Most organisations aren’t ready for AI. The uncomfortable bit is that many of them won’t find out until they’ve already spent the money.
We work with complex organisations trying to make transformation stick: NHS bodies, government departments, banks. Places where progress is rarely linear, where decisions are shared across too many people, and where the distance between a good idea and real operational change is measured in years.
So when leaders ask, “Are we ready for AI?”, they usually mean a technology question:
- Do we have the data?
- Have we identified the right use cases?
- Do we have the right tools and partners?
Those questions matter. They’re also the questions that are easiest to answer, at least on paper. And answering them creates a dangerous illusion: that you’re closer to impact than you are.
I’ve seen organisations with strong technical foundations stall after a successful pilot, with AI and other technology, because nobody could agree who owned the decision on the decision, on if we could deploy, what risks were acceptable, or who would fund the work once it moved beyond the transformation team. I’ve also seen organisations with creaking legacy systems make genuine progress because they got the organisational conditions right and chose a small number of deployments that mattered.
Tech is necessary. It’s rarely what stops you scaling.
This isn’t a call to slow down. It’s a call to speed up, but to speed up the right things first. If you rush technology into an organisation that cannot decide, cannot absorb change, and cannot govern at pace, you don’t move faster. You just fail and frustrate.
One reason I’ve come to this view is my own background. I’m a product person, and I’ve spent a lot of time working with user-centred design as an approach. Not UCD as “do some research and make the UI nicer”, but UCD as discipline around outcomes and adoption. What are we trying to achieve, how will this actually change things, and what will it be like for the people who have to live in the post-change world?
AI readiness is the same. It’s not primarily a question of capability. It’s a question of whether the organisation is set up to absorb the change AI introduces.
If AI is going to change how you operate, not just generate demos and pilots, four conditions matter more than your tech stack.
1) Clear decision rights
AI speeds up analysis and increases the number of options on the table. It doesn’t remove the need to decide. If anything, it forces decisions earlier, with less time to deliberate and more information to argue about.
If your organisation defaults to consensus, endless escalation, or “let’s run another pilot”, AI will amplify that pattern. You’ll just get faster cycles of the same indecision.
Before you deploy at scale, you need named accountable owners for each use case, explicit decision rights, and a clear risk appetite. Not a steering group. Not a working group. A person who can say yes or no, and is supported when they do.
2) Real capacity for change
Most large organisations are already running multiple programmes at once. They all draw on the same finite pool of people who can actually land change in operations: the ones who understand the workflow, can bring teams with them, and can navigate the messy reality between policy and practice.
AI transformation draws on that same pool and adds extra load on data, security, information governance, legal, procurement, and service operations.
If those people are already maxed out, adding “AI” doesn’t accelerate transformation. It creates gridlock, frustration, and a longer list of initiatives that never quite make it into day-to-day work.
This isn’t a reason to slow down. It’s a reason to choose fewer things, stop lower-value work, and back deployments that genuinely have potential to improve metrics.
3) Data that matches reality
“Garbage in, garbage out” is true but shallow. In complex organisations, the bigger issue is that data often describes the official version of the organisation, not the real one.
Workarounds, informal triage, local variations, manual fixes, duplicate recording, shadow systems, the stuff that makes services function, is frequently invisible in the record. And where it is visible, it’s often encoded in ways that reflect reporting needs rather than operational truth.
If we train models or automate decisions against the official view, you don’t just get errors. You get systems that optimise for how things are supposed to work, not how they actually work. That’s when AI produces confident answers that look plausible and fail in practice.
Readiness here isn’t only about data quality. It’s about whether you have an honest, shared understanding of your real workflows, and whether you can align the data to them.
4) Governance that can deploy, not just evaluate
In most environments, governance exists for good reasons: patient safety, financial risk, privacy, accountability. The problem is that many governance mechanisms were built for a slower cycle of change.
If every model update requires months of approvals, or if procurement can’t handle iterative delivery, you’ll stay permanently in “exploring” while the organisation builds a museum of pilots.
The organisations that move don’t abandon governance. They make it reusable and proportionate. They define risk tiers, pre-agreed controls, standard templates, clear red lines, and a practical route to production. They create a system where good teams can move quickly without improvising compliance every time.
The question leaders should be asking
Instead of “Are we ready for AI?”, ask:
What would have to be true for AI to change how we operate safely, at scale, and repeatedly?
That question forces you into the hard parts: who really owns decisions, where change capacity actually sits, whether your data reflects reality, and whether governance is set up to enable deployment rather than slow-motion assessment.
Most organisations are partially ready. They have real strengths alongside structural gaps. The ones that make progress are the ones that can see both clearly, and deal with the gaps before they try to scale the technology.
The ones who struggle are the ones that mistake pilot success for organisational readiness.
If you’re leading an organisation right now
A few moves consistently make the difference in complex organisations:
Start with decision-making, not infrastructure. You can improve data and tooling over time. Decision paralysis is much harder to fix once AI becomes another thing that can be argued about.
Treat change-ready people as a constrained resource. Map where they’re already committed, across delivery, assurance, operations, and comms, then make deliberate trade-offs before you add AI on top.
Bring governance, IG, legal, security and procurement in early as delivery partners. Not for sign-off theatre, but to design a repeatable path to production with predictable timelines and clear controls.
Be ruthless about what “progress” means. A small deployment that changes a real workflow beats ten impressive demos. If you can’t point to an operational metric that moved and stayed moved, you haven’t really started.
The pressure to show AI progress is real. But deploying AI into an unready organisation doesn’t just risk a failed project. It creates cynicism, the kind that makes the next attempt slower, harder, and more political. We’ve seen this with Agile transformation programmes.
Speed comes from removing friction in the organisation, not from running more pilots. Get the foundations right and you do not just reduce risk. You move faster towards something that actually works.
