People before technology. Define the problem, measure the impact and choose the role of AI: augment, assist or automate?
AI dazzles us. Every time a powerful technology emerges, we jump straight to ideas, solutions and use cases. But when we start with the solution, we forget what truly matters: what real problem are we solving, for whom, in what context, and with what human impact?
AI is remarkable, yes, but it remains a tool. The order matters: people first, technology second.
The dangerous shortcut: from "what can we do" to "what for"
The temptation of solutionism pushes us to:
- Seek use cases before understanding the problem and its context.
- Optimise internal metrics and forget the outcome for the user.
- Build "intelligent" prototypes that don't fit into real-world workflows.
Result: brilliant solutions that resolve nothing of substance, or that add cognitive and operational friction.
"When you adopt technology, you always have to make sure there's a business case: find a problem to solve or a process to improve… you apply technology to it and then the business case for doing so emerges."
Ángela Gómez
Starting with the human being (and their context)
Before talking about flows, prompts or agents, do the groundwork:
- Who: which segments (personas/roles) will be the beneficiaries?
- Context: tasks, tools, constraints and the physical/social environment.
- Cognitive friction: where mental effort, ambiguity or working-memory load is concentrated.
- Desired outcome: what changes for the person? What does "success" mean to them?
Does it make sense to solve it with AI?
Not every problem needs AI. Run it through this filter:
- Differential human value: does it improve time, confidence, quality, access or accuracy for the person?
- Non-AI alternatives: can I achieve a meaningful improvement without AI?
- Appropriate metric: how will I measure the outcome that matters to the user?
Decide the role of AI: augment, assist or automate
- Augment (human-led): the person decides; AI suggests, summarises and prioritises. Ideal where expert judgement is required.
- Assist (human-in-the-loop): AI does the heavy lifting; the person validates or corrects. Useful for scaling with control.
- Automate (hands-off): only for tasks that are repetitive, low-risk and well-defined.
Human First, Next AI is not a slogan: it is a way of working. When we start with people — their cognitive friction, their context and their idea of success — AI stops being "magic" and becomes a tool. The formula is simple (and demanding): define the right human problem, measure the outcome that matters and choose the appropriate role for AI (augment, assist or automate). That is how we move from brilliant prototypes to real impact.
AI suggests, summarises and prioritises; people decide, adjust and iterate. If we preserve that division of responsibilities, we gain speed with judgement and quality time to think strategically. People first; technology second.
An invitation to reflect
- What specific human friction do you want to ease within your team or for your client?
- How will you know that your solution improved that person's life or work (what user-centred metric will you use)?
- What part should remain in human hands and which can AI take on (augment, assist or automate)?
- What risk will you eliminate or what trust will you build to make adoption sustainable?
Which problem would you start with, and what role would you give AI in your case? I'd love to hear from you — write to me.
