Our Vision on Behavioural Design x AI

What does craftsmanship mean if AI takes over every task?

The big question for the next ten years is not in the technology. It sits in the identity of craftspeople. Because AI now does what experts used to need years to learn. A copywriter sees a hundred headlines appear in twelve months that she would have polished for a week. A UX designer gets three wireframes back on a single prompt. An analyst needs three minutes where she used to spend three days in PowerPoint.

So the question is no longer "can I do this?". AI can. The question is: what am I still adding?

AI Won't Replace Humans, But Humans With AI Will Replace Humans Without AI.

This is the 1999-moment of the internet, all over again. The people who took it seriously back then have lived in a different gear for the twenty years since. The same split is happening now, only faster. The difference between craftspeople who take this seriously and those who wave it away will become visible in their output within two years. In their job descriptions within four.

Deep human insight becomes the scarcest skill

Our hypothesis, sharper every month: AI has the IQ of a PhD across every field. What AI does not have is a body, a social context, or the motivation of a real human standing in front of a choice on a Tuesday morning. AI knows everything about behaviour in general. AI knows nothing about your customer, your colleague, your audience, in that one moment.

That is exactly where professional value is shifting. Not to whoever can type, design or analyse fastest. To whoever can see most clearly what is going on under a behaviour. Who asks the right question, checks the right assumption, hands AI the right mental model.

The difference between mediocre and superior is your demanding standards as a prompter.

Ethan Mollick puts it this way: keep the human in the loop. Whoever briefs AI mediocrely gets mediocre output, even if a top model is running under the hood. Whoever knows what good is, and what good is not, can lift AI upward. Craftspeople with taste win.

Behavioural Design is the most systematic language for human behaviour

If deep human insight is the new scarcity, the question becomes: how do you develop it systematically? Not through life coaching, not through gut empathy, not by accumulating experience until you intuitively know what customers want.

Behavioural Design is a synthesis of four fields: behavioural psychology (Kahneman, Tversky), behavioural economics (Thaler, Sunstein), Jobs-to-be-Done thinking (Christensen) and persuasion science (Cialdini, Heath). It is not a single framework. It is a method to analyse, predict and change behaviour.

Why does that work so well with AI? Because 96% of human behaviour runs unconsciously. Language alone does not have access there. Behavioural Design gives you the mental models that make the unconscious part designable. And those models are exactly what AI needs to truly reach your audience.

The mental models we feed AI every day:

• System 1 thinking: images, metaphors and framing to address the automatic brain
• Outside-in thinking: Jobs-to-be-Done, pains, gains, comforts, anxieties (see the SUE Influence Framework)
• SWAC: Spark, Want, Can, Again to explore interventions
• 4C: Catch, Convert, Confirm, Continue to design influence journeys

Behavioural Design x AI: add human intelligence to artificial intelligence

This is where it gets concrete. AI has raw power and no specific audience intelligence. Behavioural Design has deep audience intelligence and hands that cannot execute at scale. Combine them, and you multiply your craftsmanship by a factor.

Two examples from our daily work.

A European agricultural policy paper. We used to read it cover to cover, run a thinking session, brainstorm framing. Now we prompt: "This document is very System 2. Give me three ways to make this vision paper deeply System 1, in which the farmer is the hero of the transition to sustainable agriculture. Make a Guardian cover that brings that story to life." Within five minutes you have three directions you would only have found in week two.

A private banking campaign. Prompt: "Design a campaign intervention for private banking that taps into a moment-that-matters and works outside-in from the audience's Job-to-be-Done. Give me five ideas before we design a campaign poster." The idea itself is not the point. The accelerated exploration is.

"Communicating with AI" is the wrong frame. You carefully conduct the many intelligences inside AI: the researcher, the behavioural expert, the coach, the copywriter.

The verb matters. Not communicating. Conducting.

The biggest mistake: AI is not a technical challenge

74% of employees do not use AI tools, even when they could. You hear this number in every leadership meeting, and the framing is always the same: "We need to do more training." Or: "We are waiting for IT to roll out the right tools." Or: "It is a communication issue."

All wrong. It is a behavioural problem. And inside that behavioural problem sit four framing problems we worked out in our webinar: how management sees AI, how we see AI, how we see AI skills, and how we see ourselves in relation to AI.

The third framing problem matters most to us: we think AI skills are technical skills. The image of the young tech whizz who can do magic with AI is harmful, because it makes everyone else feel they will never get there.

The truth is radically different. The three real AI skills are:

1. Briefing. Knowing how to instruct AI with the right mental models, role and context.
2. Reviewing. Assessing AI output critically from the question "for whom does this not work, and why?"
3. Anchoring. Feeding AI your best work so it gets sharper in your field over time.

None of these are technical skills. All three are craftsmanship.

"AI-First" is an unfortunate frame. Better: craftsmanship first.

From AI tools to AI advantage (webinar)

In March 2026 we hosted a webinar for the Behavioural Design Club where the first five insights came together. More than 300 professionals joined. The full write-up, with the four framing problems and the three superpowers of the AI conductor, sits in the webinar blog. The video and slides:

Download the slides (PDF) →

Behavioural Design frameworks are AI superpowers

Here the picture comes together. The three AI skills, briefing, reviewing, anchoring, are exactly the three things Behavioural Design has had a language for since two decades.

Briefing with Behavioural Design. We rarely prompt with an open question. We hand AI a mental model: "Assess this as System 1 communication. What images and metaphors are missing?" Or: "Analyse this audience through Jobs-to-be-Done. What is the functional job, the emotional, the social?" Or: "Apply SWAC to this adoption challenge. What would the Spark be? How do we make it Want? How do we make it Can? How do we design Again?"

Reviewing with Behavioural Design. The difference between mediocre and superior sits here. A regular reviewer says "this does not work for me" and stops. A Behavioural Design reviewer says: "this does not work for me because you are starting from a negative pain point, which makes the visitor feel bad. Try to reach the same goal through a transformative Job-to-be-Done: how can this product transform the user into a better version of themselves?" That is no longer feedback, that is direction.

Anchoring with Behavioural Design. We build environments in Claude Projects and custom GPTs where our best cases, frameworks and analyses live. Our SUE Behavioural Design Guru knows the Influence Framework, knows SWAC, knows our cases. Ask it for a critique of your campaign and it does not give you generic advice, it gives you Behavioural Design advice. That anchoring is the step most teams skip, and it is precisely the step that lifts an AI from mediocre to superior.

Prototyping at the speed of thought

Behavioural Design has a principle we have been repeating for years: behaviour only becomes visible when you make something concrete that people can react to with their unconscious. Slides change no one. A tangible intervention does.

Prototyping used to take weeks and expensive designers. Now a prototype costs a prompt. An Instagram post, a landing page mockup, a three-frame storyboard, a physical poster design. That changes something fundamental about our method. Where we used to test three interventions in a sprint, we now test thirty.

The cost of bringing something to life is now near zero. That shifts the bottleneck. No longer to production, but to judgment: which prototype is most interesting to put in front of someone? Which behavioural hypothesis do I want to test? AI has no answer to those questions. That is your work.

Astrid and I are rebuilding the entire SUE website with Claude Code right now. I prototype iPhone apps in an afternoon. Not because I can code. Because I know what I want to test.

Synthetic personas: simulate your audience before you spend a euro

This is the insight we think will reshape our work most over the next two years. Feed AI your interview transcripts, customer feedback, NPS quotes and sales call notes. Ask it to build a synthetic persona that simulates your audience. Then test your campaign, your copy, your product, your email against that persona.

Tools like artificialsocieties.io are making this mainstream: you upload your research, the system builds personas, and you can interview them as if they were real people. They are not real people. But they are realistic enough to save you ten iterations before you bring in real research.

Important: this is not a replacement for real research. Synthetic personas hallucinate, miss the subtle non-verbal signals, and only know what you have fed them. But as accelerators of iteration they are unmatched. We use them to sharpen hypotheses before we put the actual audience in the room.