It doesn’t take much for an idea to wedge itself into the collective consciousness where it calcifies into an accepted ‘truth’. The problem is, there are lots of things that are true – but also come with heavy caveats. Or are only truish. There are others still that are questionable, or simply false. History is replete with them, some are highly damaging, others amusing.
Consider the belief that ‘10,000 steps a day’ is a scientific benchmark for health, when it was actually made up for a pedometer advert. Now consider how embedded the idea has become that ‘95% of AI projects fail’. You have almost certainly heard the statistic, which comes from a widely cited 2025 MIT report on enterprise AI adoption.
The research was sound within its parameters – specifically measuring pilots that failed to show direct P&L impact. But as the stat has circulated, it has hardened into a cultural belief and is often used as evidence by detractors that the technology itself is the issue.
I would argue that we’re missing the nuance in what these figures ‘reveal’; the bigger problem is organisations need to radically change how they think about AI full-stop.
Faster horse… or donkey?
I have yet to see a business deliver ‘transformation’ just by coding faster. I have seen organisations, however, use AI to discover problems early – and treat that discovery as a success in its own right. Here, it can expose hidden technical debt, such as continuity risks, duplicated technologies and architectural weaknesses that limit business performance, which humans often miss.
The 95% failure rate would drop steeply if AI were brought in earlier in the transformation process as a diagnostic tool. Yet too many projects still ‘fail’ because so many businesses are locked into viewing AI purely as a productivity tool; something to cut costs on admin or coding.
This persistent view sets all the wrong expectations because it’s only once these structural issues have been ironed out that AI should be applied to the more sophisticated and practical aspects of a business’ day-to-day operations.
My team, for example, worked recently with a client for which we deployed an AI agentic platform to map everything inside the organisation. The system flagged a regulatory compliance flaw hiding right under their noses while identifying a six figure number in unused cloud spend within its first week. The discovery paid for itself instantly, not through automated output, but through visibility. That demonstrates that AI’s biggest early return isn’t automation, but uncovering issues humans struggle to identify.
Business strategy comes first
As the stage shifts from diagnosis to remedy, it pays to take stock before rushing in to ‘turbocharge’ the business. We must abandon the quest for blanket miracles and focus on a ‘thin slice’ strategy which executes just one high-impact, executive-sponsored use case to demonstrate clear value before scaling.
The simple fact is we can’t do everything at once. Transformation should be iterative with each small progression setting up the next and building towards a longer-term business objective.
However, without a clear destination in sight and a structured roadmap to reach it, the risk is that money and time will be wasted. Businesses can’t be endlessly experimenting, hoping something works; a behaviour many indulge in as they panic that competitors are firing ahead of them, when in reality they’re very likely in the exact same predicament.
We must remember that AI is not priced like traditional software, which usually carries a simple fixed cost per user licence. Enterprise level AI often uses consumption-based pricing models meaning costs scale directly with higher processing. That makes judicious and strategically meaningful use even more important before deploying AI across the business.
Success, therefore, isn’t defined by the tech you buy, but by the questions you ask – and these should start with two fundamentals: which core organisational challenges do we need to solve, and in what order?
Ultimately, AI will not save anyone any money if it is used as a lightning fast typewriter and increased productivity doesn’t necessarily equate to the types of transformation that enable long-term growth. Leaders need to stop using AI to do the wrong things faster, and start by using it to tell you what needs fixing first and then align this to clear business strategies that are timetabled, measurable and, most of all, achievable.
