Connect your business to Claude and ChatGPT
Also called: Custom MCP connector
Service 4 · Agent integration
Hand one routine job to an AI helper, and keep a person in charge
An AI helper (people also call it an agent) is a piece of software that reads, sorts and drafts the way a person would, inside the tools you already use. We pick one job that takes up your team's time, connect a helper to it, and set clear rules for what it can do alone and what a person signs off. We only move to the next job once the first one is working.
Assistant: 12 new emails. 7 are delivery questions, and I've drafted replies from your delivery policy. 2 are refunds over £50, so they're waiting for you.
Person: Send the delivery replies. Show me the refunds.
Assistant: Sent. Here are the two refunds, with the order history for each.
A made-up business. Refunds over a limit always wait for a person.
Why start small
One job, done well
A helper that answers the ten questions you get every day is worth more than a grand plan that never launches. It also shows quickly whether the idea works for you.
People stay responsible
Anything that costs money, changes an order or speaks for you in a new way goes to a person first. You decide where that line is, and you can move it.
You see what it did
Every action is logged in plain words, so you can check its work, spot mistakes and decide when to trust it with more.
What setting up an AI helper involves
Most of the work is in choosing the right job and setting the rules. The software is the easier part.
Pick the job
We look at where your team's time goes and choose one job that's frequent, rule-based and low-risk to start with.
Connect it to your tools
Email, your shop or booking system, a spreadsheet, a help desk. The helper works where your team already works.
Write the rules
What it may do alone, what needs a person, what it must never do, and how it should sound. In writing, agreed with you.
Test it on real cases
We run it on past examples before it touches anything live, and compare its answers with what your team actually did.
Launch with a safety net
It starts by drafting only. As it proves itself, you can let it act alone on the simple cases.
Review and next step
After a few weeks we look at the log together, adjust the rules and decide whether to add a second job.
How an AI helper project runs
Step 1: A short chat
Which jobs take your team's time, and which tools you use for them.
Step 2: A proposal
The one job we'd start with, the rules we'd suggest and a quote.
Step 3: Build and test
Connected to your tools, tested on past cases, with you checking the results.
Step 4: Live, then review
Drafting first, acting alone later, with a review after a few weeks.
Is it right for you?
A good fit if
- Your team spends hours a week on the same kinds of questions or tasks.
- The job follows rules you could explain to a new starter.
- You'd like to try AI on something real, with a person still in charge.
Probably not for you if
- You'd like AI to make important decisions on its own from day one.
- The job changes every time and depends on judgement. A helper can assist, but it won't replace that.
What we need from you
Someone who knows the job well to spend an hour or two with us, examples of past cases, and access to the tools the helper will use.
The helper gets only the access the job needs, and you can switch it off at any time.
Questions about AI helpers and agents
What's the difference between a helper and a chatbot?
A chatbot answers questions on your website. A helper works behind the scenes: it reads emails, sorts orders, drafts replies and updates your systems, within rules you set.
Which AI does it use?
An AI helper usually runs on Claude or ChatGPT, chosen for the job. We'll tell you which, why, and what it costs to run.
What if it gets something wrong?
If an AI helper gets something wrong, it shows in the log. That's why a helper starts by drafting, why the important things wait for a person, and why we adjust the rules when it slips.
Does our customer data leave our systems?
The helper has to read the data it works on, which means sending it to the AI provider. We use providers' business terms, which don't train on your data, and we keep what it sees to what the job needs.
How much does it cost?
A one-off fee to set up the first job, quoted after a short chat, plus the AI provider's running costs, which we estimate for you up front.
Tell us about the job
Which routine job takes up the most time, and which tools your team uses for it.
A real person reads every note. There's no obligation, and we'll tell you honestly if we're not the right people for it.
Often goes with
Be ready for AI agents that do business with each other
Also called: A2A readiness
Keep your connector working after launch
Also called: Connector care plan