leadership August 24, 2026 By Blake Colman

Pearl Meyer’s 34% AI Problem Puts $2.5T of Spending at Risk

If only 34% of your executive team knows who owns AI decisions, you do not have an AI strategy. You have an expensive workplace argument.

Only 34% of C-suite executives say it is consistently clear who makes the calls on AI. Yet worldwide AI spending is forecast to hit $2.52 trillion in 2026.

That is not a technology story. It is a leadership failure with a massive invoice attached.

I have seen this movie in businesses long before anyone called it AI. A company buys the shiny thing, announces a transformation, gives everyone a login, then discovers six months later that nobody owns the commercial result. The tech team owns the platform. The legal team owns the risk. Finance owns the budget. HR owns the training. The CEO owns the PowerPoint. And the customer, unsurprisingly, gets bugger-all benefit.

The expensive gap between ambition and accountability

Pearl Meyer’s Q2 2026 survey of 116 board members, CEOs, C-suite executives and senior managers found a properly awkward disconnect: the people closest to implementation are the least convinced anybody is actually in charge.

Just 34% of C-suite executives said AI decision ownership was consistently clear. Directors were more confident, at 53%, and executives below the C-suite came in at 57%.

Read that again. The people furthest from the operational mess think the job is more sorted than the people expected to make it work.

That is classic boardroom theatre. Everyone nods at “AI strategy” because nobody wants to sound behind the times. But “strategy” without a named owner, a defined commercial problem and permission to kill bad projects is just corporate cosplay.

The spending makes this dangerous. Gartner forecasts total worldwide AI spending of $2.527 trillion this year, up 44% from 2025. Infrastructure alone is expected to account for $1.366 trillion. Companies are building the engine room at record speed. The question is whether management has bothered to decide where the bloody ship is going.

Pearl Meyer’s research puts the human problem in plain English: AI ambition is moving faster than the leadership system required to execute it. The survey found that 80% of CEOs believe their executives can make enterprise-wide trade-offs, while only 30% of other C-suite leaders agree. That is not a minor difference of opinion. It means the CEO may believe the leadership team is acting like one company while the leadership team experiences the place as competing departments with better slide decks.

Why “the CIO owns AI” is the lazy answer

The worst response to this is to dump it on the CIO or CTO and call the problem solved.

Technology leaders should own technology architecture, vendor choices, data access, cyber controls and reliability. That is their lane. But they cannot own every business decision that AI touches. They should not be deciding whether a sales workflow is worth changing, whether customer-support quality has improved, which jobs need redesigning, or whether an automation project creates more margin than misery.

Likewise, a chief AI officer with no authority over budgets, operating processes or business-unit leaders is often just a very well-paid concierge for software demos.

Real ownership has to be split properly, not vaguely. The CEO owns the enterprise priorities and trade-offs. A single executive owner owns each use case and its measured result. Technology owns whether the system is safe, reliable and integrated. Finance owns whether the numbers stack up. Legal and risk set the guardrails. HR owns the changes to roles, incentives and capability.

That may sound obvious. Good. Most useful management is obvious. The problem is that obvious things are rarely done with enough discipline.

Fortune’s reporting on the Pearl Meyer findings captured the other trap nicely: handing out ChatGPT or Copilot licences is easy; redesigning the surrounding workflow so the tool creates value takes much longer. I would go further. Buying licences is not adoption. A staff survey saying people like a tool is not ROI. A hundred pilots are not innovation if none becomes a standard operating process with an accountable owner.

The board and CEO are looking at different companies

The most revealing numbers are not even about AI.

Pearl Meyer found that 88% of CEOs and 79% of C-suite executives believe achieving strategic goals will require significant changes to how their organisation operates over the next three years. Only 42% of directors agreed.

That gap should make every founder and director sit up.

Boards commonly see capital allocation, risk dashboards and progress updates. Management sees the operating reality: brittle systems, missing data, overloaded middle managers, undertrained teams, conflicting incentives and customers who do not care how clever your model is. A board that believes the company is fundamentally set up for the future while management knows it requires major rewiring is a board preparing to be surprised.

The same mismatch appears on change fatigue. Sixty-three percent of CEOs believed employees could take on more organisational change without being stretched too thin. Only 33% of C-suite leaders agreed, alongside 40% of non-C-suite executives.

Now, I am not arguing that leaders should spare people every difficult change. That is how good businesses become museum pieces. I am saying the CEO needs an unvarnished view of capacity. Every new AI rollout consumes management attention, data-cleaning effort, training time, process redesign and trust. If you pile five transformations onto a team already carrying three, you do not get eight transformations. You get quiet resistance, shortcuts and a bunch of initiatives that die in committee.

The overlooked angle: this is a decision-rights problem, not an AI problem

Here is the contrarian bit: the companies most at risk are not necessarily the slow adopters. They are the frantic adopters with fuzzy decision rights.

A slower business that chooses two valuable problems, assigns serious people and measures the result can beat a faster business that buys every tool and calls the chaos progress. Gartner itself says organisations are increasingly prioritising proven outcomes over speculative potential, and that enterprise scaling requires more predictable ROI.

That is the grown-up view. It is not anti-AI. It is anti-waste.

Look at the practical example from Yum Brands. Its technology chief, Jim Dausch, did not start by trying to replace every worker with a robot. At Pizza Hut, his team improved the order flow so kitchens had a better chance of making food when a driver would actually be available. The reported outcome was hotter deliveries and higher customer-satisfaction scores. Across Yum’s 63,000 locations, the company is also clear that franchisees will resist technology unless it improves customer experience in a way that lifts sales, or reduces food waste enough to pay for itself.

There is your template. Start with the operational bottleneck. Put a number on the outcome. Give an operator responsibility. Then decide if the technology earns the right to scale.

Not every AI project needs a grand enterprise committee. But every material project needs a single throat to choke and a scorecard that cannot be spun. If the owner cannot tell you, in one minute, what changed in revenue, margin, cycle time, error rate, risk or customer retention, the project is not ready for more money.

The CEO’s actual job now

Dataiku found that 80% of global CEOs believe their job is at risk if AI fails by 2026, while 81% of US CEOs expect a peer to be ousted after an AI failure or crisis. That sounds dramatic, but the logic is simple: boards and investors have been told AI is strategic. Eventually they will ask where the returns are.

The chief executive’s job is not to become the company’s best prompt engineer. It is to stop ambiguity breeding in the executive team.

A capable CEO should be able to answer five blunt questions:

1. Which three AI use cases matter commercially this year? Not 30. Three. 2. Who owns each result by name? A title is not a name. 3. What baseline are we improving? You cannot claim productivity if you never measured it. 4. What authority does the owner have to change the workflow? Tools without process authority produce theatre. 5. What do we stop doing if this works? If nothing stops, you have added cost rather than created leverage.

That final question matters most. AI should remove work, reduce errors, speed decisions or improve the customer experience. If it merely creates another dashboard, another approval layer or another meeting to discuss “insights”, congratulations: you have invented an expensive new form of admin.

What this means for you

Whether you run a startup, lead a division or manage a small team, steal this for tomorrow morning: make a one-page AI ownership register.

List every live AI initiative. Against each one, write the executive owner, the operating owner, the specific business metric, the current baseline, the next 90-day target, the total cost, the key risk and the date it will be stopped if it fails.

Then take the list into your next leadership meeting and ask each owner to explain their project without using the words “transformation”, “enablement”, “ecosystem” or “journey”. That alone will save you a fortune.

If nobody owns the number, cancel the project. If the number is not worth moving, do not fund it. If three executives claim ownership, none of them does.

The winners from the $2.52 trillion AI spending boom will not be the firms with the flashiest tech stack. They will be the ones where a smart operator owns a real business problem, has authority to change the work, and gets judged on the result.

Everything else is just an expensive argument in a nicer font.