The AI Buyout Thesis: Why Pilots Fail and Ownership Wins
The MIT study tells us that pilots are doomed, full ownership is only real path to AI transformation
The headlines this week were stark: MIT says 95% of corporate AI pilots are failing. Fortune and Entrepreneur both ran with the story, pointing out that despite billions being poured into AI projects, almost all have delivered “little to no measurable impact.” Cue the “AI is over-hyped” and “AI Bubble Goes Pop” think pieces (which is not to say we aren’t in a bubble — we probably are!).
On the surface, this looks like a failure of AI (and the market certainly read it that way, NVIDIA, Palantir, etc. all saw large stock price slides in the wake of this study). But it isn’t. It’s a failure of companies to properly understand and embrace AI.
The Adoption Valley Problem
As
’s excellent piece on the “AI Adoption Valley” shows, every major technology goes through a dip. PCs, the internet, smartphones — all saw early productivity declines before native workflows emerged and real gains were unlocked.That’s where we’re at with AI today. The tools are powerful, incredibly so given how long LLMs have been around. But most organisations are bolting them onto old workflows, asking AI to do the wrong things (cheap content, customer support scripts) instead of redesigning entire systems around it.
MIT’s researchers were pretty blunt on this: corporate AI fails because workflows are brittle, adoption is shallow, and companies lack the willingness to rebuild from the ground up. I agree!
Why Pilots Fail, and Why They Always Will
The reality is, most companies will never get AI transformation right. It’s not because they’re stupid, it’s because true AI transformation isn’t an experiment you bolt onto business-as-usual.
It means removing whole departments. Rewiring infrastructure. Hiring for entirely new skills. It means telling sacred cows — entire layers of middle management, for example — that they’re no longer needed (and then hiring a ton of new people, as an aside I think AI will net-net create a hiring boom). It’s hard work, brutally hard.
framed the challenges of embedding AI perfectly in his piece The Bitter Lesson versus The Garbage Can: most organisations aren’t neat machines. They’re Garbage Cans — sprawling messes of undocumented processes, workarounds, and duplication. CEOs don’t even know how half their organisations really work. That makes AI especially hard to embed because AI thrives on clarity of inputs and outputs, not chaos.Mollick’s second insight — the Bitter Lesson — makes it worse. Human process-mapping, hand-crafting, and encoding of “how we do things here” is often wasted energy. In AI, brute force computation plus outcome training beats painstaking process encoding almost every time. In other words: the more you try to make AI fit existing workflows, the less it delivers.
That’s not something most CEOs are willing to take on whilst also running the day-to-day of their core business. So they tinker. They pilot. They settle for incrementalism. And they fail.
Why AI Buyouts Are the Answer
In an AI buyout, buyers acquire companies outright and rebuild them around AI from the inside. They do the hard work because that’s the entire point of their strategy. Easy? No! But potentially extremely powerful.
Instead of pilots, you get owner-level mandates. Instead of experiments, you get actual transformation (eventually, at least). That means:
Workforce redesign — AI isn’t an assistant at the edges but the core operating system.
Infrastructure overhaul — systems and processes rebuilt for AI-native workflows, not patched legacy ones.
Outcome focus, not archaeology — don’t waste years mapping every broken process. Define the outputs that matter, then let AI + new ops models find straighter paths to them.
Margin expansion, not hype — the focus isn’t on multiple arbitrage or narrative, but on real OPEX reduction and productivity lift.
In other words, AI buyouts aren’t about sprinkling AI into existing structures. They’re about rebuilding the structures themselves — whilst knowing where not to waste effort.
The Trend of the Next Decade
Look closely and you can already see where this is headed. MIT’s own study admits that the 5% of companies that succeed do so by picking a pain point, executing deeply, and often by partnering with outside firms. That’s the DNA of buyouts: focus, execution, transformation.
Most AI pilots will fail, but AI buyouts — hard, messy, expensive transformations that require patience — can unlock true transformation. Or at least they can with the right team.
This is why I believe AI buyouts are not a niche play but the trend of the next decade.
We’re all-in on this at Thunder. We’re currently working with a handful of companies quietly initiating buyouts. We’re not at liberty to share results yet, but what I can say is: this works. Not every time — I actually believe a good chunk of the teams going after this will fail — but when executed well, this is the value capture strategy for the decade.