AI Organisational Readiness
Why most organisations aren't built to capture what AI makes possible
In my recent thoughts on what I'm seeing with our customers, I wrote about how AI has shifted the software delivery bottleneck from the development team to the decision-makers. Code is faster, but projects aren't, and the reason is almost never technical.
Microsoft's 2026 Work Trend Index puts a number on it that I think every leader should think about.
65% of AI users fear falling behind if they don't adapt quickly. Yet 45% say it feels safer to focus on current goals than to redesign how they work with AI.
This tension between the pressure to move and the pull to stay is not a technology problem. It's an organisational readiness issue, and it's the gap beneath every delivery challenge I described in my previous article. This is often true even in non-AI-related work; AI is compounding the tension around how long we wait to move forward.
What is the difference between AI adoption and AI organisational readiness?
Microsoft surveyed 20,000 workers across 10 markets and found that only 19% of AI users sit in what they call the ‘Frontier zone’ (not a term I love, but one Microsoft is now embedding in the industry), where individual capability and organisational readiness reinforce each other. About half sit in the emergent zone, where both are still taking shape. And 10% are what the report calls "blocked agency" people who have built real AI capability but work inside organisations that aren't set up to let them use it.
That last group is the one I find struggling the most. These are the people who have done the work, tested, developed the skills and are ready to operate differently and implement. But the systems and processes around them haven't caught up. The metrics, incentives, guardrails, and management norms haven't adapted to the new way of working.
The report's most important finding: organisational factors like culture, manager support and talent practices account for more than twice the reported AI impact of individual mindset and behaviour.
We have a juxtaposition of learning, models shifting daily, and organisations that aren't applying the change management needed to implement the capability in front of them.
Why is AI ROI still so hard to measure?
This is one of the most consistent questions from C-level and business leaders, and I understand why it's hard to answer. When you weigh the cost of the tooling (which comes with an experimentation tax on the business currently), the time investment in getting people genuinely productive (general learning and equipping themselves with understanding of AI, which is different from the experimentation tax), and the ongoing cost of models that keep changing, the calculation is shifting. As model costs shift from early loss-leader pricing toward commercial rates, the return depends almost entirely on whether users have the judgment to make them worthwhile.
Microsoft's research found that only 13% of AI users say they're rewarded for reinventing how they work with AI, even when it delivers immediate results. Adoption is encouraged, but transformation isn't rewarded.
The organisations that will come out ahead are what Microsoft calls Frontier Firms: those focused on AI absorption rather than just AI adoption, redesigning how work gets done and turning the output into insight that compounds over time. Most organisations are still far from this, as BAU frays their ability to tackle the imposing shifts AI is forcing.
What does AI organisational readiness require from business leaders?
Microsoft's data shows that when managers actively modelled AI use, employees reported a 17-point lift in perceived AI value, a 22-point lift in critical thinking about their AI use and a 30-point lift in trust in agentic AI. Only one in four AI users say their leadership is clearly and consistently aligned on AI.
The practical version of this, from where we sit, looks like three things:
Clear intent about why AI is being used and what outcome it's meant to drive
Quality gates deliberately redesigned rather than inherited from a pre-AI workflow
Psychological safety for people to discard work quickly, experiment and learn from what worked and what didn’t.
None of this requires a large budget or a sophisticated technology stack. It requires leadership that's genuinely willing to rearchitect how the organisation works. We’ve held this view for a long time, even when developing software before AI. If you are “rebuilding” an application, don’t just build it as you did before; you are rebuilding for a reason, and a rebuild gives you a chance to reimagine and evaluate workflows, etc. It also generally requires a good amount of change management, since most people like their comfort zone and want to keep things the way they work today; that shift takes time, and AI doesn’t change that. AI is like rebuilding from old to new; it gives you a chance to step back, reimagine how your business and workflows run today, and consider how they might work better in the future. Use AI to help you with this type of planning first before you proceed to building.
What comes next
If you haven't already, read the delivery-level version of this argument first: The faster we build software, the more the real delivery bottleneck is exposed. The organisational readiness challenge sits above it, but the two are connected.
If you're navigating this inside your own organisation and want to understand where your teams sit on the readiness spectrum, AIly is where we start that conversation with clients.
