Tournesol Book a call

We build agents

Custom AI systems, built in weeks.

Every system we build grows from the same roots and the same platform, and becomes something of its own.

Typical build: 2–4 weeks

The problem

Every project starts in your ground.

  • Manual work that consumes expert time
  • Rules that change faster than teams can follow
  • Systems that don’t talk to each other

We start where the work really happens: your documents, your rules, your systems.

The platform

One base under everything.

Data goes up. Decisions come down. Every system we build stands on the same platform, with TAG-engine running its workflows, so each new one starts from proven ground instead of from zero.

The method

Inside every part, the same method.

01

Find the impact

We start in your processes, data and rules, and find where AI changes the most.

02

Prove it in 2 weeks

A working demonstration on your real data, before you commit to a build.

03

Build for production

What we prove is what we ship: secure, documented and running every day.

AI drafts, people decide. Calculations and approvals stay explicit — never left to a model — and where a rule can be written down, we write it down instead of guessing.

What we’ve built

Same foundation. Different systems.

Each petal is a different system, built on the same base. Here are two of them.

Field service · In production

A Teams agent for field technicians

Technicians on cryogenic and medical-gas installations enter a job number in Microsoft Teams. The agent pulls the job from the CRM, walks them through a bilingual checklist chosen for that job type, and sends the approved report straight to document generation. They can pause mid-install and pick up later.

Languages
FR · EN
Status
Live
Read the case →

Customs · Product

HS code classification

Classifies products to EU CN/HS codes step by step under the General Interpretative Rules, and returns the code with its reasoning. TAG workflows call a dedicated set of tariff tools over a graph of the full nomenclature.

CN codes indexed
24,000+
WCO opinions
1,235
Read the case →

Your turn

Your project grows here.

Bring us the problem. In a free one-hour call we’ll tell you honestly what’s possible, and what we’d prove first.

Book a call

The team

The people at the roots.

Jaime Melon

Jaime Melon

Co-Founder & CEO

Arnaud Vandevondel

Arnaud Vandevondel

Co-Founder