Chaining actions in your tools
Looking up information, checking it elsewhere, filling in a form, alerting the right person — without a human triggering each step.
AI agents
Having an AI agent built for your company
An AI agent is a program you give an objective rather than an instruction: it decides the steps itself, uses your tools, and hands you back a result. The difference from a simple assistant that answers questions is that it acts — it creates a record, sends a reply, updates a file.
It is also the category where the gap between the demo and production is widest. An agent that works nine times out of ten is unusable on a process that matters: the provider’s real skill lies there, in what they put in place for the tenth time.
227 providers in the directory cover this need. You pay nothing, neither to search nor if you hire one of them.
What a provider does on this subject
Handling a case from start to finish
A customer request, a quote, a complaint: the agent gathers the information, prepares the reply and flags what it cannot decide on.
Knowing when to stop
The limits, the cases where the agent must hand over to a human, and the record of what it has done. That is what makes an agent usable in production.
Plugging into your information system
Access to the right data, with the right permissions, without opening your tools wide.
Situations that look like yours
- Handling an incoming quote request: qualifying it, looking up prices, preparing the document, submitting it for approval.
- Chasing unpaid invoices: spotting overdue invoices, preparing reminders, escalating above a threshold.
- Preparing sales meetings: gathering the customer history, the company’s latest news, the open issues.
- Monitoring published calls for tender and flagging those that match your criteria.
If your case is not on this list, describe it anyway: we will tell you honestly whether it falls within what our providers know how to do.
What it costs you
€0
to post your project, receive qualified providers and talk with them.
And €0 too if you hire one of them: Claude Partners is paid by the provider you choose, at 20% of the engagement amount excluding VAT. Nothing is added to the quote you receive.
See pricing in detailHow much does an AI agent cost?
There is no list price: an agent is not software you buy, it is development work done on your tools and your rules. A serious quote reads in three lines — the days of design and putting into production, the monthly running cost once the agent is in service, and the cost of rework the day one of your tools changes. A provider who only prices the first line should be ruled out.
What makes that first line vary is almost never the AI model: it is the number of tools the agent needs to access, and whether it writes or only reads. An agent that consults three sources and prepares a draft is a short project. The same agent allowed to send to the customer, approve an amount or modify a case file requires safeguards and acceptance testing that often weigh more than the development itself.
The second line is the one that surprises people: an agent consumes resources every time it runs. Have it estimated on your real number of cases per month, not on the demo — it is what decides whether the agent is still profitable after a year.
To give an order of magnitude, a day is billed at the provider’s day rate: our market grid sets €300 to €700 excl. VAT per day depending on seniority, with €500 for an experienced profile. These are market benchmarks, not the rates of the people in the directory: we only display a provider’s price if they have given it to us. The method is explained in the day-rate barometer.
Scope your request before posting it
Four answers are enough to turn an intention into a request that can be qualified — and to get comparable quotes rather than a gap of 1 to 10 between two providers.
What objective is the agent given?
In one sentence, with a verifiable result. “Prepare a quote reviewed by a salesperson” is an objective; “improve our responsiveness” is not.
Which tools must it access, read-only or with write access?
This is the question that weighs most on the project’s cost, timeline and risk.
Where does it stop?
Decide in advance what it is not allowed to do on its own: send to the customer, approve an amount, delete data.
How many cases a month?
Ten cases a month do not justify an agent. Several hundred do. In between, it is open to discussion.
Four questions to ask the provider
- Show me an agent you have put into production, and what broke in the first week.
- How do I see what the agent has done, case by case?
- What happens when it gets something wrong: who is alerted, and how do we put it right?
- How much does it cost to run per month, once delivered?
The mistakes that cost a lot
Confusing a demo with production
A successful demo proves nothing. Ask to see how it behaves on awkward cases, not on the example chosen by the provider.
Ignoring running costs
An agent consumes resources every time it runs. Have the monthly cost estimated at your real volume, not just the development.
Granting too many permissions “to start with”
Broad write access is very hard to close down afterwards. Start read-only, and open up step by step.
Five minutes to describe your need
No account to create. We get back to you within 48 business hours, even to say no.