An analysis by US insurers (Blue Cross Blue Shield) puts the extra cost of hospital AI medical-coding tools at $942 million over two years. The mechanism: the AI documents patients as having more 'complex' conditions, with no real change in the care delivered — a 'clear disconnect between coding and treatment.' On one side hospitals automate billing; on the other insurers automate disputes: 'bots fighting bots, agents fighting agents.'
The lesson goes beyond healthcare. AI saves nothing by magic: it amplifies whatever you ask it to optimise. Point it at 'maximise billing' and it maximises billing. At Devside, we always start from the real question — what outcome for the user, at what actual cost? — before bolting AI onto a feature. Measure the impact, keep a human on the decisions that matter, and refuse AI when it just shifts the problem around: that's the difference between AI that helps and AI that costs.