The conversation in care about artificial intelligence tends to start in the wrong place. Providers ask which tool to use, which vendor to trust, which process to automate first. The question that actually determines whether any of it works comes earlier: is the underlying data good enough to build on?
In most care organisations, the honest answer is not yet.
The data problem nobody talks about
AI doesn’t create insight from nothing. It finds patterns in what’s already been recorded. Which means if your recording is inconsistent — different staff logging the same information in different ways, fields left blank, free-text notes that vary wildly in detail — the patterns the system finds will reflect that inconsistency back at you.
Garbage in, garbage out is an old idea. It applies directly here.
The care sector has a specific version of this problem. Data is recorded primarily for compliance — to prove something happened, not to capture it in a way that’s useful for analysis. Visit notes, care plans, incident records: they exist to satisfy auditors, not to feed algorithms. That’s not a criticism of the people filling them in. It’s a structural problem with how recording has been designed.
What good data actually requires
Before AI can add value, a few things need to be true: fields need to mean the same thing across your workforce; free text needs enough structure to be comparable; recording needs to happen consistently, not selectively. None of this is glamorous. It’s data hygiene — the unglamorous work that makes everything else possible.
The organisations that will get the most from AI in the next five years aren’t necessarily the ones investing in it now. They’re the ones cleaning up their data now.
A better starting question
Instead of “which AI tool should we use?”, try: “if we ran an analysis of the last six months of our care records, would the data be reliable enough to tell us something true?”
If the answer is uncertain, that’s where to start. Not with the tool — with the foundation the tool will need to stand on.