I would never let IT run wild in a business. And I certainly would not let AI. That is a lesson learned, not a theory.
Let me be clear about what I mean, because this is not anti-technology. IT and AI are not the enemy. Letting either run unchecked, with no one senior able to question it, is. The danger is never the tool. It is what happens when a function the rest of the business does not understand is left to spend, build and decide with nobody holding the reins.
The mechanism is always the same. Most senior leaders do not really understand what their IT function does. So when IT says a project is complex, expensive and on track, the board nods, because nobody in the room can say otherwise. Jargon stops being description and becomes a shield. A bad project can hide behind it for a long time, because the people who should be challenging the spend cannot understand it well enough to challenge anything.
This is not a rare failure. McKinsey, with the University of Oxford, studied more than 5,400 large IT projects. On average they ran 45% over budget and delivered 56% less value than promised. Half massively blew their budget. And 17% went so badly they threatened the very existence of the company. Read that again. Roughly one in six large IT projects does not just waste money, it can take the whole business down with it. Across the projects studied, the overruns added up to $66bn, more than the GDP of a small country. One retailer in the data spent $1.4bn trying to modernise its systems and abandoned the lot.
I have watched it happen up close. A build that should have been manageable got wrapped in enough technical language that nobody above it could tell how badly it was going. By the time the truth surfaced, we were effectively held to ransom by a single developer, the only person who properly understood the system, and an IT lead who had buried the problem rather than raise it. It came within touching distance of taking us offline. The business was large enough to absorb the blow. A smaller one would not have survived it. I am not going to dress that up as anyone's villain story. It is simply what happens when a decision sits with people no one can question.
Now point that same dynamic at AI. If your leaders could not challenge IT, they have no chance of challenging AI. It is the same problem with a fresh coat of mystique and a faster way to spend money. The numbers are already worse. RAND found that more than 80% of AI projects fail, twice the rate of IT projects that do not involve AI. An MIT study this year found that 95% of generative AI pilots delivered no measurable return at all. Hand AI to a team you cannot question, and you have roughly doubled your odds of the disaster I just described.
The uncomfortable part is that these projects rarely fail on the technology. McKinsey's own analysis puts most of the failure down to strategy, stakeholders and alignment, not code. The fix is not a cleverer developer or a better model. It is leadership that owns the decision instead of outsourcing it to the one function nobody else can see into.
So what does owning it look like
So what does owning it look like. You do not need to learn to code, and you do not need to understand the architecture. You need to understand enough to ask hard questions and refuse to be bamboozled.
Own the business case, not just the budget. What is this for, what does good look like, and how will we know if it is working. If the answer only ever comes back in jargon, that is the warning sign, not the reassurance. Anyone who really understands a thing can explain it to you in plain English. If they will not, ask why.
Build in stage gates and independent challenge. Bring in someone who knows the technical side but works for you, not for the project, so no single person, inside or out, can hold you hostage. McKinsey's research is clear that the projects that beat the odds are the ones with proper checkpoints and business ownership, not the ones with the biggest budgets.
And never let the spend run ahead of your understanding. The moment you are approving money for something you cannot explain to a colleague, stop. That gap between what you are paying and what you understand is exactly where these things go wrong.
IT and AI are not the problem. Both are tools, and used well, both are how good businesses pull ahead. The problem is the abdication, the quiet decision by senior people that this part is too technical for them, so someone else can own it. That is the moment you stop running your business and it starts running you. Keep the decision. You do not have to understand everything. You do have to refuse to sign off on anything you do not.
Sources
- McKinsey and University of Oxford, Delivering large-scale IT projects on time, on budget, and on value (study of more than 5,400 large IT projects: 45% average cost overrun, 56% less value than predicted, half massively over budget, 17% threaten the company's existence, $66bn in total overruns, the $1.4bn retailer that abandoned its modernisation).
- RAND, The Root Causes of Failure for Artificial Intelligence Projects (more than 80% of AI projects fail, twice the rate of non-AI IT projects).
- MIT report on generative AI in business, 2025, as reported by Fortune (95% of generative AI pilots show no measurable return). Secondary source.
Ready to move the numbers.
Book a 30-minute impact discovery, no obligation. A clear read on where brand, e-commerce and AI can move growth.