Train your team and put its first AI agent in production.
Colombani.ai trains your team on a use case you choose, builds the agent with them on your tools and your documents, and puts it into service.
- From €6,000 excl. VAT
- firm quote after scoping, once the use case is chosen with your team
- Training
- by a Qualiopi-certified provider, OPCO funding possible
- On your tools
- Claude, ChatGPT, Copilot or a local model
- Guarantee
- build not invoiced if the agent is not in production
Teams that already use AI, with nothing yet in production
Your people use Claude, ChatGPT or Copilot every day, each on their own. Nothing yet runs as a shared tool that saves time week after week. The offer is for executives, partners, HR directors, team leads and CIOs who want that first tool, whatever the size of the organisation.
Four stages
- 01
Scoping
We start from a precise task, not from a tool: who does it, where it gets stuck, who owns the result. We identify three cases, choose one and measure the time it takes today.
- 02
Training
Training is built on your case and your examples, rather than a tour of features. Everyone knows when to use the agent, how to review what it produces and where to raise a problem. Certificates of completion issued.
- 03
Build
The agent is built with your team, on its tool, and tested on your real documents. The rules of use are written down.
- 04
Go live
The agent goes into service. We measure again, tell you what the measurement shows and what it does not show yet, then hand over.
How the first case is chosen
The first agent is not the most ambitious one. It is the one your team will still be using the following week, and the weeks after.
- Frequent
- the task comes back every week, so the team builds the habit
- Precise
- one step of the work, one group of people, one expected output
- Supplied
- the documents it needs exist and can be used
- Measurable
- the time or quality can be compared before and after
A case of average value that is ready beats a high-value case that is still blocked. The blocked one comes next.
- We start
- We first check one point
- We narrow the case
- We wait: something essential is missing
- An agent in service on your tool: Claude, ChatGPT, or a local model if your data must stay in-house
- A trained team, with its certificates of completion
- Written rules of use: what the agent does, what it does not, who checks
- A before and after measurement of the time spent on the task
- The documentation and the code, which belong to you
A business law firm checked every court decision cited in its briefs by hand before filing them. The tool built for them reads the text, extracts each reference, checks it in official databases and flags the decisions whose content does not match what is attributed to them. The firm remains in control of what it files in court, and now only rereads the flagged references.
Before you book
How much does it cost?
From €6,000 excl. VAT. The final price depends on the agent to build, so the firm quote is issued after scoping, once the use case is chosen. The training part can be funded by your OPCO.
Can the training be funded?
The training part is delivered by a Qualiopi-certified provider (training provider registration number 11757549975, this registration does not constitute State approval) and can be covered by your OPCO according to its criteria. The build of the agent is not training and is not funded.
What if the agent is not in production at the end?
The build part is not invoiced.
Which tool will the agent run on?
On the one your team already has: Claude, ChatGPT or Microsoft 365 Copilot. When data must not leave your premises, the agent runs on a local or European model.
Who needs to attend the scoping?
The people who do the task today, the person who owns the result and, depending on the data involved, IT, security or legal. Each judges the result differently, and comparing those views is what makes it possible to pick a case that lasts.
Does the agent act on its own?
Not by default: it prepares, your team reviews and decides. An agent that writes, sends or changes something needs more safeguards: who authorises it, what it may touch, where it stops to ask. They are written down before go live.
What happens after go live?
The agent, its documentation and its code are yours. At the end, depending on what the measurement shows, we recommend one of these paths: continue, adjust, extend the agent to a neighbouring case or stop. What comes next is not included in the offer.
Start with 30 minutes
Describe the task that takes your team the most time. We tell you in thirty minutes whether it makes a good first agent.