Most AI pilots in care planning measure the wrong side of the equation.
We measure how fast the document gets written. Nobody measures whether it gets read.
Writing was never really the bottleneck. Reading was. A support worker arriving at a home has a few minutes to work out what matters about this person today. A comprehensive twelve-page plan and no plan at all fail the same way — the worker skims, guesses, and falls back on memory.
AI removes the cost of producing text. It does nothing about the cost of absorbing it. Get that backwards and you don't get better care planning. You get more of it, landing on the same worker, in the same hallway, with the same four minutes.
The useful question isn't "can AI write this faster?" It's "what would make the person at the door better informed?"
Sometimes that's a full generated plan. Often it's three sentences and one flag.
Being responsible about AI in community care means being honest that volume was already the problem. A tool that makes volume cheaper is not automatically a tool that helps.
What this means for health organizations
When you evaluate AI documentation tools, measure what happens downstream: whether frontline staff find what they need in the time they have. A short, prioritized summary is often a better goal than a longer, more complete document.
