What I can take on
Custom AI agents
Voice or conversational agents connected to your tools: answering, qualifying, scheduling, logging. With business rules hard-coded, not left to the model's judgement.
Automatisation de processus
Automated chains between your software: calendar, invoicing, CRM, inbox, spreadsheets, forms. What used to be copy-pasted now happens on its own.
Integrations and APIs
Getting two tools that don't know each other to talk. Business calendars, ERP, SaaS platforms, web services.
Analysis and mapping
Laying out an existing process, costing it out, and honestly saying what deserves to be automated — and what doesn't.
Dashboards
Making visible what used to live in scattered spreadsheets: volumes, turnaround times, usage, incidents.
Handover and documentation
Everything I build is documented and belongs to you. You must be able to take it over yourself, or hand it to someone else.
How it works
Four steps, no commitment upfront.
01
We look
A call, then observing what you actually do. Often the real problem isn't the one we thought it was.
02
I cost it out
What the current situation costs, what automation would cost, and how long it takes to pay for itself. In writing.
03
We test small
A prototype on a real case, not a trade-show demo. If it doesn't hold up, we find out then — not six months later.
04
We go live
Rollout, monitoring, documentation, and knowledge transfer. You're not locked into the provider.
One principle that explains the rest
A language model is excellent at understanding and phrasing things. It's poor at counting, following a rule to the letter, or being reproducible.
So in my work, AI talks and understands — but business decisions (hours, durations, deadlines, amounts) are computed by tested, deterministic code that gives the same result every time.
It's less impressive in a demo. It's also what separates an impressive prototype from a service your business can actually run on.
Tell me what's taking up your time
The first conversation costs nothing, and it sometimes ends with “it's not worth it”. That's a useful answer too.