Your AI OKR is a great start. But how will you know if it's working?
More and more businesses have an AI OKR. And honestly? That's a good thing. It signals ambition, it puts AI on the agenda at a leadership level, and it gives people a reason to take it seriously. Setting the intention matters.
But here's the question I keep asking when I work with organisations on this: how will you actually know if it's working?
Because most AI OKRs, when you look closely, are measuring inputs rather than outcomes. Things like:
% of employees who have completed AI training
Number of AI tools rolled out across teams
% of employees who have activated a licence
These are easy to track. They feel like progress. But they don't tell you anything about whether your people are actually working differently — which is the only thing that matters.
The difference between activity and adoption
There's a meaningful gap between someone completing an AI training module and someone genuinely integrating AI into how they do their job. Between a licence being activated and a person confidently using a tool to draft, analyse, summarise or plan in ways that save them time and improve their output.
Activity is easy to measure. Adoption is harder — but it's the only measure worth having.
The organisations that get AI right are the ones that get specific about what adoption actually looks like in their context. Not in abstract terms, but in observable, behavioural terms. What does it look like when someone in your marketing team is genuinely confident with AI? What does it look like for someone in operations? What are the specific things they do differently?
When you can answer those questions, you have something meaningful to measure.
So what should you be measuring?
This is where I'd encourage you to think about three levels:
Confidence and understanding. Can your people explain what AI is and isn't? Do they understand where it adds value and where it doesn't? Confidence is a prerequisite for adoption — and it's something you can assess, track and develop over time.
Practical application. Are people actually using AI to do real work — not just experimenting once and moving on? The question here isn't whether someone has used AI, but whether they're reaching for it regularly as part of how they work.
Critical judgement. This is the one most organisations overlook. Do your people know how to evaluate an AI output? Do they understand hallucination risk? Do they take accountability for the final product rather than treating AI as a source of truth? This is what separates confident, responsible AI use from chaotic, unreliable AI use.
You might also consider a fourth dimension: whether people are sharing what they learn. AI capability grows faster in organisations where people swap prompts, talk about what works and help each other get started. If that's happening, it's a signal that adoption is genuinely embedding.
The role of managers
None of this happens without managers. They are the single biggest lever in whether AI adoption takes hold or quietly fades away — and they're often the most overlooked part of the equation.
Managers need to be equipped to have developmental conversations about AI. To set expectations, model the behaviour themselves, and create the psychological safety that makes people feel comfortable experimenting and occasionally getting it wrong. If your AI OKR doesn't have a component that addresses manager capability, it's likely to underdeliver.
A question worth sitting with
If I asked you today to show me the evidence that your AI OKR is on track — not the training completion rates, but the actual behavioural change — what would you point to?
If the answer is unclear, that's not a failure. It's just a signal that the measurement piece needs as much thought as the ambition. And getting that right is entirely possible — it just requires being specific about what you're actually trying to change.
That's exactly the work we help organisations do.