Claude for your team.
So it actually saves work.
Many teams have an AI subscription but no agreed way of working. We set up Claude for one recurring task, test it on your real work and train your team.
After three weeks you know whether it works. Measured, not guessed.
A good fit if
- —
Your team already uses an AI tool here and there, or wants to start, but there is no agreed way of working.
- —
There is a task with a lot of reading, searching or retyping: quote requests, supplier documents, customer emails, work instructions.
- —
The result is easy to check for an employee who knows the work.
Not the right start if
- —
Information has to move between systems by fixed rules, with error handling and a log. Then an integration or agent is the better fit.
- —
A mistake would go unnoticed and have serious consequences. That is not a task for a first trial.
How the pilot runs
This is the usual planning. Missing access to software, unclear data or an extra privacy review can take longer. We tell you that before we start.
Task and baseline
We pick one task and measure how it runs today: how many cases, how much working time, how much rework. We agree which source is leading and what Claude should deliver.
Set up and test
We set up the workspace: access, instructions and one checked set of documents. Then we test on normal and unusual cases from your own work.
Guided use
The pilot group uses it on the daily task. Short instructions: what may go in, how do you check an answer, who do you report a mistake to? We keep track of corrections.
Measure and decide
We compare against the baseline, including the time spent checking. Then you decide: scale up, adjust or stop. No annual contract.
What you get
- 01
A configured workspace for one task, with instructions and approved sources.
- 02
A written way of working: what may go in, who checks, what happens with the result.
- 03
Training for the pilot group on their own work, not on a demo.
- 04
A before-and-after measurement: working time per case, rework and use by the team.
- 05
Advice on the next step. Sometimes that is: don't scale up.
What you arrange
- 01
One person responsible for the pilot, who knows the work.
- 02
A business Claude subscription in your company's name. We help you choose which plan fits.
- 03
Real examples of the work: permitted or anonymised, including the difficult cases.
- 04
Time from the pilot group in week 3 to work with it and note corrections.
From tool to agent
Not every task needs an agent. We start as light as possible and only go a step further when the work asks for it.
Claude for your team
Adopting the tool. No integration, but clear agreements. This page.
A simple integration
When information has to reach another system the same way every time.
An AI agent
Recurring work with rules and exceptions, connected to Exact Online. Every outgoing action waits for your approval.
About AI agents →Frequently asked questions
Why not just buy a subscription ourselves?
You can. The subscription is the easy part. What is usually missing: which task, which sources are correct, who checks, and how you know whether it saves time. Without those agreements everyone uses it differently, and after a few months nobody can say what it delivers.
Does this also work with ChatGPT or Copilot?
We set up Claude. Already using a different AI tool? Then we discuss in the Process Review whether switching makes sense. Sometimes the honest answer is: stay where you are and make better agreements.
Is NineMinds an Anthropic partner?
No. Claude is a product of Anthropic. NineMinds is an independent firm that helps you put it to work in your business.
What happens to our company data?
Anthropic says it does not use inputs and outputs of its business products for model training by default. That does not automatically mean data stays in Europe. So before we start, we agree which data may go into the pilot and which stays out. Together we check whether a DPIA is needed.
What if the pilot delivers little?
Then you see it in the measurement and it stops there. That is an outcome too: you know where not to spend more time.
When is an AI agent better than Claude for the team?
When the work runs the same way each time and information has to move between systems: orders into Exact Online, invoices matched against purchase orders. Then you want fixed checks, error handling and a log. We build that as an agent. A Claude pilot often shows where that work is.
What does it cost?
A fixed price, agreed after the free Process Review. It depends on the task and the size of the pilot group. Claude subscription costs are separate. No hourly billing, no annual contract.
Want the full method first? Implementing Claude in an SME: step by step (in Dutch) →
Which task first?
In 45 minutes we look at where your team loses time and whether Claude is the right tool there. Free, no obligation. Sometimes the answer is that something simpler will do.