The volume is there
Dozens of quotes, invoices or requests every week. One hour saved repeats itself, so the setup pays for itself fast.
Systems & AI agents · design, install, train
Custom AI systems and agents: I build them, you own them
The tasks that come back every week run without you, on your own accounts. You approve what commits the business, the system does the rest.
What I run myself
2 businesses
running on these systems
Claude + n8n
installed on your accounts
Remote or on site
in English, French or Spanish
Short answer
What is a custom AI system?
A system that takes one recurring task from start to finish: it receives the information, processes it against your rules, prepares the document or the action, then stops where your approval is needed. It runs on your own accounts, with the tools you already use, and you own it.
When it fits
When to have it built rather than learn to build it
Training is the right format when you want the skill in house. Custom work takes over when volume, time or complexity leave you no room.
Your tools do not talk to each other
Inbox on one side, CRM on the other, accounting somewhere else. Copying between tools is exactly what a machine does better.
You need your evenings
You could learn. Not right now. I build the system, your team learns to run it.
Examples
Systems already running in my own business
I run these exact systems in my own business every day: drafting quotes, chasing clients, processing supplier invoices and writing meeting notes.
Sales
The sales assistant
When a client asks for a price, a quote is prepared automatically in your inbox.
With no answer after five days, a follow-up is scheduled without any action from you.
Admin
The admin assistant
When a supplier invoice arrives, the key data (amount, VAT, identity) is extracted and filed automatically.
Only the anomalies are flagged for you to check.
Meetings
The meeting assistant
Your video calls are processed to produce meeting notes automatically.
It picks out the decisions taken and assigns the tasks to the right people, straight into your tools.
More cases are covered in the guide AI agent examples for small business.
The method
The four steps of an integration project
01 · Scoping
Rules get written before any code
The process as it actually happens, your decision rules, the edge cases, and what must stay reviewed by a human.
02 · Build
I build the system on your accounts
Your tools, your access, your environment. You watch it take shape and test on real cases as we go.
03 · Go live
We start gently
First alongside your current method, long enough to compare results. Only once it is clean do we switch.
04 · Handover
Your team takes over
Training, documentation, control points. You know what to watch, what to fix, and who to call when a case falls outside.
Maintenance optional
Your tools change, so do your rules. A monthly follow-up is possible, never required: the system keeps running without it.
Ownership
Who owns the system?
Your accounts, your access
Everything is installed on accounts in your name, with nothing hosted by me. You keep the power to shut it all down without telling me.
A system someone else can pick up
Every step is documented in plain English, with explicit rules. Another provider can take over without having to rebuild anything.
You see what it consumes
Subscriptions are in your name and never marked up, and guardrails are in place to prevent runaway AI usage.
Limits
What I do not automate
Anything that commits the business unreviewed
A price, a reply to an unhappy client, a payment. The system prepares, a person approves. Always.
A process nobody can describe
When the rules change depending on who handles the file, they have to be written first. Automating vagueness produces vagueness faster.
Anything that costs more than it returns
A task done twice a year does not justify a system. I say so before the quote, not after.
FAQ
Common questions about custom systems
How long before it runs?
A first useful system can be up and running within days. A full plan coming out of an audit is usually deployed over 4 to 8 weeks.
Which tools do you use?
Usually Claude to read and draft, n8n to orchestrate, and your existing tools for the rest. The choice happens during scoping, never before.
Where does my data go?
The systems run on your accounts, with your subscriptions. During scoping we decide what can be sent to a model and what stays with you.
Am I locked in with you?
No. Documentation handed over, teams trained, accounts in your name. Maintenance is an option, not a condition.
Do I need an audit first?
Not mandatory, but useful as soon as several projects compete. The AI audit exists to set the order.
How much does a custom system cost?
Quoted after scoping, broken down piece by piece. It depends on the number of processes, the tools to connect and the level of control you want.
Getting started
Book a free call
Thirty minutes to describe your heaviest process. I tell you whether it can be automated, how, and what it takes.
Book a free callOr in writing: theo@my-agent.io