Deloitte's CTO Admits What We Already Knew About AI Spending
How to Waste Millions on AI
Deloitte’s latest research confirms companies are buying jet engines and duct-taping them to bicycles. Worker trust in corporate AI has collapsed by 38% as a result.
The £5,000 Paella That Never Was…
Last summer I spent an obscene amount of money on a fancy espresso machine because I’d convinced myself that owning the right equipment would automatically make me good at something. The logic was flawless: professionals use this exact model, therefore purchasing it would transfer their competence directly into my incompetent hands through some sort of chrome-plated osmosis.
Three months later, the thing sits on my counter producing coffee that tastes like dishwater squeezed through a sock, and I’m back to instant because at least that comes with instructions I can follow.
Which brings me, in a roundabout way, to the fact that companies are currently spending 93% of their AI budget on shiny technology and precisely 7% on the humans expected to operate it.
The Diagnosis: Throwing Money at the Wrong Problem
This delightful statistic comes from Bill Briggs, Deloitte’s Chief Technology Officer, who admitted he was genuinely surprised by this ratio when he finally quantified what he’d been feeling in his travels. Surprised. As if decades of watching corporations buy expensive solutions to problems they don’t understand hasn’t been the entire plot of modern business.
The 93-7 split means companies are obsessing over the ingredients, the models, the chips, the software, whilst completely ignoring what Briggs calls the “recipe”, which includes that tedious business of culture, workflow, and actually teaching people how to use any of this expensive rubbish. He compared it to trying to make paella but ending up with just cilantro. Which is optimistic, frankly. Most companies aren’t even getting cilantro. They’re getting a pile of uncooked rice and a vague sense that Spanish food was meant to be involved somehow.
I watched this exact stupidity play out across four decades in the tech industry, where the pattern was always identical: buy the most expensive solution first, then act shocked when nobody knows how to use it. The logic is that if you spend enough money, competence will simply materialise through the sheer gravitational pull of your budget. It won’t.
The Mechanism: Fear, FOMO, and Frozen Indecision
Here’s what’s actually happening in boardrooms right now. CEOs and boards are paralysed by something Briggs identifies as “buyer’s remorse” in a market moving at breakneck speed. They’re terrified of committing to an AI vendor today only to discover a better model releases tomorrow. Or next week. Or the week after that, when the next big announcement drops, and suddenly their multi-million-pound investment looks like a museum piece.
So they delay. And whilst they’re delaying, they’re still spending money, because doing nothing looks worse than doing something stupid. But they’re spending it on the easiest thing: the technology itself. Because buying software is straightforward. You sign a contract, cut a cheque, and you can tell shareholders you’re “investing in AI transformation.” Job done.
What you can’t do easily is explain why you need to spend months training your workforce, redesigning workflows, and fundamentally rethinking how your organisation operates. That’s messy, slow, and requires admitting that your current processes are probably rubbish. So instead, companies bolt the new technology onto old workflows like duct-taping a jet engine to a bicycle and wondering why nobody wants to ride it.
The consequences are already visible. Despite increasing access to AI tools in the workplace, overall usage has actually dropped by 15%. And here’s the bit that should terrify every CTO: 43% of workers with access to corporate AI are bypassing employer policies to use unapproved tools instead. These “Shadow AI” users report that the unauthorised consumer tools are “easier to access” and “more accurate” than the expensive corporate solutions their companies spent millions implementing.
Worker trust in corporate AI collapsed by 38% between May and July of this year. Your employees would rather risk disciplinary action using ChatGPT on their phones than use the approved system you spent a year rolling out. That’s not a technology problem. That’s a catastrophic failure of implementation.
What You Can Actually Do About This (Since Nobody Else Will)
First, accept that the 93-7 ratio is you. If you’re involved in any AI implementation at your company, you’re probably watching exactly this stupidity unfold in real time. The first step is admitting you’re part of the problem, which I realise is deeply unpleasant but necessary.
Second, flip the spending ratio before you buy anything else. Not literally to 7-93, because that’s equally stupid, but at least aim for 50-50. For every pound spent on technology, spend a pound on making sure someone can actually use it effectively. This means proper training, not a 90-minute webinar where someone reads PowerPoint slides at people. It means redesigning workflows before implementing the tool, not after. It means accepting that change is slow, uncomfortable, and requires admitting your current processes might be terrible.
Before your next AI purchase, ask yourself three questions:
Can you explain in one sentence what problem this solves?
Can the people who’ll use it daily describe how it’ll change their workflow?
Have you allocated actual budget and time for training, not just “we’ll figure that out later”?
If the answer to any of these is no, you’re about to become another data point in the 93-7 statistic.
The workers using Shadow AI aren’t rebels or idiots. They’re telling you that your approved solutions are worse than free consumer tools. Listen to them. Find out which unauthorised tools they’re using and why. The gap between what you provided and what they’re actually using is a detailed map of everything you got wrong. Don’t punish them for finding better solutions. Learn from the fact that they bothered to look.
Finally, stop waiting for the perfect moment. Briggs compared the CEO paralysis to trying to time the stock market perfectly, and he’s right. Every week you delay because a better model might be coming is another week your competitors are learning how to actually use this technology. The fastest path to competence is starting with whatever’s available now and learning from the inevitable mistakes.
Perfectionism is just cowardice dressed in business casual.
The Expensive Metaphor, Revisited
That espresso machine still sits on my counter, a monument to the belief that equipment creates competence. I could’ve spent a fraction of the money on a basic machine and actual lessons on how to use it properly. But that would’ve required admitting I needed help, which felt embarrassing at the time.
Your company’s AI strategy is currently that espresso machine. Expensive, impressive, completely useless because nobody knows how to operate it. The difference is that my mistake cost me a few hundred quid and marginally wounded my pride. Your mistake is costing millions and wounding your entire workforce’s trust in management’s ability to make sensible decisions.
But at least the PowerPoint slides about your AI transformation look spectacular.