When a business finally decides to do something about AI, the instinct is to treat it as a technology problem. Which tool, which platform, which integration. That framing feels responsible, and it is the one most of the market is happy to sell into. It is also, more often than not, the reason so much AI spend quietly disappears.
The technology is the easy part now
The technology is largely the easy part now. The tools are mature, well-supported, and capable straight out of the box. What is hard, and what actually determines whether adoption works, is a series of judgement calls that no piece of software makes for you. Which problem is worth solving first. Where the time is really being lost. How costly a mistake in this particular task would be, and therefore how much checking it needs. Whether the work should be redesigned around the tools, or left alone.
The sceptical reading is that this is consultant's framing, and that a capable team can simply pick good tools and get on with it. Sometimes that is true. But the pattern in projects that fail is consistent, and it is rarely the model or the computing power that let them down. It is more often that the problem was fuzzy, the work was never redesigned, nobody owned the change, or the wrong thing was measured. Those are all judgements, made early, that the technology then faithfully executes or cannot rescue.
This is why I look at adoption through operations rather than through the tooling. The useful questions are operational ones: where is your time and clarity going, which one workflow would be worth winning back, and what would have to change around it for a tool to actually help. The craft sits in that design, in choosing well and shaping the work around mature tools, rather than in engineering something bespoke. Judgement governs the risk, too. Deciding where human oversight has to stay, and designing a contained, checked way of working so that risk is engineered out of the system rather than left to people remembering the rules, is a judgement call, not a feature you switch on.
Why this matters for Scotland's smaller firms
For Scotland's smaller firms, this is quietly encouraging. It means the thing that decides whether AI pays off is not access to a data-science team or a large budget; it is clear thinking about your own business, which is something a well-run small firm can do better than a sprawling one. The advantage is not technical. It is the judgement to point a capable tool at the right problem, and the discipline to adopt it properly.
So the question worth asking before any tool is chosen is not which AI to use. It is where this would actually help, and what you would have to get right around it. Get the judgement right, and the technology tends to look after itself.