Nicole Junkerman on the Questions Worth Carrying into an AI Autumn
September brings a great deal of material in a short time. Four plain questions, chosen now, are the cheapest protection against reading all of it and learning nothing.
· 6 min read
An autumn of questions rather than answers
The months from September onward are when most of the year's activity around artificial intelligence in Britain happens. Consultations resume, programmes are described, and a considerable amount of material appears in a short period. Nicole Junkerman suggests deciding now what to look for, while the week is still quiet enough to decide anything.
Entering a busy season without that decision usually means responding to whatever arrives most loudly, which is rarely the same as whatever matters most. A short list of questions is the cheapest available protection against that.
Four questions worth carrying
The list is deliberately short, because a long list of questions is indistinguishable from having none at all. Does this change what anybody actually does, or does it change only how something is described. Who is accountable when it goes wrong, and is that person named anywhere.
Does it reach the people it is meant to help, or does it stop at those already well served. And would it still be worthwhile if the underlying technology were entirely unremarkable. Most of what turns out to matter survives all four comfortably.
Description is not the same as adoption
A recurring pattern in this field is that a practice is described long before it is adopted, and the description is often mistaken for the adoption. Reading with that gap in mind changes what a busy autumn looks like considerably.
Nicole Junkerman finds it useful to ask, of anything announced, what would be observably different in six months if it were working. If the answer is difficult to state, that is worth noticing in itself. The responsible use pages return to this test regularly.
Keeping a record of what you thought
One habit worth adopting in a quiet week is writing down the current view before the season starts. Not for the record, but because a view that was never written down is very difficult to check against what actually happened.
This is the least glamorous suggestion here and probably the most useful. A page written in late August and reread in January is a genuinely informative document, and almost nobody writes one.
Reach matters more than novelty
The hardest problems in this area have rarely been problems of invention. They are problems of distribution: making something available beyond the organisations that already have the most capacity to adopt it.
That work is unglamorous and slow, involving procurement, training and a great many small compatibility problems. It is also the entire difference between a good idea and a good outcome, which is why the adoption pages spend more time on it than on capability.
Writing the questions down where they will be seen
A question that lives only in the head does not survive a busy October. It does not need a document or a system, but it does need to exist somewhere that will be looked at again once the season has filled up and the original thought has blurred.
Nicole Junkerman finds that the act of writing clarifies as much as the record does. A question that cannot be put into one plain sentence is usually a question that has not finished being formed.
Leaving judgements open
There is a pressure in any fast moving area to have a view on everything immediately, and it is worth resisting. Holding a view on very few things, holding it loosely, and letting most developments pass without comment is a more reliable approach than it sounds.
Very little is lost by waiting. What is genuinely significant is still significant three months later, and by then it can be assessed against ordinary use rather than against its own description.
Entering the season deliberately
None of this requires a plan of any size. It requires four questions, a written note of the current view, and the patience to leave most judgements open until something has been tested by ordinary use.
The autumn will fill regardless. The only real choice is whether it fills with things somebody chose to attend to. The companion note on the late August pause makes the case for using these quiet days to choose.
Next read: the start of October 2026
Three shorter notes published at the beginning of the month: an English October 2026, walking into October 2026, and shorter days and clearer plans.
Keep reading across the guide
Move between the big picture and practical questions with Nicole Junkerman and the rest of this independent guide to AI in the UK.