Service

AI support agents trained on your business

An AI support agent answers your customers' routine questions from your own documents, policies and tone of voice. Anything it is unsure about goes to a person rather than being guessed at. What separates a useful agent from an embarrassing one is where it looks things up, what it is allowed to say, and when it hands over. That is where most of the build time goes.

How the process works

Sound familiar?

The signs this is your problem

If two or more of these are true, there is almost certainly a worthwhile automation hiding in it.

  • The same twenty questions arrive every week in slightly different words
  • Response times slip at weekends and out of hours
  • Support quality depends on which person picks up the ticket
  • Your team answers order-status and policy questions instead of doing real work
  • You have good documentation that nobody reads

What gets built

What an agent build includes

Four things, in roughly this order. The proportions shift by project — the shape does not.

Retrieval over your real knowledge

The agent answers from your help centre, policy documents, product data and past tickets — not from the model's general knowledge. Every answer can carry a citation back to the source document it came from.

Guardrails and refusal rules

Clear limits on what the agent may say. It does not invent policy. It does not quote a price it cannot check. It does not promise a refund or a delivery date unless that rule is written down. Anything outside those limits goes to a person instead of being improvised.

Clean human handoff

When it hands over, the person picking it up sees the whole conversation, what the agent already checked, and why it stopped. Not a blank ticket the customer has to explain from scratch.

Evaluation before it goes near a customer

The agent is tested against a set of real historical questions with known-correct answers, and I show you the pass rate and the failure cases before anything is switched on.

What you get

Handed over, documented, yours.

  • A working agent on the channels you choose — web widget, email, or WhatsApp
  • An evaluation report showing measured accuracy on your real questions
  • Documented escalation rules and refusal boundaries
  • A knowledge-gap report: the questions the agent could not answer, so you can fix the docs
  • Full transcripts retained in a system you control

Typically built with

OpenAIAnthropic ClaudeGoogle Geminin8nPythonWhatsApp Business API

Chosen per project, not by habit. Everything runs in accounts registered to you — you hold the keys and see the bills directly.

When this is not the right service

I do not build agents that pretend to be human. Every agent I ship identifies itself as automated on first contact — that is both a trust decision and, increasingly, a legal requirement.

How we'd work

From audit to autopilot in weeks.

Fixed scope. One fixed price. Working software every week. You always know what ships next and what it saves.

  1. Week 0 · Free

    Audit

    Thirty minutes on where your hours actually go. I map the repetitive workflows, time them, and score each one by what automating it would return. You keep that map either way — including the honest note on which processes to leave alone.

  2. Week 1

    Blueprint

    A written proposal: what gets built, which systems it touches, what it costs to run each month, the timeline, and one fixed price for the whole engagement. Nothing starts until you approve it. The price does not move unless you change the scope.

  3. Weeks 2–4

    Build

    I design, build and test inside your own accounts, with evaluations and guardrails on anything using a language model. You see working software every week rather than a status update — which means you can redirect early, while redirecting is still cheap.

  4. Ongoing

    Run & improve

    Launch, monitor, iterate. Thirty days of support included with every build. After that: take a care plan, hand it to your own team using the documentation, or run it yourself. All three are genuinely fine.

Questions

AI support & sales agents — the usual questions

It can, which is exactly why the build is designed to prevent it. The agent answers from retrieved passages of your own documents rather than from general model knowledge, refuses questions it has no source for, and escalates instead of guessing. Before launch I test it against real historical questions with known answers and show you the measured failure rate rather than asking you to take it on trust.

Yes, always. Every agent I build says it is automated at the start and offers a way through to a person. It is the right thing to do, and it is also the safe thing: the EU AI Act and the FTC both take the view that hiding it is deceptive.

No. I use the business accounts from OpenAI, Anthropic and Google, where what you send is not used to train their models. Your documents stay in storage you control, and the exact terms of whichever provider we use are written into your project documentation.

It hands over to a person, with everything attached. They see the whole conversation, what the agent looked at, and why it stopped — so the customer never repeats themselves. Every handover is logged too, and that log becomes a list of the gaps in your documentation worth filling.

Ready to get the hours back?

A free 30-minute audit call. You leave with a written map of your automatable workflows and what each one is worth — whether we end up working together or not.

Replies within 24 hours · No sales team — you talk to the builder

Chat on WhatsApp