The company policyPolicy excerpt
Order changes and delivery
Delivery dates are confirmed after the order is reviewed.
A customer may request a change by replying to the order email.
AI Training
Your team practises AI on the work they already do.
A session is a short explanation and then practice: people work on tasks from their own job, check what the tool gives back against a real source, and leave with the exercises and a guide for checking the next answer.
- Two hours, four hours or a full day
- Up to 8 or 12 people
- Your own tasks, your own sources
What your team practises
Five steps, practised on each role’s own work.
We cover choosing a suitable task, preparing a useful request, checking the answer against its sources and deciding what information can be shared. Each session includes an exercise and a review of where the output needs human judgement.
Who a session is for
Choosing a role changes the examples below.
-
01
Step 1 of 5
Choosing the task
Something the person does often, with a source they can check the answer against.
A customer asks about delivery or returns.
-
02
Step 2 of 5
Preparing the request
The task, the source the tool may use and the form the answer should take, written down before asking.
The policy excerpt goes in, with a request for a short reply.
-
03
Step 3 of 5
Checking against the source
Every factual statement in the answer is traced back to the source, or marked.
Does the policy really say the change is free?
-
04
Step 4 of 5
Handling a missing fact
When the source doesn’t answer, the result says what still needs checking and who can confirm it.
The fee isn’t in the policy, so the reply says it will be checked.
-
05
Step 5 of 5
Applying the checking guide
The same checks on a new example, so the habit carries into everyday work.
A second customer question, checked without help.
Try one exercise from a session
Check this reply before it goes to the customer.
A customer asks to change an order. An AI assistant has drafted the reply, and it sounds fine. Do what your team practises in a session: check each promise in the draft against the company policy, then see how your check compares with ours.
- 1Check the three promises
- 2Write a better reply, if you like
- 3See the answer
The customer asks
Hi, can I still change my order? And could it arrive tomorrow?
The exercise uses no customer data and no AI model. Your reply stays in your browser.
The AI’s draft reply
Yes, you can change the order1 free of charge2, and it will arrive tomorrow3.
Promise 1: “Yes, you can change the order”
Promise 2: “free of charge”
Promise 3: “it will arrive tomorrow”
How your check compares with ours
1Yes, you can change the order
- Your mark
- Our mark
- Partly
Different from ours
The policy lets a customer request a change by replying to the order email. It doesn’t say every change will be accepted.
2free of charge
- Your mark
- Our mark
- No, it doesn’t say
Different from ours
The policy says nothing about a fee for changes, so the reply can’t promise there isn’t one.
3it will arrive tomorrow
- Your mark
- Our mark
- No, it doesn’t say
Different from ours
Delivery dates are confirmed after the order is reviewed, so tomorrow can’t be promised before that.
The reply we’d send
You can request a change by replying to your order email. The delivery date needs to be confirmed after the order is reviewed. We also need to check whether the change has a fee.
The policy supports requesting a change, not confirming that every change is possible. It gives no fee waiver and no next-day promise. A good answer keeps those distinctions and tells the customer what still needs checking.
How a session runs
What happens before, on the day, and after.
The day itself is the middle of the work. What makes it useful is choosing the right tasks beforehand and having something to keep afterwards.
-
Before
We pick the tasks together
A call with whoever knows the work: what your team does most often, which tools you already have, and which source an answer should be checked against. The exercises are built from that.
-
On the day
Short explanations, longer practice
Each step is explained in a few minutes and then practised for longer. In a mixed group people work in tracks by role, so nobody spends the day on somebody else’s task.
-
At the end
A review of what needs a person
We go back over what the tool got right, where it was confidently wrong, and which decisions stay with the person who owns the task.
-
After
What your team keeps
The exercises, the checking guide and a short list of tasks to try next. The Team Session adds a prompt library. The Team Day adds two weeks of answers by email to the questions that come up afterwards.
Checking guide5
- Identify the source the answer should come from.
- Check each factual claim against it.
- Mark any information that’s missing.
- Confirm you may use that information in this tool.
- Name the person who approves the result.
Sessions and prices
Three formats, one fixed price each.
Every format works on your own tasks. The group size is what the price covers, and each person beyond it is added at a fixed rate.
-
Leadership Briefing
Where AI fits your work and what the EU AI Act asks of companies that use it, for the people who decide.
€690per session
Runs in one day
-
Team Session
Practice in tracks for the roles on your team, with a prompt library your team keeps.
€890per session
Runs in one day
-
Team Day
A full day on your team’s real tasks, then two weeks of answers by email to the questions that come up.
€1,490per session
One day, then two weeks by email
Extra Participant €60 per person beyond the group size
Which format fits depends on how many people join and how much practice you want. The calculator adds the extra participants and shows the total.
Calculate your priceLimits and responsibility
What a session does not settle.
-
It doesn’t make the output reliable.
Training makes people check. It doesn’t make the tool right, and a checked answer still needs the person who owns the task to approve it.
-
It doesn’t write your rules.
Which tools are approved, what may be typed into them and how long anything is kept are your organisation’s decisions. We can lay out the choices; we don’t make them for you.
-
Sometimes a session isn’t the answer.
If a task repeats and is well defined, connecting the tools you already use will do more than teaching people to do it by hand. We’ll say so when that is the case.
The EU AI Act
Companies that use AI are asked to support their staff’s AI literacy.
Article 4 of the EU AI Act applies to providers and to deployers, and a company using an AI system in its work is a deployer. Since the July 2026 amendment, it asks them to take measures that support the AI literacy of their staff and of others who operate AI systems on their behalf.
It doesn’t require a particular level for each person, and a session isn’t a certificate. Training is one of those measures, and your other legal and data-protection responsibilities stay separate.
EUR-LexRegulation 2024/1689, Article 4
Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, […]. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual.
Questions
The questions we get asked.
Do we need to know anything about AI beforehand?
No. The sessions assume no background. People need a laptop and an account for the tool they will use, and we say which one after the first call.
Which tools do you teach?
The ones your team will actually use. If nothing is chosen yet we work with a general assistant and show what changes between tools, because the checking habit is the part that transfers.
Is there a certificate at the end?
No. Article 4 asks companies to take measures that support AI literacy and states that no specific level has to be guaranteed for any individual, so a certificate would claim more than the law asks. You keep the material, the exercises and the checking guide.
What if more people want to join than the group size?
Each person beyond the size is added at a fixed rate. Past twelve people a second session works better than a fuller room, because everyone is meant to be working, not watching.
We already use ChatGPT. Is this still useful?
Then the gap is usually checking rather than using. The session spends its time on what the tool cannot know, where it is confidently wrong, and what to do when the source does not answer the question.
Do you use our data in the session?
We build the exercises from examples you choose. Before the session we agree what may go into the tool and what may not, and that list is part of the material your team keeps.