## Which one should I buy? The question your product pages leave open

Source: https://agentmatica.com/blog/which-one-should-i-buy
Published: 2026-09-28 · Author: Robin · Category: Growth

If you sell three grinders, four sizes of the same jacket, or a starter kit next to a pro kit, a lot of your product page visitors are asking one question: which one is right for me? Hardly any of them email you about it. They flip between tabs for a while, then some buy the cheaper one to be safe and some buy nothing.

It's a harder question than shipping. Shipping has one right answer sitting on a policy page. Which-one depends on the shopper: what they'll use it for, what they already own, what they want to spend. A static page can't ask them any of that.

### What the jam study does and doesn't tell you

The study everyone quotes is Iyengar and Lepper's jam table from 2000. With 24 jams on display about 3 percent of passers-by bought one, and with 6 jams about 30 percent did. Later work made the story messier. A 2010 meta-analysis by Scheibehenne and colleagues found the average effect across studies was close to zero, with big swings depending on the setup.

A 2015 meta-analysis by Chernev and colleagues did find conditions where it bites, mainly when the options are hard to compare and the shopper isn't sure what they want. That describes plenty of product pages. You can usually keep every product and help the shopper narrow the range instead.

### Fix the page first

Start with what a torn shopper sees. Put a one-line who-it's-for near the top of each product in a range, because "for one or two cups a day" tells them more than a spec list does. Where two products really are close, add a small comparison table with the three or four rows that differ, and leave out the twenty that don't.

Then name the default. If most people should buy the middle option, say so on the page and say who should pick the others, the way someone behind a counter would.

### Where the agent takes the order

A page only gets you so far, because the missing input is the shopper's own situation. Someone on a shop floor solves that with one question back, like what are you making or what do you have at home. Agentmatica does the same on your product pages.

A shopper types "which of these is better for espresso?" and the agent answers from your product pages and catalog, shows the two or three that fit as product cards, and says in a sentence why each one suits. If the shopper is still torn, it can ask what they brew and how often, then point them to one. On a connected Shopify or WooCommerce store it puts the one they pick in the cart with the right variant and hands over a checkout link.

Your Sales notes shape the rest. If they say the grinder usually goes out with a particular scale, the agent mentions the scale when it fits the order, so the recommendation follows your own rules as well as your product descriptions.

### Read what they asked

After a couple of weeks the conversations show you which pairs of products people can't tell apart. That's your list of comparison tables to write and product descriptions that need more work. Fix the top few, and the page and the agent's answers both get better on the next crawl.

---

## How to raise average order value without another popup

Source: https://agentmatica.com/blog/raise-average-order-value-without-popups
Published: 2026-09-28 · Author: Robin · Category: Growth

Most advice on average order value ends in a widget: a frequently-bought-together strip, a cart drawer upsell, a spin-to-win that turns into a discount. All of them guess. The widget knows what's in the cart and has no idea why it's there.

Someone buying a coffee grinder as a gift wants different add-ons from someone replacing a broken one. The gift buyer might happily take a bag of beans and a gift note. The replacement buyer already owns everything and reads the suggestion as noise.

### The add-on that gets accepted answers a question

The suggestions shoppers take are usually the ones they were about to ask about. Do I need a filter for this? Will this cable work with my laptop? Is there a case that fits? When the add-on is the answer to their question, it doesn't read as selling.

So timing matters more than placement. A shopper asking about compatibility is a far better moment to mention the adapter than a popup on page load, because they've just told you what they're worried about.

### Use your free-shipping line

If you have a free-shipping threshold, it may already be your best order-value tool. Shoppers a few pounds short will often add something to clear it, and they'd rather be told what than go hunting. A progress bar in the cart helps. A sentence that names something that makes sense with their order helps more: "Add the 250g sampler and shipping's free."

### Add-ons that answer the question

A shopper on your grinder page asks, "will this do espresso, or do I need anything else?" Agentmatica answers from your product page, and if your Sales notes say espresso buyers need a tamper, it shows the tamper as a product card with one line on why. When they say yes, on a connected Shopify or WooCommerce store it adds the tamper to the cart and hands over a checkout link.

You write those rules in Sales notes, in plain sentences: what goes with what, where your free-shipping line sits, anything it should never push. The agent uses them whenever a shopper asks for advice, so the agent mentions the 250g sampler when the order is a few pounds short.

On a connected Shopify or WooCommerce store you can also turn on the product-page suggestion. The agent picks one complementary item from your catalog, following your pairing rules if you've written them, and offers it once per conversation. If the chat is closed it shows as a small bubble beside the launcher. It's off by default, so you choose when it starts. Turn it on and read the first week of conversations to see how the pairings land with your catalog.

The chat can also open a few seconds after a cart add with a relevant suggestion, or when a shopper lingers on checkout without moving on.

### Leave the checkout offers at checkout

Some order-value work belongs in checkout itself, like a free gift over a threshold or an upsell block on the checkout page. Those sit fine alongside a conversation. On Shopify, our sister app Smart Checkout Widgets handles that side.

### How to tell it's working

Read the conversations where the agent suggested something and ask two things: did the shopper take it, and did the suggestion fit what they were buying? Then watch items per order over a few weeks.

If shoppers keep passing on a suggestion, the pairing is probably wrong, and you can change it in Sales notes in a minute.

---

## The Black Friday support surge: where is my order, sizing and returns, before they reach your inbox

Source: https://agentmatica.com/blog/black-friday-support-surge
Published: 2026-09-23 · Author: Robin · Category: Ecommerce

Your busiest sales day is also your busiest support day, and on a small team they land on the same people. Shopify says its merchants took $14.6 billion over the 2025 Black Friday Cyber Monday weekend, up 27 percent on the year before, and that more than 94,900 of them had their highest-selling day ever. Few small stores have a support team sized for their best day. They have the owner and whoever is free.

The questions are the ones you get every week, arriving all at once while you're also watching stock, fixing a discount code and chasing the courier. Black Friday 2026 is on 27 November and Cyber Monday on the 30th. Most of what decides how that week goes for your customers can be done in October.

### Which questions arrive, and when

The week before the sale is mostly pre-purchase. Is the discount starting on Thursday or Friday, and in which time zone? Will the price drop further on Cyber Monday? Does this code stack with the one from the newsletter? What size should I get? These shoppers haven't bought yet, and a slow answer usually costs the order.

From Thursday night through Saturday it shifts to problems at checkout. The code doesn't work, or the size sold out between the product page and the cart. Then comes the post-order wave on Sunday and Monday: did my order go through, can I change the address, can I add something to it, can I cancel the one I placed by mistake.

Order-status questions start in the middle of that and often keep climbing for a week or more after the sale ends, as parcels queue at the carriers. That part gets its own post, linked below, because it's the one that catches people off guard.

### Publish the answers before the sale starts

Most of those questions have a factual answer that already exists in your head or in a supplier email. Put it on the site, in plain words, before the first shopper asks.

Delivery cut-offs come first. Write a page that says the last order date for arrival before the holidays, per region and per shipping method, and link it from the cart and the shipping policy. Carriers usually publish their holiday dates in the autumn, so build yours from theirs and leave a day of margin. If you sell abroad, give international orders their own dates and say who pays duties; Shopify's 2025 recap put cross-border orders at 16 percent of all orders over that weekend.

The returns window for gifts is next. Something bought on 27 November and opened on 25 December is outside a standard 30-day window, and the buyer knows it. Decide whether you extend the window for holiday orders, when the clock starts, whether a gift recipient can return without the buyer, and who gets the refund. Then write all of it down. The NRF's 2025 returns report, done with Happy Returns, found 82 percent of shoppers call free returns a major consideration when they buy, so expect this page to be read before checkout as well as after.

Sizing matters most for apparel and anything that has to fit something else. Put fit notes on the product pages that get the most size questions, say which way to go between sizes, and make sure the size guide link works on mobile. A size question that goes unanswered during a sale tends to become either an abandoned cart or a return.

Last, the sale terms. Start and end times with the time zone, which products are excluded, whether codes stack, what happens if the price drops after someone buys, and whether gift cards count. Discount-code questions come up a lot while the sale is running, and each one has a one-sentence answer if you've decided it in advance.

### Tell people how long a reply will take

A customer who knows a reply will take a day will usually wait a day. One who hears nothing tends to write again, and then again. Put a response time for sale week on your contact page and in your email auto-reply, and pick a number you can keep on your worst day. 'We answer within one working day during the sale' beats 'we'll get back to you shortly' followed by three days of nothing.

If you're going to be offline for part of the weekend, say so in the same place, and say what happens to messages in the meantime. People tend not to write twice once they know their first message has been seen.

### What should always reach a person

Some conversations should never be answered from a page, however good the page is. Decide which ones before the sale, and decide who picks them up.

Lost or damaged parcels are the obvious case, including 'it says delivered but it isn't here'. Those need a person, a photo and a decision about a replacement. Angry customers are another, because a correct answer in the wrong tone makes things worse, and what they usually want is to be heard by someone who can act. Refund exceptions belong on the list too: a return just outside the window, a price-drop claim, an order that arrived after the date you promised. Any case-by-case call that costs you money should be yours.

Add address changes after dispatch and large or wholesale orders if you get them. Write the list on one page, share it with anyone helping out over the weekend, and set up a notification that reaches you somewhere you'll actually see it.

### Sale week with Agentmatica

The pages you've just written are what Agentmatica answers from. It crawls your site and replies to shoppers in the chat, with a link to the page it used on each answer grounded in one. A shopper asks 'will this arrive before Christmas if I order on Saturday?' and gets your cut-off date with the delivery page linked, so they can check it and buy. When a page isn't there, it says it doesn't know and offers a way to reach you.

The pre-sale questions turn into orders in the same chat. Someone asks which size to get in the jacket they're looking at, and the agent answers from your fit notes, shows the jacket as a product card and, on a connected Shopify or WooCommerce store, puts the right size in the cart and hands over a checkout link.

On a Shopify store it also looks up an order once the visitor has identified themselves with their email or order number, which covers the 'did my order go through' and 'has it shipped' questions on Sunday and Monday. Refund decisions stay with you. Refund requests, complaints, an explicit request for a person and conversations where the visitor's mood turns are escalation triggers, and you get notified by Slack, Telegram or push, depending on what you've connected.

You can watch conversations as they happen from the dashboard. If something changes mid-sale, such as a code that's been switched off or a product that's delayed, add a private directive and the agent follows it in its next replies, or take the chat over yourself and hand it back when you're done. The settings also have a field for what the agent tells people when nobody can reply right away, which is where the response time from the section above goes.

Check your reply allowance before the sale. The free plan covers 100 replies a month, which a busy sale weekend can use up; paid plans start at $29 a month for 1,000 replies, and Surge Protection is an optional switch on paid plans that keeps the agent answering past the limit at $5 per 100 extra replies, capped at your plan's monthly price. Then set it up in October. Try it on the dashboard's Test your chat page before it goes live, and use a few quiet weeks to read the questions it couldn't answer and fill those gaps on your site.

---

## The week after Black Friday: order-status questions, returns and gift exchanges

Source: https://agentmatica.com/blog/after-black-friday-returns-wismo-week
Published: 2026-09-23 · Author: Robin · Category: Ecommerce

The sale ends on Cyber Monday, and the support load usually keeps climbing for another week or more. Orders placed between Thursday and Monday are now sitting in carrier networks that are carrying some of their heaviest volume of the year, and the customers who bought on Friday start asking where their parcel is around Wednesday. At the same time the first returns arrive, and the gift questions begin.

Sales go quiet while support gets busier, and it's the week most small teams plan for least. A few decisions made now take a lot of the pressure off.

### Why order-status questions peak after the sale

Tracking pages often go quiet during peak. A label gets created on Friday, the parcel is collected on Tuesday, and between the two the tracking says nothing useful. To a customer, 'label created' four days after ordering looks like nobody has packed it. Scans get skipped at busy depots too, so a parcel can appear to sit still for two days while it's actually moving.

Two things can reduce the load before it starts. Send a second shipping email once the carrier actually has the parcel, and say plainly that tracking can lag during the holiday period. And update your delivery page with your real dispatch times for that week. If you're running three days behind, say three days, because a delay customers were told about usually produces fewer emails than one they discover.

The questions that still come in split into two kinds. Plain status questions ('has it shipped, when will it arrive') can be answered from the order record and the carrier status. Everything else needs a person: delivered but not received, a parcel stuck for a week, a gift that won't arrive in time. The earlier post on order-status tickets goes through that split in detail.

### Returns start early, and gifts make them messy

The NRF's 2025 returns report, produced with Happy Returns, has retailers expecting 17 percent of holiday sales to come back, and 19.3 percent of online sales across the year. The first wave is people returning what they bought for themselves, often the wrong size, or two sizes of the same thing ordered to keep one. That wave usually starts within days of delivery.

Gift returns come later and raise questions your normal returns page probably doesn't answer. Can the recipient return it without the buyer? Does the refund go to the buyer's card or to the recipient as store credit? Does the extended holiday window start on the order date or the delivery date? Can someone return an item that came with a free gift, and do they have to send the gift back? If you ran a gift-with-purchase offer during the sale, that last one will come up, so decide it now and write it on the returns page.

Answer the refund amount question too. A customer who bought at a sale price normally gets back what they paid, and people do ask. The same goes for items bought in a bundle or with an order-level discount, where the refund for one item isn't obvious from the receipt.

### Exchange or refund: decide before the first request

An exchange keeps the revenue, and for a size problem it's usually what the customer wants. The same NRF report found 76 percent of shoppers are more likely to choose a return option that gives them an instant refund or exchange. The catch after a big sale is stock. The size they need may have sold out on Friday, and an exchange you can't fulfil turns into a refund plus an annoyed customer.

So write the rule before the requests come in. Offer an exchange first when the size is in stock, store credit as the next option if you want to keep the sale, and a refund when neither works or the customer asks for one. Say how long each takes. Whatever the rule is, make sure everyone helping you applies it the same way. Two customers who got different answers will find out, and one of them will write in about it.

### What still goes to a person

Lost and damaged parcels, returns outside the window, refund disputes and anyone threatening a chargeback need a human decision. So does a customer who has written more than once about the same order, because by the second message they're usually more annoyed with the store than with the parcel. Keep those on the escalation list from sale week and keep the notification going to someone who will act on it the same day.

### Agentmatica in the week after

A customer writes, 'I ordered on Friday and it still says label created, has it shipped?' On a Shopify store, Agentmatica looks the order up in the chat once they give their email or order number, and answers from the order record in plain language.

Returns and exchange questions it answers from your returns and shipping pages, with a link to the page used. A gift recipient asking whether they can return without the buyer gets the rule you wrote, with the page linked. When your pages don't cover something, it says so and offers a way to reach you.

Refund requests, complaints and requests for a person are escalated to you, and you can read the conversation, add a private directive or take it over and hand it back. The Insights page lists conversations that need attention, including ones where the agent had no source to answer from. That's a working list of what to add to your returns page before the gift returns arrive in late December and January.

### Measuring what the week cost you

Skip the resolution percentage and measure this week from your own inbox, against an ordinary November week. How many order-status messages did you answer by hand each day? How many customers wrote twice about the same order? How many hours did you or your team spend on support instead of the business? How many answers had to be corrected after they were sent?

Then tag the returns by reason as they come in. If one product line's returns are mostly sizing, that's a size guide to fix before next year, and if they're mostly damage, it's packaging. The questions nobody could answer from your site are the pages to write. Put all of it in one note with the dates, because next October you'll want to know when the order-status questions peaked and how long they took to fall back.

---

## "Where is my order?" How to stop answering it yourself

Source: https://agentmatica.com/blog/where-is-my-order-tickets
Published: 2026-09-08 · Author: Robin · Category: Ecommerce

'Where is my order?' is somewhere between 20 and 40 percent of ecommerce support volume, depending on which survey you read, and more than half of it in peak season. Every one of those tickets runs the same way: find the order, open the carrier page, paste the status back to the customer, close the ticket, wait for the next one.

It's high-volume work with no judgement in it, and it makes the customer wait for information that already exists in two systems they can't see. Your team can answer it for years and get no better at their jobs. The way out is to change where the question lands.

### Why the tracking page doesn't fix it

Most stores already put a tracking link in the shipping email, and the tickets still come. The customer has usually seen the tracking number. What they want to know is whether it will be there by Friday, why it says delivered when nothing has arrived, or why only one of the two things they ordered has shipped. A tracking page answers none of that. A person does, which is why the email gets sent.

Look at your own inbox and count the WISMO tickets that open with the customer pasting the tracking number. They have seen the page and want someone to interpret it.

### How Agentmatica answers the order question

Agentmatica takes the order question before it becomes a ticket. A shopper types 'where's my order', the agent asks for the email or order number, looks the order up in Shopify, and answers in plain language: what shipped, what didn't, the carrier status and the delivery estimate. When the answer is 'it's marked delivered but you don't have it', the agent hands the chat to you once you've set that as an escalation rule, because that one needs a human.

Two rules keep this safe, and both are built in. Agentmatica only reads order data after the visitor has identified themselves, and it only says what it can source: the order record for status, and your shipping policy page for the 'what if it's late' part, with that page cited under the reply so the customer can check. It has nothing to improvise a delivery promise from, and when your pages don't cover something, it says it doesn't know.

The same agent keeps selling once the parcel question is settled. A shopper who asks what goes with the thing they just bought gets the matching products as cards and can buy them in the same chat.

### Three WISMO questions that need a rule

Once the plain status questions stop reaching you, what's left is more interesting, and it's where you set the agent's limits. 'It says delivered but I don't have it' is a claim about a missing parcel, and the right move is a person and a photo of the doorstep. 'Can I change the address?' depends on whether the order has left the warehouse, which the agent can read, and on whether you're willing to pay the courier's redirect fee, which is your call. 'Only half of it arrived' is a split shipment nine times out of ten, and the order record shows that.

Write those three down as rules before you switch anything on. Delivered-but-missing always goes to a person. Address changes are answered from the fulfilment status and then handed over. Partial deliveries are answered from the order record. Agentmatica follows escalation rules and private directives you write in the dashboard, which is the difference between an agent that answers the easy questions and one that also mishandles the hard ones.

### What changes the week after

Stores that route WISMO to an agent usually see the category shrink to a handful of genuine exceptions, the lost parcels and wrong addresses and split shipments that went sideways. Those get an escalation rule and a notification in Slack or Telegram, and the person who picks them up has time to care about them. They take over the live chat, sort it out, and hand it back to the agent.

The second effect takes longer to notice. Every unanswered question the agent logs is a gap in your content. If people keep asking whether you ship to Ireland, that's a page you're missing, and once you write it the next crawl picks it up. The support queue starts working as a to-do list for your website.

### How to know whether it worked

Skip the resolution percentage. Every vendor publishes one and none of them tell you whether your customers were treated well. Measure the things you can see in your own inbox: how many order-status emails you answered by hand this week against the week before, how many customers wrote twice about the same order, and how many answers you had to correct after the fact. If the first two fall and the third stays near zero, it's working. If the third climbs, read the corrections; they tell you which page on your site is wrong or missing.

Give it two weeks and one busy weekend, because the Sunday night after a promotion tells you far more than a quiet Tuesday.

### What it costs against a helpdesk seat

A helpdesk with an AI add-on is usually priced per seat or per resolution, so the cost grows with the very tickets you're trying to make disappear. Agentmatica's free plan covers 100 replies a month with no card and no time limit, which is enough to start handling order-status questions on a small store; paid plans start at $29 a month for 1,000 replies. If you're comparing against a dedicated ecommerce chat tool, the Zipchat comparison below lays the plans side by side, including the overage fees that decide the real monthly bill.

---

## "The chatbot said it, not us." Two courts have now said: yes, you.

Source: https://agentmatica.com/blog/the-chatbot-said-it-not-us
Published: 2026-09-01 · Author: Robin · Category: AI accuracy

Whatever the AI agent on your site promises, you've promised. In 2024 a Canadian tribunal ordered Air Canada to honour a bereavement discount its chatbot had invented. The airline argued that the bot was 'a separate legal entity responsible for its own actions'. The tribunal called that remarkable and ruled against it. In May 2026 a German appeals court reached the same conclusion in a different dispute: the company publishes the answer, so the company owns the answer.

Small teams are putting AI agents on their sites under that rule, and they should keep doing it. What the agent is allowed to say is now a business decision, and the owner makes it.

### Where general-purpose bots go wrong

Independent measurements of live support bots put the wrong-answer rate somewhere between one in seven and one in four. That's what a general-purpose language model does when it's asked something it was never given: it produces the most plausible shape of an answer, in your brand's voice, with total confidence. Your returns window becomes 60 days because that's a common number. Your plan includes a feature because the sentence flowed.

A bot that's wrong 15 percent of the time on policy questions is writing 15 percent of your policy for you, and you find out which parts when a customer holds you to one.

### Three controls that change the risk

One: ground every answer in your own pages. The agent retrieves from your live site at question time and can only assert what it found. If retrieval comes back empty, the right output is 'I don't know, here's how to reach us'. That one rule takes invented policy off the table.

Two: show the receipt. A citation on each grounded answer lets the customer check and lets you audit. When something looks off, the citation points to the page that produced it. Nine times out of ten the page was ambiguous and needs a rewrite, so the fix is on your website, where you can make it.

Three: define what the agent must not decide. Refunds over a threshold, cancellations, anything with a legal edge: these get an escalation rule that pulls a human in, notifies them wherever they are, and lets them take over the live chat. The agent explains and the human commits. That split is effectively what the courts are asking for.

### All three controls, built in

Agentmatica ships with all three. A shopper asks whether they can return a jacket after 60 days. The agent reads your returns page, answers with the window that page states, and shows the page under the reply. If none of your pages cover it, the agent says it doesn't know and tells the shopper how to reach you.

A shopper asks for a refund on a large order. Your escalation rule pulls you in, you get the notification in Slack, Telegram or email, and you take over the live chat to make the call yourself. When it's done you hand the chat back and the agent carries on. Private directives set the lines it holds in between, such as never quoting a delivery date or always sending cancellations to you.

The same grounding is what makes it safe to let the agent sell. When a shopper asks which jacket suits wet weather, it recommends from your catalog, shows the options as product cards, and on Shopify or WooCommerce adds the one they pick to the cart with a checkout link.

### What a careful setup gets you

A grounded agent that cites its sources and declines when it has nothing to cite is more trustworthy than the average human reply. It answers from the written policy, it doesn't get tired at 11pm, and it leaves a permanent, searchable record of exactly what was said and why.

Start with your returns, refunds and shipping pages. Read them the way a shopper would, fix anything a stranger could misread, and decide which requests always come to you, because those pages and rules are what your agent will say.

---

## Many abandoned carts start with one unanswered question

Source: https://agentmatica.com/blog/shipping-cost-cart-abandonment
Published: 2026-08-25 · Author: Robin · Category: Growth

Your shoppers are using checkout to find out what shipping costs, and a lot of them leave when they see the number. Baymard's running meta-analysis of checkout studies puts the average cart abandonment rate at just over 70 percent. Ask those shoppers why and the top answer hasn't changed in a decade: extra costs they didn't see coming. Shipping, taxes and fees turn up on the final screen, and about 48 percent of abandonments trace back to that moment.

Treat that as a pricing problem and you'll start offering free shipping you can't afford. Treat it as an information problem and the fix is much cheaper. The shopper had a question, and the only way to get it answered was to go through checkout.

### The questions people don't email you about

Think about what's going through a shopper's head on a product page. Do you ship to me? What will it cost? When will it arrive? Can I return it if the fit's wrong? Is this the right size, or the right cable or version for my machine?

Almost none of these become support tickets. A ticket takes effort and a wait, and checkout is faster, so the shopper goes to checkout for the answer, reads the number and leaves. Part of your abandonment rate is a count of the questions your pages leave open.

### Where to answer each question

Shipping cost and delivery time go on the product page, next to the price. One line such as 'Free delivery over 60, otherwise 4.90, two to four days' covers most shoppers. If the rate changes by country, link to a shipping page that lists it country by country, and keep that page current. On Shopify, a free-shipping progress bar in the cart or checkout shows how far the shopper is from the threshold, which gives them a reason to add something instead of a reason to leave.

Returns belong in the same spot. Say how many days they have, who pays for the return label, and whether they get a refund or store credit. A shopper who isn't sure about the fit buys more easily when sending it back looks simple.

Sizing and compatibility need a size guide or spec table on the product page itself, where the shopper already is. If two of your products are easy to confuse, say who each one is for.

Payment questions come last and still lose orders. Whether you take PayPal, buy now pay later or cash on delivery should be visible in the cart, before the payment step. Once all of this is written down, watch which questions still come up. Those tell you what to write next.

### Answer shipping before checkout does

Put shipping information on the product page. That's the obvious fix and you should do it. Static copy still can't cover every combination of destination, weight and promotion, and shoppers rarely read the FAQ. Agentmatica sits on the product page and answers the specific question from your actual shipping policy, in one exchange, with the page it used cited under the reply: 'Delivery to Portugal is 6.90, three to five days, free over 60.'

The same agent reads your size guide for the sizing question and the product spec for the compatibility question. When a shopper asks which charger fits their model, it shows the ones that fit as product cards, and on Shopify or WooCommerce it adds the pick to the cart and hands over a checkout link, so the answer and the item arrive in the same step.

Your Sales notes tell it what goes together. If you pair a jacket with a waterproofing spray, the agent can suggest the spray on the jacket's product page, which is handy for a shopper a few euros short of free shipping.

### Measure it by the questions

Judge this by the questions shoppers ask, and leave the overall abandonment rate out of it. After a couple of weeks you'll have a ranked list of what they wanted to know before buying, which is customer research you'd otherwise pay for. Maybe people keep asking about shipping to a country you don't serve, or about sizing on one product line, or about a compatibility detail nobody thought to document. Each one is either a page to write or a decision to make.

Fix the top three in your content. The agent's answers improve on the next crawl, and so does the page for everyone who never opens the chat.

---

## Your AI chatbot isn't broken. Your website is out of date.

Source: https://agentmatica.com/blog/chatbot-wrong-because-website-stale
Published: 2026-08-18 · Author: Robin · Category: Support

When a grounded agent gives a wrong answer, the page it read is usually the thing that's wrong. Say a support lead sends over a screenshot: the bot told a customer the Starter plan includes API access. It doesn't, and it hasn't since March. The instinct is to blame the AI, tweak the prompt, maybe shop for a smarter model.

Then someone checks the citation and finds the sentence, word for word, on a pricing comparison page that was never updated after the March change. The agent had read that page and repeated it accurately, so the fix belonged on the website.

### Invented answers and faithful ones

There's invented-wrong, where the model fills a gap with something plausible. Grounding and citations fix that structurally, and we've written about it before. Then there's faithfully-wrong, where the agent accurately repeats content that's no longer true. No model upgrade touches the second kind, because the model did its job.

Faithfully-wrong comes up far more often than you'd expect, and it's good news. The agent is doing what you asked, and the fix is a page edit on your own site.

### An agent is only as fresh as its last crawl

Every grounded agent works from an index of your site taken at some point in time. If that index was built once at setup, it starts ageing the same afternoon. A price changes, or a policy gets softened, or a product is discontinued, and the agent keeps quoting the old world with full confidence and a valid citation.

So the feature to check for, boring as it sounds, is the re-crawl schedule. Your agent should re-index your site automatically, at least weekly, and whenever you ask it to after a big change. Update the shipping policy page on Tuesday and the answer should change by Wednesday, citation and all, without anyone touching the agent.

### Keeping the agent as current as your site

Agentmatica re-crawls your pages on a schedule, and the Re-crawl button in the dashboard refreshes the index straight after a big change. A shopper asks whether you still offer free returns. The agent answers from the returns page as it reads today and shows that page under the reply, so if the answer is out of date, you can see exactly which page said it.

When nothing on your site answers a question, the agent says it doesn't know and tells the shopper how to reach you. That conversation, along with any the agent escalated or a shopper marked with a thumbs-down, lands under Needs attention in Insights. Write or fix the page, the next crawl picks it up, and the same question gets a sourced answer.

### Turn the misses into a maintenance list

The other half of staying current is knowing what the agent couldn't answer. Read the Needs attention list weekly. It's a ranked set of pages that are missing, ambiguous, or contradicting each other.

After a month of this the website is clearer for humans too, because you've been fixing the exact sentences that confused a machine, and support volume drops because the answers are now where people look. Put the list on your Monday to-do and fix the top page each week.

---

## Why AI chatbots make things up, and how citations fix it

Source: https://agentmatica.com/blog/why-ai-chatbots-make-things-up
Published: 2026-07-10 · Author: Robin · Category: AI accuracy

An AI agent that invents your policies is worse than having no agent at all. In 2024, an airline's chatbot invented a bereavement refund policy, and a court ordered the airline to honor it. The bot did what language models do by default: it produced the most plausible-sounding answer, and the plausible answer was wrong.

If you're putting AI in front of your customers, plan for that exact risk: an agent that answers confidently and wrongly, in your brand's voice, about your own policies, and sounds helpful while it does it.

### Why models invent answers

Language models are trained to continue text convincingly. Ask one something outside what it knows and it fills the gap with the statistically likely shape of an answer. That's fine when you're brainstorming. It's a disaster when the question is about your returns policy.

Vendors often answer this with 'better prompts' or 'guardrails', which are instructions layered on top of the same mechanism. They make invention less frequent. They can't remove it, because the model's underlying objective, to always produce an answer, hasn't changed.

### Retrieval with receipts

The fix is in how the system is built, and it has three parts. Restrict the agent to answering from a specific body of content, retrieved at question time: your website, your docs, your policies. Require it to cite the passage behind each claim. And when retrieval comes back empty, make the correct output 'I don't know'.

Citations do two jobs at once. Your visitor gets an answer they can check by clicking through to the source. You get an auditable record of every conversation: when an answer looks off, the citation shows exactly which page produced it, and usually the page is what needs fixing.

### "I don't know" is a feature

Visitors trust an assistant more after they have seen it decline once, because the refusal proves the confident answers mean something.

Every refusal is also logged, which gives you a precise list of gaps in your content. Write the missing page, the next crawl picks it up, and the same question gets a cited answer the following week. The agent answers or logs the gap, and each round makes the website and the agent better.

### How Agentmatica answers a policy question

Agentmatica is built on this design. A shopper asks, 'Can I return something I bought in the sale?' The agent pulls the answer from your returns page, replies in plain words, and links the page it used so the shopper can read the policy for themselves. If your returns page doesn't mention sale items, the agent says it doesn't know and offers to bring in a person. Your team takes over the live chat, answers, and hands the conversation back to the agent when they're done.

You stay in charge of what it says. Private directives let you tell the agent how to handle topics you care about, and the gaps it logs show you which pages to write next. The same grounding carries into sales: when a shopper asks which product fits, the agent answers from your catalog and shows the matching products as cards.

---

## How small teams cut repeat support tickets without hiring

Source: https://agentmatica.com/blog/cut-support-tickets-without-hiring
Published: 2026-07-03 · Author: Robin · Category: Support

Most of your support queue is questions your website already answers. Pull up your last hundred conversations and sort them into two piles: 'the answer exists on our website' and 'a person needed to think about this.' For most small teams the first pile is the bigger one. Shipping status, return windows, sizing, stock questions, asked slightly differently every day and answered identically every day.

The fix for that pile is to answer it on the website before it turns into a ticket.

### Step one: find your repeat rate

Measure before you buy anything. Tag a week of incoming questions by topic. The distribution is always lopsided: a handful of topics generate most of the load. That's good news, because a system that handles the top ten topics well absorbs the majority of the volume.

The same exercise gives you your acceptance test. Whatever you put on the site should answer those top ten questions correctly, from your own content, before it's allowed in front of customers.

### Step two: answer at the point of intent

A ticket is a question that left the website. The visitor looked (or didn't), gave up, and emailed. Every hour they wait for a reply costs you twice: their frustration and your queue.

An on-site agent answers the question in the moment it comes up, on the page where it came up. Grounded retrieval is what makes that safe. The agent answers from your live content and cites the page, so the answers stay accurate as the ticket count drops. A visitor who got a wrong answer still emails you, later and angrier.

### Step three: make escalation excellent

Once the repeats stop reaching you, you have time for the real issues. Escalation rules catch the high-stakes topics (refunds over a threshold, cancellations, anything legal). Mood analysis flags frustrated visitors before they ask for a human. Alerts reach the right person in Slack, Telegram, email, or push, and anyone on the team can take over the live chat mid-conversation.

### Your top ten questions, answered on the page

Agentmatica reads your site and catalog and answers the repeat questions where shoppers ask them. A shopper on a product page asks, 'How long does delivery take, and can I send it back if it doesn't fit?' The agent answers both from your shipping and returns pages and links each page it used. When the content isn't there, it says it doesn't know, and the question lands in your list of gaps to fill.

Plenty of those support questions are buying questions in disguise. When the same shopper asks which size to get, the agent answers from the product details, shows the right product as a card, and on Shopify or WooCommerce adds it to the cart and hands over a checkout link.

For the second pile, a teammate takes over the live chat, sorts out the refund or the damaged parcel, and hands it back to the agent afterwards. Private directives let you set how the agent handles the topics you want kept close.

---

## Your website gets traffic at 2am. Here's how to capture it.

Source: https://agentmatica.com/blog/capture-leads-while-you-sleep
Published: 2026-06-24 · Author: Robin · Category: Growth

Many night visitors leave because nobody answers the one question standing between them and a purchase. Check your analytics for sessions between 10pm and 7am, plus weekends. For most businesses that's 30 to 40% of all traffic, and it arrives when nobody can answer a question. Those visitors want the same things the daytime ones want, and they get the same pages with no help.

The usual tools for this are bad trades. A contact form asks the visitor to do your team's work and then wait. An email address on the page gets you a message, maybe, with no context, and a 'leave a message' chat widget is the same form in a costume. Put an agent on the site that answers from your own pages, then asks for an email.

### Answer first, ask second

The order matters. A visitor who has just had a useful answer, cited from your actual content, is warm. One polite ask at that point ('want us to follow up by email?') converts because it comes across as service.

The popup that demands an email before the visitor has received anything makes the same request in the opposite context. The conversion is nowhere near the same, and neither is the impression it leaves of your brand.

### A lead with a transcript beats a lead with a name

A conversation gives you more than a form fill. Along with the contact details you get the whole exchange: what they asked, what they were comparing, what almost stopped them. Your morning follow-up can open on that instead of 'just checking in.'

Send those leads wherever your team already looks, whether that's Slack, Telegram, email, or a push notification, and the overnight queue is the first thing sales works through with their coffee.

### What Agentmatica does at 2am

Some night visitors want to buy there and then. A shopper asks at 2am, 'Which of these two jackets is warmer, and do you ship to Ireland?' Agentmatica answers the shipping part from your delivery page and links it, compares the two jackets from your catalog details, and shows both as product cards. The shopper picks one, and on Shopify or WooCommerce the agent adds it to the cart and hands over a checkout link.

If the shopper isn't ready yet, the agent asks for an email and saves the conversation with it. When the content isn't there, the agent says it doesn't know, and the saved conversation tells your team exactly what to follow up on in the morning.

### The compounding part

Every after-hours conversation also feeds the insight side. Questions the agent couldn't answer become content to-dos, and questions that keep coming up show which pages need to work harder. The agent catches the leads the site was dropping, and the transcripts tell you why it was dropping them.

---

## AI agent vs. live chat: what your website actually needs

Source: https://agentmatica.com/blog/ai-agent-vs-live-chat
Published: 2026-06-12 · Author: Robin · Category: Strategy

Most websites need an agent on the first conversation and a person ready to take over. The three options usually get lumped together, so it helps to separate them. Live chat is humans typing during business hours, with a queue when it gets busy. Scripted chatbots are decision trees that handle exactly the questions you predicted, phrased the way you predicted them. AI agents are models that understand the question and, built properly, answer from your content.

### How the three compare

Live chat wins on empathy and on judgment calls, and loses on coverage (nights, weekends, spikes) and on cost per conversation. Scripted bots win on predictability and lose the moment a visitor goes off script, which happens constantly. AI agents win on coverage and on the long tail of questions. Their known weakness is that an ungrounded model improvises.

That weakness has a fix. Restrict answers to your indexed content, cite the source, and say so when the content doesn't cover the question. Coverage can't be fixed for humans without adding payroll, and the off-script cliff can't be fixed for decision trees at all.

### Agent first, human always available

The setup that works in practice: the agent takes every first conversation, answers the routine questions with cited replies, and escalates on triggers you define, such as a topic rule, signs the visitor is frustrated, or an explicit 'talk to a human.' Your team watches live and can take over any chat partway through, with the transcript as context and the widget telling the visitor a teammate has joined.

Visitors get an answer straight away, including at hours when nobody on your team is online, and a person when it matters. Your team sees the conversations that need their judgment, and nobody has to cover a Tuesday 3am shift.

### The first conversation, start to finish

A shopper on a running shoe store's site asks, 'Which of these works for wide feet, and how long does delivery take to Leeds?' Agentmatica answers the delivery half from your shipping page and links the page it used. For the fit half it searches your catalog, shows the matching shoes as product cards, and says why each one fits. When the shopper picks one, it adds that size to the cart and hands over a checkout link on Shopify or WooCommerce.

If the shopper then asks for a custom order your pages don't cover, the agent says it doesn't know and your teammate takes the chat live from the dashboard. When they're done, they hand the conversation back and the agent carries on. Private directives let you set the rules it follows along the way, such as which questions always go to a person.

### The migration path

If you're on live chat today, keep it and put the agent in front of it, then watch your first-touch volume drop. If you're on a scripted bot the change is bigger: delete the flow builder, index the site, and let retrieval handle the phrasing you never predicted.

---

## The Shopify owner's guide to an AI agent that actually sells

Source: https://agentmatica.com/blog/shopify-ai-sales-agent-guide
Published: 2026-06-02 · Author: Robin · Category: Ecommerce

An AI agent on a Shopify store earns its place when it answers from the live catalog and current policies, checks real order status, and sells only what you stock. Store chat has one specific failure mode: the enthusiastic bot that recommends a product you discontinued, quotes a shipping time from an old page, or agrees to a discount that doesn't exist. In ecommerce a made-up answer costs you money one order at a time, so grounding comes first.

### Ground it in your store's own pages

Use your product pages, collection pages, and shipping and returns policies as the knowledge base. They get indexed and recrawled, so a price change or a policy update reaches the agent's answers on its own. The agent cites the page it used, which is how the visitor learns to trust it: 'ships free over $40' links to the shipping page that says so.

Connect Shopify and the agent gets a live layer on top of that: order lookups and product data inside the conversation. 'Where's my order?' gets answered right there in the chat.

### Selling like a good floor associate

When a visitor asks about espresso, the agent searches your real catalog, shows your espresso gear as product cards, and suggests the follow-up questions buyers actually have. Because it works from the catalog, it quotes your real products and prices, and an off-catalog request gets a plain 'we don't carry that.'

The most useful sales work happens at the moment of hesitation. 'Will this fit a 15-inch laptop?' asked on a product page is a buying signal with one blocker. Answering it on the spot, with the spec cited, is probably the highest-ROI sales work on your site.

### How Agentmatica sells on Shopify

A shopper asks, 'Which grinder should I get for espresso at home?' Agentmatica checks your catalog, shows the two grinders that suit espresso as cards, and explains the difference between them in a sentence each. The shopper picks one, and the agent adds it to the cart and hands over a Shopify checkout link without the shopper leaving the chat.

You decide what it pairs. Sales notes tell the agent which products go together, so with the product-page suggestion turned on, it can suggest the item you've paired with the grinder. Private directives set the rest of the house rules, and answers about shipping or returns link the page they came from.

### When the order is big, bring a human

Set escalation rules for bulk orders, custom requests, and high-value carts. The agent keeps gathering details (quantity, timeline, contact) while your team gets pinged in Slack or Telegram, and anyone can take over the chat live and hand it back when they're done. The visitor sees one continuous conversation.

Setup is one script tag before </body>, or follow the Shopify guide if you want the agent to take actions like order lookups.

---

## How to get your site cited by ChatGPT, Claude, and Perplexity

Source: https://agentmatica.com/blog/get-your-site-cited-by-ai
Published: 2026-05-20 · Author: Robin · Category: AI visibility

If AI assistants can't read your site, they won't cite it. A growing share of buying research now happens inside these assistants. Someone asks ChatGPT 'best AI chat widget for a small store' and gets a synthesized answer with a handful of citations. If your site isn't legible to these systems, you're missing from that answer entirely.

The work is mostly things worth doing anyway: clear structure, a plain description of what you do, and content a crawler can reach.

### Ship an llms.txt

llms.txt is a plain markdown map of your site written for machines: what you do, the facts that matter, and annotated links to your important pages. When an assistant fetches your site, it gets your framing in one request and skips the guesswork of reading navigation menus.

Keep it current and accurate, and link it from your footer. Assistants find it faster that way, and so do the people testing them.

### Let in the crawlers you want

Audit your robots.txt. Plenty of sites block GPTBot, ClaudeBot, and PerplexityBot by default. That's a reasonable stance for a publisher who makes money from content. For a product company that wants to be recommended, it's a self-inflicted wound. Allow the assistants' crawlers on your marketing and docs pages, and keep dashboards and APIs disallowed.

### Write answer-shaped pages

Assistants extract and cite passages. Pages that state facts plainly get quoted: pricing with numbers, integration lists, fair comparisons, FAQs with real questions as headings. Pages that are all adjectives get skipped.

The same discipline makes an on-site AI agent good. If your own agent can answer a question from your content with a citation, an external assistant probably can too.

### Use Agentmatica to test your pages every day

Agentmatica answers your shoppers from the same pages you want ChatGPT to quote, so every conversation checks how legible they are. A shopper asks, 'Do you ship to Canada, and what does it cost?' If your shipping page says so, the agent answers and links that page. If it doesn't, the agent says it doesn't know and your team can take the chat over live.

Those gaps show up under Needs attention in Insights as the questions your content can't answer yet. Add the missing fact to the page, Agentmatica recrawls it, and the next shopper gets a cited answer. The page an outside assistant reads is now better too.