Not Everything Needs AI: The Questions That Come Before the Decision

Last month at the IoD Chartered Director Conference, on a panel about governing AI, an audience member asked how I decide which AI to use. My answer was that I do not start there. I start with three questions, and by the time I have worked through them the AI question has often answered itself, sometimes by disappearing. The task before every Board is remaking the business around AI. The mandate is settled. The judgement inside it is not. The first act of that judgement is not choosing a tool. It is deciding what is even worth remaking, because the honest answer cannot be everything. Remaking everything with AI is not transformation. It is fashion with a budget. Better business judgement produces better AI decisions.

Three honest things you can do with any piece of work

Most conversations about AI begin in the wrong place. They begin with the model. Which one, how large, how expensive, hosted where, and from which provider. That is a real decision, but it is the last one, and starting there quietly assumes the decisions before it have already been made correctly. Usually they have not been made at all.

The Great Remaking is best understood as a magnifying glass held over the whole business. Hold it over any single piece of work, look at that work honestly, and there are only three things you can do with it. You can leave it as it is. You can stop doing it. Or you can remake it. Only the last of those reaches a tool decision, and even then the choice is between AI and something simpler. The three questions are how you find out which of the three you are looking at.

The first is what are you trying to do, and why. Ask it and you will usually get back a description of the thing someone has already decided to build. A system. A dashboard. An assistant. What you wanted was the outcome underneath: a decision someone has to make, a customer who needs telling something, a number the regulator wants by the end of the month. The why is what gets you from the first answer to the second. Keep asking it and the requirement tends to shrink, sometimes to a fraction of what was proposed, because the mechanism had quietly become the goal somewhere between the first conversation and the third.

The second question is how you do it today, and do you still need to do it. This is the one that does the real work, and it is the one nobody expects. Ask how something is done at the moment and you get an account of the current process, which is useful. Keep listening and you often get something more valuable, which is the sound of someone hearing themselves describe a piece of work that stopped making sense years ago. Businesses accumulate work the way houses accumulate objects, quietly and without decision. A report is produced this quarter because it was produced last quarter. An approval sits in a chain no one has revisited since the person who designed it left. A monthly reconciliation still runs to catch a discrepancy that was engineered out of the system three years ago, and nobody has thought to ask whether it is catching anything now. A committee pack grows a section every time something goes wrong and loses one only when someone retires.

None of it is doing harm, exactly, which is precisely why no one stops it. The same pattern showed up in cloud migrations, where published guidance on application portfolios put the share of an enterprise estate that turned out to be no longer useful, and could simply be switched off, at as much as 10 percent and sometimes 20. That figure is about software, not AI, and the point is the habit rather than the number. Look hard at any body of work and some of it turns out not to be earning its place. The most valuable answer this question gives you is that the work should stop, not that it should be given a better tool. Buying a clever tool for work that should have stopped is the most expensive mistake on the list. You pay to build it, you pay to run it, and you pay again in the attention it takes from work that matters. The best saving comes before any technology does.

Only then do you ask why you think you need AI. Not which AI, and not how much of it. Why any. Some work has one right answer. A balance reconciles or it does not. A count is right or it is wrong. Other work is judgement, where the best you can do is a reasoned view of an uncertain situation and there is no right answer sitting somewhere waiting to be found. The four indicator types I have set out elsewhere separate the confirmed fact from the signal, the forecast, and the reasoned conclusion, and it is the first of those that matters here. A model is the right instrument for judgement. It is the wrong instrument for a question that already has one right answer, and that is where the trouble usually starts.

The task that taught me this

I learned this from my own work. I wanted a list of tasks pulled from a system of record and shown in a file that refreshed itself. It looked like an AI job, so I built it as one. It returned a different answer almost every time it ran. The task had exactly one right answer, and I had handed a fixed question to a probabilistic tool that was entitled to work it out afresh on every pass, and did, arriving somewhere slightly different each time.

The tool was not broken. It worked exactly as designed, reached the source, and got back the right data. It still gave a different answer each time, because deciding what counted, and what counted as complete, had quietly become the model’s job rather than mine. That is the danger, and it is easy to miss. The wrong answer did not look wrong. It looked like a right answer: plausible, confident, and slightly different from the last one. If I had not already known the correct number, I would have had no way of telling which run to believe.

Rebuilt as a few lines of ordinary code reading the data directly, it returned the right answer every time, for a fraction of the cost, and it has not varied since. I want to keep the lesson from this narrow, because it is easy to inflate. I am not saying that ordinary code is trustworthy and models are not. Deterministic systems can return the wrong answer perfectly consistently. The lesson is only this: a task with one right answer should not be handed to a tool that decides it afresh on every run. There is a quieter lesson underneath it, too. The elaborate self-refreshing file I first imagined was never really the thing I needed. The simpler tool did the actual job, which means the honest answer to the second question had already reshaped the whole exercise before the tool question ever arrived.

The discipline of remaking

The remaking is real and the mandate stands. Nothing here is a retreat from that case. But remaking well is not remaking everything, and it is certainly not putting a model into everything. The three questions are simply how a Board exercises judgement inside the mandate. They stop the work that should stop. They leave alone the work that already does its job. And they spend AI where the work calls for reasoning, synthesis, or judgement, rather than where a simpler tool is better, cheaper, and more reliable.

What makes the second question a governance question rather than a procurement one is that nobody below the Board has the standing to answer it. The people closest to a piece of work are the least likely to conclude that it should stop, because their expertise, their team, and often their standing are bound up in it continuing. Asking whether a body of work should exist at all has to come from somewhere with the authority to accept the answer, and that is precisely the Board’s work. It is uncomfortable, unglamorous, and absent from almost every AI strategy deck I have seen.

The organisations getting the remaking right are not the ones putting AI into everything and calling it transformation. They are the ones doing the harder work of deciding, task by task, what to stop, what to leave, and what to remake with a model. Better business judgement produces better AI decisions, and there is no shortcut that skips the judgement and arrives at the good decisions anyway.

The willingness to say a task does not need AI, or does not need doing at all, is what makes AI investment credible, not a retreat from it.