Most organizations think they're not ready for AI because their data is messy. That's usually not the real problem.
The real problem is that no one has ever written down how the work actually gets done.
How do we decide what gets prioritized? What actually happens behind the scenes? When does a case get escalated, and to whom? Ask five people and you'll often get five answers, or a shrug and "it depends."
You can't automate, delegate, or even safely oversee a process that only lives in people's heads. And here's the useful part: the work of defining it is valuable whether or not any AI ever touches it. Clear definitions reduce errors, make training faster, and mean the process survives when a key person leaves.
AI doesn't fail because your data is imperfect. It fails when it's asked to automate a process the organization was never clear on to begin with.
What this means for health organizations
Pick one process you're considering for AI, such as referral routing or monthly reporting, and write it down step by step with the people who do it. Wherever their descriptions differ, you've found a decision that needs to be agreed before any tool can help.
