There is a conversation happening in boardrooms right now that sounds something like this.
The CEO opens: "We need to be doing more with AI." Everyone nods. Someone from technology mentions a pilot they read about. The CFO asks about ROI. A consultant in the room quotes McKinsey. Everyone leaves the meeting feeling like progress was made.
Six months later, people have been very busy, but little has changed.
I have been in those rooms. I have also been in the rooms where something different happened, where organisations moved with genuine clarity, made decent decisions, and built AI capability that actually stuck. The difference between those two rooms is not budget, and it is not the sophistication of the technology. It is something far more basic: the ability to answer the question "what problem are we actually trying to solve, and do we have what it takes to solve it here?"
That sounds obvious. It is surprisingly rare.
The urgency is real, but it is creating the wrong pressure
I am not going to tell you AI is overhyped. It is not. The economic case is well-established, the productivity gains in targeted use cases are genuinely significant, and the gap between organisations that move now and those that wait is widening. That part is real.
What is also real is that urgency, without clarity, produces waste.
Organisations often invest six figures in AI tools without first establishing a clear understanding of their data landscape. I've been involved in companies launching AI pilot projects in customer-facing processes where the underlying workflows are so inconsistent that the models struggle to generalise effectively, the customer impact was real. Additionally, governance and compliance are frequently neglected during the early stages of planning, leading to significant challenges later on when regulators begin to raise questions.
None of this is stupidity. It is what happens when the pressure to "do something on AI" outpaces the structural thinking needed to do it well.
The five things you need to understand before you commit
Over the last few years, I have developed a fairly consistent view of what separates AI investments that return genuine value from those that become shelfware or expensive lessons. It comes down to five questions, and most organisations cannot answer all five cleanly when they start.
1) What is the business value, and how will you measure it?
Not "AI will make us more efficient." Something specific. Revenue impact. Cost reduction in a defined process. Cycle time across a named workflow. If you cannot quantify what good looks like before you begin, you will not be able to evaluate whether you got there. This is also the question that forces prioritisation. You cannot and should not pursue every AI opportunity simultaneously. You need to know which ones move the needle most, and in what order.
2) Is your data actually ready?
This is the question that most deflates AI ambition when answered honestly. Data readiness is not just about volume. It is about quality, consistency, governance, and access. I have seen organisations with genuinely impressive datasets that were unusable for AI purposes because data ownership was fragmented, definitions were inconsistent across business units, or access rights were so locked down that getting data to where it needed to be took months. You do not solve this after you start building. You need to understand it before.
3) How complex is the technical integration?
There is a meaningful difference between an AI use case that runs on clean, structured data through a modern API-connected architecture and one that requires extracting information from legacy systems, reconciling formats, and building new data pipelines from scratch. Both might be worth doing. But they carry very different cost, risk, and timeline profiles. Knowing this upfront shapes your phasing, your investment case, and your sequencing.
4) What is your regulatory and governance exposure?
Regulation is moving fast, and the cost of retrofitting compliance into an AI system that was built without it is significant. This is particularly sharp in financial services, insurance, healthcare, and any organisation handling personal data at scale. GDPR obligations do not pause because you are excited about a use case. Emerging AI governance requirements, including sector-specific rules that are coming even if they have not landed yet, need to be in the room when you build your roadmap, not reviewed as an afterthought six months into deployment.
5) Does this actually align with where the organisation is going?
This one is softer, but it matters. I have seen technically sound, commercially viable AI initiatives stall and fail because leadership appetite was not there, because the initiative sat at odds with a strategic priority elsewhere, or because the step change required was more than the organisation was ready to accept. Strategic alignment includes executive sponsorship, readiness for change, and the honest question of whether this is the right time. Sometimes the answer is not yet, and that is a legitimate finding.
The speed trap
The pressure to move fast on AI is understandable. Nobody wants to be the organisation that missed it. But the organisations I have seen move the fastest, successfully, are not the ones that skipped the thinking. They are the ones that compressed it: they did the structural work quickly, got to clarity fast, and then executed with confidence because they knew what they were doing and why.
The organisations that moved fast without thinking mostly have a collection of disconnected pilots, a data mess they are still untangling, regulators looking concerned, and a board that is increasingly sceptical about whether AI is delivering anything.
Speed and rigour are not opposites. But you do have to choose to do the rigour, and you have to do it early.
What a proper starting point looks like
Structured AI assessments are not a new idea, but they are unevenly executed. Done well, a good assessment should do four things. It should surface the real opportunities across your organisation, not just the obvious ones that people in the room already know about. It should score those opportunities honestly against value, feasibility, data maturity, regulatory risk, and strategic fit. It should produce a sequenced, phased roadmap that is investable and realistic. And it should do all of this quickly enough to be useful, not as a six-month consulting engagement that produces a report nobody reads.
I have carried out exercises like this in five days. Through structured interviews with senior leaders across every function over three days, a prioritised heatmap of ten to fifteen scored use cases, a 90-day quick-win plan, an 18-month roadmap with indicative investment ranges, and a board-ready presentation at the end of it. While it may not be possible to do this in five days for every organisation, the approach is valid and scalable.
The point is not to produce documents; don't go down the analysis rabbit hole on this. The point is to get your organisation to a clear, defensible decision about where to go with AI and in what order. AI-enabled solutions are best delivered through experimentation, and with the acceptance that you can fail fast and pivot.
If that is where your organisation is right now, you can find out more by reaching out to us in the comments.
A final thought ...
Here is the thing that I keep coming back to when I talk to senior leaders about AI. Most organisations have some version of an AI plan. Very few have a genuinely honest picture of their AI readiness.
The plan is the easy part. It is the discipline to stop, look clearly at where you are, and build a sequence that actually works from your current position, not an idealised version of it, that most organisations find difficult.
AI will not wait for you to be ready. But moving without knowing where you stand is not speed. It is just noise.
The difference between the two could be worth five days of your time.
No comments yet