The most dangerous moment in an AI transformation isn't when you make a wrong move. It's when you don't know you're making one.
The most dangerous moment in an AI transformation isn't when you make a wrong move.
It's when you don't know you're making one.
Most of the senior leaders I talk to arrive at the same problem from different directions. Some have been told by their board to "do something with AI." Some have watched a competitor make an announcement and aren't sure whether to panic or dismiss it. Some have already run a pilot — it went fine, everyone said the right things, and then nothing changed.
All of them are dealing with the same underlying problem: they don't have a clear picture of where they actually are.
Not where they want to be. Not where they think they should be. Where they actually are — right now, with their specific team, their specific operations, their specific moment.
That's the starting point for everything. And it's the part most companies skip.
When a leader tells me they don't know where to start with AI, I take that seriously — not as a gap in their knowledge, but as an accurate description of a genuinely complex situation.
They don't know:
What AI can realistically do for their type of business right now, not in a demo, not in a case study from a different industry, but in their actual operations with their actual team.
What their team's real relationship with new technology looks like. Not what people say in meetings. What actually happens when something new gets introduced.
Which of their current processes are genuinely worth automating, and which ones only look like low-hanging fruit from the outside.
What they'd do with the time if certain things did get faster.
These aren't easy questions. They're the right questions. And most AI strategy advice never gets to them, because it starts from a different place — usually from a tool, a framework, or someone else's success story.
Here's a useful way to think about it.
Imagine you're trying to plan a route somewhere. You know roughly where you want to end up. But your map is wrong — some roads marked as open are closed, some that look fast are actually slow, and a few of the most useful shortcuts aren't on the map at all.
You could still plan a route. You'd just be planning based on a map that doesn't match reality.
This is what most AI strategies look like. They're carefully constructed routes based on an inaccurate map. And the reason the map is inaccurate isn't because the leaders involved are uninformed — it's because they haven't yet had the conversations that would tell them where the map is wrong.
The conversations I mean are specific. Not "what are our goals" or "what's our AI vision." Those conversations produce documents, not maps.
The conversations that produce accurate maps sound more like:
Where does time actually go in this team's week? Not where it's supposed to go — where does it actually go?
What breaks, regularly, that nobody has time to fix?
What would people do differently if they weren't spending half their day on [insert the thing that everyone knows is inefficient but nobody has addressed]?
When we've tried to change something before, what actually happened?
These are operational questions. They require talking to people other than the leadership team. They produce answers that are specific, sometimes uncomfortable, and genuinely useful.
Starting from your actual situation means resisting the gravitational pull of the general case.
The general case is everywhere right now. Every conference, every consultant, every tech vendor has a story about how Company X used AI to do Y and got Z result. These stories are real. They are also, almost always, irrelevant to your specific situation — because they're about someone else's map.
A real starting point has three things:
An honest picture of where you are. Not where you wish you were, not where you think you should be. What your team can actually do, what your operations actually look like, what your real constraints are. This requires talking to people who will tell you things you don't want to hear, and asking questions that don't have impressive answers.
A specific problem worth solving. Not "we should use AI more" — that's a direction, not a problem. A specific problem sounds like: we spend forty hours a week on a process that produces outputs nobody acts on or our customer support team answers the same twelve questions three hundred times a day. When you find problems that specific, the question of whether AI helps with them becomes answerable.
A realistic view of what change actually costs. Not the technology cost — that's usually the smallest part. The real cost is the time it takes to train people, change habits, debug the gap between what a tool promises and what it delivers in your context, and maintain momentum when the results take longer than expected to appear. Most AI pilots underestimate this by an order of magnitude.
None of this is technically complicated. You don't need to understand how large language models work. You don't need a data science team or a cloud migration or a dedicated AI task force.
What makes it hard is that getting an accurate map requires being honest about some things that organisations are usually not very honest about.
How decisions actually get made. How resistant certain parts of the business are to change. What the leadership team actually prioritises when resources get tight. Where the bodies are buried, metaphorically speaking.
I've sat in a lot of these conversations. The ones that go well are the ones where someone in the room is willing to say: actually, that's not quite right — here's what's really happening. Those moments are where the real map starts to emerge.
The ones that don't go well are the ones where everyone agrees with each other. Where the map stays neat and the routes stay clean and nobody mentions that half the roads are closed.
If you're trying to get oriented and don't know where to begin, this is what I'd suggest.
Pick one area of your business — not the whole thing — and spend two weeks finding out what actually happens there. Not what the process documentation says. Not what the last all-hands presentation showed. What actually happens, day to day, according to the people who do the work.
Ask: Where does time go? What breaks? What would you fix first if you had a week and no other obligations?
You'll find the map is different from what you thought. Some roads you thought were open are closed. Some shortcuts nobody told you about. A few places where you expected problems don't have them, and a few places where you assumed things were fine turn out to be where most of the friction lives.
That's your starting point. Not a vision, not a framework, not a strategy deck. An accurate map of a small piece of territory.
From there, the question of what AI might actually do for you becomes specific enough to answer.
Everything in an AI transformation that goes wrong goes wrong because someone was working from an inaccurate map. Everything that goes right goes right because someone took the time to find out what was actually there.
That's not a technology problem. It never was.
If you're a senior leader trying to figure out where you actually stand — not where you should be, but where you are — that's the conversation I'm here for.
Berk Bayri
Independent AI Transformation Advisor
I help leaders navigate AI transformation — starting from their actual situation, not a framework. 27 years of professional work across design, code, product, and AI systems in production.
Work with me →When there's something worth reading — not before.