In a fisheries office in Peterhead, an officer is working through a complicated licence case, chasing a piece of missing information. Then a vessel arrives, and compliance information needs to be gathered and updated. The officer stops working on the case and turns their attention to the quay.
We only saw this because we were there in person. None of the documentation we had analysed described a moment like it.
Scott Logic is working with the Scottish Government’s Marine Directorate to replace FIN, the service through which it issues and updates licences for every Scottish fishing vessel. AI has dramatically increased the pace of that work. Research, analysis and prototyping have moved far faster than a traditional approach allows, and the time saved is measured in weeks and months rather than hours and days. AI will help accelerate the build phase too.
Yet the faster we have gone, the clearer one thing has become: change still has a human speed. Understanding, alignment and insight do not arrive at the speed of an AI output. They come from productive friction: the effortful human engagement that AI cannot do on our behalf and that we would not want it to.
That might sound like a constraint. I see it as an opportunity: using AI with more focus and intent leads to better outcomes. Drawing on my background in user research, service design and design thinking, I see productive friction at work on three levels: in how individuals absorb what AI produces, in how teams make sense of it together, and in what only direct contact with users can reveal.
Fast analysis needs slow reflection
AI has been a huge help in the Discovery phase, and we have been measuring how much faster it makes us, compared to a traditional, manual approach. We have moved more quickly at every stage, from understanding the legacy system to specifying and designing the features of its replacement.
What has not changed is the need for our experts to do the thinking themselves. AI can analyse reams of material at phenomenal speed and produce an impressive set of findings. Those findings are of limited use until someone absorbs and interrogates them.
It is the act of analysis that builds understanding in the analyst. Psychologists call this the generation effect: we remember information better when we generate it in our own minds than when we simply read it.
Thinking is not only about gaining knowledge. It is also about prompting questions, sparking ideas and spotting assumptions to validate, all of which draw on distinctly human qualities. You will know from your own AI use that it can require an expert eye to pick out what’s important, spot any mistakes, and detect biases in the AI’s outputs. That effort is productive friction at the level of the individual and it is how each expert sharpens their understanding before testing it against the views of others.
Speed must not come at the expense of understanding. When AI compresses the analysis, projects need to plan time for people to sit with the results, question them and form their own view.
Frictionless outputs need human friction
Zoom out from the individual to the team and the same pattern holds. AI speeds up each team member’s work, but the biggest gains come from our conversations as a multidisciplinary team. That is where we question each other’s readings of the findings, challenge assumptions and extend our understanding, rather than simply accepting what AI has produced.
Those conversations matter because each discipline brings a different perspective to the same problem. A user researcher, a business analyst and an engineer will each surface questions, risks and opportunities that others, including AI, may not see.
The Discovery artefacts we can now generate quickly help us articulate ideas and get past communication barriers, within the team and with our Marine Directorate colleagues. AI helps start the conversation, but people still have to finish it. Alignment comes from the team and stakeholders engaging with those artefacts, debating them and making decisions together. Projects should plan these structured review and challenge sessions as a deliberate step, not something squeezed in once the outputs exist.
Knowing what to ask isn’t knowing what you’ll find
FIN is a service used by people, and its replacement will only be successful if it also works for people. User research and service design have been central to our Discovery and Alpha work, including a visit to Peterhead in Aberdeenshire, Europe’s largest white fish port.
To prepare, we used AI to search the Marine Directorate’s large body of existing user requirements and pull out those relevant to the visit. That saved us at least five days of reading and gave us a set of assumptions and questions to validate. Any human-centred design practitioner will tell you that is only half the story. The real insight comes from being open to the unexpected.
Watching fisheries officers at work, we saw the complexity behind what looks like simple licence processing. We heard the terminology and distinctions that matter to the people doing the job. And we saw at first hand the practical judgement and adaptability officers bring to keeping cases moving while balancing other duties, like the officer who set aside a complex case to monitor an arriving vessel.
None of this tacit knowledge was in the existing documentation, yet all of it needs to shape the design of the new service. AI is now a powerful tool for human-centred design, but it can only analyse the evidence it is given. It cannot generate new knowledge through observation.
Projects should therefore reinvest the preparation time AI saves into the aspect of human-centred design that matters most: observing people doing real work in real places.
Making room for human speed
AI has given us time back. The question is how we spend it. Our experience with the Marine Directorate suggests that the answer is to reinvest it in the human work that AI can help us prepare for, but which it cannot do itself.
In practice, that means three things:
- Build reflection into the plan, so individuals have time to absorb and question what AI produces.
- Treat team and stakeholder discussion as a deliberate step, because shared artefacts are not the same as shared understanding.
- Spend saved preparation time in the field, observing the people the service is for.
Change still has a human speed. Projects that respect it, rather than trying to engineer it away, will get more from AI, not less.
My colleague Suzanne Angell will share more of what we have learned at the DigitalScotland conference on 24 November, including the positive effects of AI on people’s roles, on how teams align, and on the satisfaction they get from their work.