Your store's chatbot shouldn't invent your return policy.

I build support chatbots that answer only from your own documents — with a source link on every answer, and an honest "I don't know" plus a human handoff when they're unsure. Self-hosted in the EU. Your customer data stays yours.

Watch the 90-second demo Email me your store URL

No calls needed. I reply within 24 hours with a straight answer.

Real system, not a mockup: cited answer, honest refusal, human handoff. 68 seconds.

The problem

Your customers ask the same five questions forever: where's my order, how long is shipping, can I return this, does it fit, do you ship here. Off-the-shelf bots answer them by guessing — confidently, and sometimes wrong. A hallucinated shipping rate or invented return window costs you more than the ticket ever did.

What this is instead

Trained on your docs, not the internet.

Your shipping policy, returns, FAQ, sizing guide, product pages — chunked, indexed, and used as the only source of truth.

Every answer cites its source.

Customers see where it came from. So do you.

It refuses to guess.

When the answer isn't in your documents, the bot says so and hands the conversation to a human — instantly, with the question, to your inbox or Slack. That refusal path is the whole product.

It tells me when it breaks.

Error alerting on every workflow, so a silent failure never runs for a week unnoticed.

It runs where you choose.

Self-hosted, EU hosting available. No customer conversation leaves your infrastructure. If data residency has ever come up in your business, this is the version of AI support that survives that conversation.

Honest note: if you get fewer than ~100 support tickets a month and only need order tracking, you don't need me — a $30/month app will serve you fine, and I'll tell you which one. This is for stores where the same policy, shipping and sizing questions are eating hours every week, or where customer data can't sit on a US server.

Packages

Starter — $150

Chatbot trained on up to 50 pages of your policies and FAQs. Website widget. Cited answers. Honest refusal fallback. 1 revision. 7-day delivery.

Standard — $450most stores want this

Everything in Starter, plus product catalog sync, human handoff to your inbox or Slack, error monitoring with alerts, and a tuning pass after two weeks of real customer questions. 3 revisions. 10-day delivery. 30 days of support.

Premium — $1,200

Everything in Standard, plus live order-status lookup, multi-language, a monthly answer-quality report, 90 days of support, and an optimization review.

Keep it accurate — $99/month

Optional. I re-sync your docs whenever your policies change, monitor for errors, and send you a monthly report of what customers asked and what the bot deflected. Cancel anytime.

How it works (four steps, one week)

  1. You send: your store URL, the pages the bot should learn, and where handoffs should go.
  2. I build: ingestion, retrieval, citations, refusal path, handoff, monitoring.
  3. You test: ask it your hardest questions. We tune what it gets wrong.
  4. It goes live on your store, and I hand over documentation for everything.

FAQ

Will it make things up?
No. It answers only from your documents, cites the source, and refuses plus hands off when unsure. That's enforced in two places — a confidence gate on retrieval and a second check on the answer itself.
Where does my data live?
On infrastructure you control; EU hosting available. Your documents and conversations are never used to train public models.
What if it breaks?
Every workflow has error alerting. I find out before you do — that's what the monthly plan buys.
Which AI does it use?
Model-flexible (OpenAI, Claude, others). You're never locked to one vendor's pricing or roadmap.
Do I need technical skills?
No. You send links; I hand over a working widget and plain-language documentation.

Stop answering the same five questions.

Email me your store URL

I'll tell you straight whether a custom bot is worth it for your volume, or whether an app would serve you better.