Ilkka Rajala in a dark blue three-piece suit, hands in pockets
Ilkka Rajala
FIEN

AI training for companies

Thinking in the age of AI.

AI answers faster than you can think. So who is doing the thinking, you or the machine?

Ask about available datesFive formats

Where we are

The tools have been bought. The work hasn't changed.

Many organisations have bought the licences and run the user training. Yet the work is done as before. Part of the team doesn't touch the tools, and those who use them most check the least.

The software works. The bottleneck is in thinking, and in how people experience the change.

The second problem only shows later. In a study by Microsoft and Carnegie Mellon University, those who trusted AI the most thought critically the least themselves (Lee et al. 2025).

59% of Finnish respondents say AI has hardly changed their work
5% say AI is part of their organisation's core processes
76% of leaders believe their staff are excited about AI
31% of employees say they are excited

Sources: Professio and Taloustutkimus, Future Skills for Working Life 2026, 1,640 respondents. People Element, Employee Engagement Report 2026.

Someone else can teach the Copilot buttons.

I teach what tool training doesn't reach: how a person thinks together with the machine.

The result shows in two ways. People use AI to grow their own expertise. Routine work moves to the machine, and more time is left for their core expertise. At the same time they learn to judge which output they can trust and what needs checking.

A person who knows their own thinking and knows where to check gets more out of the machine. And still thinks for themselves.

An old warning

The appearance of wisdom without wisdom.

In Plato's Phaedrus, King Thamus warns the inventor of writing: your students will get the appearance of wisdom from you, not wisdom itself. The warning is about 2,400 years old, and it fits AI even better than it once fitted writing.

The machine writes fluently and confidently. Fluency looks like competence, which makes it hard to doubt. The same trap waits on the reader's side: you can get information in an instant, but understanding only comes when you apply it and your own thinking changes.

In the end writing became a tool for thinking, once people learned to use it. The same learning lies ahead with AI, and it goes faster when you practise it deliberately.

Five skills

Five skills that keep thinking sharp.

The same five skills are part of the keynote, the training day and the programmes. The difference is how much time there is to practise them. Programmes for managers and leadership teams also cover leading change.

  1. Think wider

    AI has never been usable like this before, so our usual way of thinking doesn't know where to look for uses. Many people use it like a search engine and get only a fraction out of it. A wider way of thinking sees the whole chain of work and finds places for AI you hadn't noticed before.

    The other half is unconventional thinking. When everyone uses the same machine, ordinary ideas get cheaper. The value goes to whoever connects two distant fields or turns an assumption upside down. That can be practised.

    Research. In an experiment with 300 writers, AI ideas made individual stories more creative, but the stories began to resemble each other. (Doshi & Hauser 2024)

  2. See your own thinking

    Before you ask the machine, stop and notice what you already think: what you assume, what you feel and what matters to you. When the answer comes, look again. What changed? Perhaps your perspective or your certainty, sometimes even what you consider right.

    I call this self-calibration. Your starting point is the yardstick against which you notice how the machine affects you. When you put your own thought and the machine's answer side by side, you also see your own thinking biases and can judge which one holds up better.

    Research. In a study at Aalto University, AI improved participants' results, but everyone overestimated their own performance. Those who knew AI best misjudged themselves the most. (Fernandes et al. 2026)

  3. Ask with structure

    The structure of the question determines the structure of the answer. A vague question gets a vague answer. You learn how the questioning techniques I have taught for more than 20 years work with AI in everyday work.

    When AI does the work itself, the question becomes a brief: the goal, the background, the limits and the point where the machine stops to ask. A good brief requires that you know how you do your own work.

    Research. In an experiment with nearly 1,900 participants, people tried to get AI to produce a specific image. About half of the improvement came from people learning to ask more precisely. When the machine rewrote the requests on the user's behalf, it added its own ideas, and the benefit shrank by more than half. (Jahani et al. 2025)

  4. Doubt in the right place

    Text produced by AI feels true. The machine goes along with whoever is asking, and a busy person trusts the automatic answer. You don't need to check everything, but you need to know what to check. You learn where errors are usually found, and where in an agent's long chain of work the human check belongs.

    Research. In an experiment with 758 consultants, AI improved results on tasks it was suited for. On a task outside its capabilities, those using AI reached the correct solution 19 percentage points less often than those working without it. (Dell'Acqua et al. 2026)

  5. Keep what matters most

    Delegate routine work to the machine. Keep judgement, values and the decision for yourself. A skill you don't use gets weaker. That is why you also learn to use AI in a way that develops your own skills.

    AI gives you information in seconds, and the answer sounds wise. Information is easy to get and remember. Applying it to your own situation and changing your own thinking are skills that only come from doing.

    Research. High school students practised maths with GPT-4. Their practice results improved by 48%, but in the exam without the machine they did 17% worse than those who never had it. When the machine only gave hints, exam results stayed at the level of the control group. (Bastani et al. 2025)

    Example. More than ten years ago I was assisting a trainer in the United States who is known for the way he influences people with language. At lunch I asked the camera operator for the morning's tape and transcribed it by hand in my hotel room, word by word. Then I studied the structure of the text: the assumptions, the metaphors, the order and how he built emotional states in the audience. Often I finished only at night. Today a machine would transcribe the same in a minute. I understood the structure only by doing that work myself.

For managers and leadership teams

Leading change

People worry about what the tool will do to their expertise. Fear, shame about not knowing how, and leadership's overly rosy picture stop adoption. Poor use also costs working time.

Research. 40% of employees had received AI-generated busywork from a colleague in the past month. Sorting out each case took almost two hours. (Stanford and BetterUp 2025)

Agents

From question to brief.

AI already does work on its own. An agent gathers the information, writes the report, updates the spreadsheet and moves through ten steps in a row without anyone checking each one. The human's work shifts to giving the brief, setting the limits and checking. That is a manager's job, and now every expert does it.

An agent does the work as well as its instructions allow. Good instructions require the expert to know how they do their own work, and few can put that into words. For more than 20 years I have helped people see the structure of their own thinking. The same skill now helps write it down for the machine.

  • The question becomes a brief: goal, background, limits and quality criteria.
  • Checking becomes oversight: where in a long chain of work the human looks.
  • Autonomy is decided deliberately: what the machine may do on its own and what a human approves.
  • The expert's tacit knowledge is written down as instructions the agent follows.
Ilkka Rajala in a blue blazer, reaching out a hand

Why me

I know this from both ends.

I have taught thinking skills for more than 20 years and trained more than 50,000 people. At the core of my work has always been metacognition, the ability to notice your own thinking. Now that the machine increasingly thinks for us, almost everyone needs that skill.

According to philosopher Andy Clark, the most important skills in the age of AI are metacognitive: knowing what to rely on and when (Clark 2025). That is what I have taught throughout my career.

In 2015 I helped found an AI company where we taught a machine to read from speech and text how a person thinks. I know where the machine is strong and where it only sounds confident. I use AI agents in my own work every day.

The other side of change is in people. When adoption gets stuck, the reason is usually what people are afraid of losing. I have worked on organisational change since 2005.

From a keynote to your own trainer.

Five formats, the same content. You can start small and expand once you see results.

0145–60 min

Keynote

For All staff, a kickoff, a client event or a seminar
Price from €2,500 + VAT

AI produces mediocre text in seconds and almost for free. That is why good thinking of your own is now worth more than before.

In the keynote I go through five skills that help everyone get more out of AI and keep their own thinking sharp. They work for anyone who uses AI at work, whatever their role.

The audience tries the effects for themselves in the room. At the start, the room estimates a number first on its own and then after seeing the AI's answer. Many change their estimate, including those who say the machine doesn't influence them.

  • Think wider: why our usual way of thinking doesn't find uses for AI, and how unconventional thinking finds places others don't see.
  • See your own thinking: what you already know and assume before you ask, and how to recognise when AI is influencing you.
  • Ask with structure: how to phrase a request or a brief so that the machine's work is useful.
  • Doubt in the right place: what in the answer needs checking and what you can trust.
  • Keep what matters most: what you give the machine and what you do yourself, so that your own expertise grows.

Each skill leaves you with something you can try the very next morning. Also available as a 30 or 90 minute version.

021 day

Training day

For A team, a unit or a group of experts
Price from €5,700 + VAT

The keynote shows the effect. In the training day the five skills are taken into your own work.

Everyone brings a real work task where they use AI or could use it. The whole day is spent practising on it, so the skills are in use the very next day.

  • Morning: thinking wider and seeing your own thinking. Where AI can help, what starting point you ask from, whether your thinking supports seeing opportunities, and how to recognise when you are being influenced.
  • Afternoon: structured asking, doubting in the right place and what you keep for yourself.
  • At the end the group agrees on shared ground rules: where AI is the default and where a human always decides.
  • A short survey before and after, so you can see what changed.
  • The exercises are done with the tools you have approved. Confidential information is not taken into public services.

The day can also be run as an expert workshop. Everyone unpacks the structure of their own best way of working and writes it up as instructions for an AI agent they can use right away.

0310 weeks

Manager programme

For Managers or a key group
Price from €14,800 + VAT

The manager decides whether the team uses AI well, badly or not at all. In the programme managers take on the five skills themselves and learn to lead a change that can't be ordered.

They take what they learn to their teams with a ready-made discussion outline. That is why the effect reaches all staff, even though only the managers are coached. Results are measured before and after.

Leading change

Managers learn to notice the fear and the shame of not knowing how that stop adoption in their team, and to talk about them openly.

  • Baseline survey for the participants and their teams.
  • Week 1, three hours: thinking wider and your own thinking. Everyone picks a use case from their own work to develop throughout the programme.
  • Weeks 2–4: a weekly assignment in your own work, 15 minutes a week.
  • Week 5, three hours: structured asking, doubting in the right place and what you keep for yourself. Plus a map of resistance in your own team.
  • Weeks 6–8: managers run a 30-minute discussion with their teams, using a ready-made outline.
  • Weeks 9–10, three hours: final survey, results and a report for the leadership.

I can also offer participants personal sparring.

04about 2 months

Leadership team programme

For Leadership team
Price from €9,500 + VAT

Many leadership teams have decided to adopt AI. Few know what their staff really think about it, or how AI is already changing the leaders' own decisions.

In the programme the leadership team practises the five skills in its own work, sees why adoption is stuck and makes the decisions in the same room.

Leading change

The leadership sees what lies behind the resistance and corrects its own, often overly rosy, picture of the situation.

  • Pre-survey for the leadership team and a selected sample of staff: use, trust, fears and their own judgement.
  • Workshop 1, 3.5 hours: where the leadership's picture and the staff's picture differ. The five skills are practised on a real strategic question. Thinking wider, we look at where AI could change entire chains of work.
  • Workshop 2 about six weeks later, 3 hours: leading the change. Which units aren't moving forward and what expertise people are afraid of losing. Finally you decide what the organisation gives to the machine, what it doesn't and who is responsible for the work agents do.
  • You get a one-page summary, a list of decisions, named owners and a baseline for the metric.

The first workshop can also be ordered on its own, from €4,500 + VAT.

052 days + materials

Trainer licence

For Large organisations with their own training or HR team
Price €17,500 + VAT

When the goal is to reach all staff, your own trainer is often the most sensible route.

I train your trainers to teach the five skills in their own voice and on their own schedule.

  • Two-day trainer training: the five skills, the exercises and how to run the demonstrations.
  • Materials: slides, exercises, a team discussion outline and a survey.
  • The price includes a licence for the whole organisation and training for one trainer. You can bring more trainers to the same training for €1,500 + VAT per person.
  • The licence covers training your own staff.

Also works as a next step once the manager programme has shown results.

EU AI Act

What the law requires.

The EU AI Act requires organisations to support their staff's AI literacy. The obligation applies to every organisation whose staff use AI at work. That includes everyday tools such as Copilot and ChatGPT.

Training has to suit people's roles, skills and the context of use. User instructions alone are not enough, and it is worth documenting the measures. In Finland only 17% of employees have received AI training, even though 63% have used or tried AI at work.

When AI is used in areas such as recruitment or credit decisions, the Act also requires that people recognise their tendency to over-rely on the machine. That is a skill I teach.

What you get

  • Training designed for each role.
  • Measurement before and after.
  • A written description of the measures for documenting AI literacy.

Sources: EU AI Act, Articles 4 and 14. CGI and Norstat 2025.

Measurement

Change is measured.

The programmes use the same survey at the start and at the end. We ask how often people use AI, whether they form their own view before asking and whether they check the output before using it in a decision.

One item is not a matter of opinion. The respondent gets an AI answer with two hidden errors. How many do they find? That measures judgement directly.

The respondent also writes down their own estimate before seeing the AI's answer, and reassesses it afterwards. How far did the estimate move, and did they notice the shift themselves? That measures self-calibration.

The results of the measurement also serve as documentation of how you have supported your staff's AI literacy.

Guarantee

If you're not satisfied, I won't invoice you. That is how I have always worked.

Frequently asked about AI and thinking.

What are AI agents and what do they change in expert work?
An agent is AI that carries out a multi-step task on its own: it searches for information, writes, calculates and reports. The expert's work shifts to giving the brief and checking the result. The training practises both, using your own work tasks.
Does AI weaken thinking?
It can, if the machine does the work people used to learn from. In an experiment with high school students, AI improved practice results but lowered exam results once the machine was gone (Bastani et al. 2025). When the machine only gave hints, there was no harm. That is why I teach using AI in a way that grows your own expertise.
What does self-calibration mean?
It is the skill of noticing your own thinking before the AI's answer and after it. First you notice what you assume, feel and consider important. After the answer you look at what changed: your perspective, your certainty or even your values. The difference shows how the machine affects you. When you compare your own thought and the machine's answer side by side, you also notice your own thinking biases and can judge which one holds up. The name comes from measurement: an instrument is calibrated by comparing it with a known reference, and here the reference is your own thinking before the question.
Which tools are used in the training?
Your own, for example Copilot, ChatGPT or Claude. The training is about thinking together with the machine, so it works with any tool. Confidential information is not taken into public services.
What does the EU AI Act require from training?
Article 4 requires organisations to support their staff's AI literacy so that training suits people's roles, skills and the context of use. Measurement before and after, and a written description of the measures, serve as documentation.

Ask about available dates. I'll tell you where to start.

Ask about available dates ilkka@ilkkarajala.fi +358 40 508 5615