Every technology team in the country is having the same budget conversation. AI tools deliver real productivity in places — code assistance, support triage, document work — yet the wider market is asking, loudly, whether AI valuations have run far ahead of AI revenue. The Financial Times’ recent bubble coverage captures a genuine split among investors. Inside companies, the split looks different: it is between tools that earn their licence fees and tools bought because nobody wanted to be left behind.
If a correction comes, that distinction will decide which budgets survive.
What a correction would actually hit first
History offers a useful map. In the dot-com correction, the internet kept growing while internet budgets were slashed — the spending that died was the spending that could not point to a customer. AI planning veterans expect the same filter now. Programmes tied to a measured outcome — cost per resolved ticket, hours returned to skilled staff, error rates — survive reviews. Programmes justified by fear of missing out do not.
That is why the Financial Times‘ bubble analysis matters to non-investors: it describes the exact test — revenue versus capital intensity — that boards will soon apply internally.
Related reporting on this site: How to Assess Online Health Information Before Acting on It — part of our continuing coverage for readers following how national decisions land locally.
The jobs question, without the panic
It is worth separating two fears that headlines merge. One is that AI replaces jobs; the other is that AI budgets evaporate and take jobs with them. A correction acts mainly on the second — and mostly within the AI supply chain itself: the startups, agencies and programme teams whose existence depends on ever-growing spend. For the wider workforce, the adjustment looks more like a reallocation, as money and management attention flow back to underfunded basics.
None of that is painless for those directly affected. But it is a normal capital cycle, not an employment cliff — provided companies plan the transition instead of improvising it.
How prudent companies are budgeting now
Smaller firms can apply the same discipline at kitchen-table scale. List every AI subscription and what it replaced. Cancel what nobody opened in thirty days. Renegotiate the rest annually, and keep one credible alternative in mind for each critical tool so no vendor owns your workflow. Above all, spend first on the unglamorous enablers — clean data, documented processes, staff training — because those retain their value in every scenario, including the one where AI prices fall and the tools become cheaper to use than ever.
That last point is the optimist’s version of bubble planning: corrections make useful technology cheaper. Prepared companies benefit twice.
Signals to watch in the months ahead
For anyone trying to read the cycle, analysts point to four signals. First, enterprise renewal rates: whether companies renew AI contracts at current prices when the first big cohorts expire. Second, vendor behaviour: discounting and bundled deals usually arrive before any official admission of overcapacity. Third, hiring data in the AI supply chain itself, where programme and implementation roles move earliest. Fourth, the language of earnings calls, where ‘efficiency’ quietly replacing ‘investment’ is a classic late-boom tell.
None of these signals requires a crash to matter. Even a gentle deflation — prices flat, budgets flat, hype fading — changes which projects get funded and which roles get hired. Companies watching the signals can adjust gradually. Companies watching only the headlines will adjust all at once, later, and at higher cost.
The tech team’s takeaway
Budget for the tools, not the narrative. Measure what each system earns, keep contracts short and portable, and invest in the data and training that make any model more useful. If the bubble pops, that discipline is protection. If it does not, it is simply good engineering management.
