The Great AI Over-Pivot
Inside the psychology of over-pivoting
Every company I’m involved with seems to be pivoting every six months — each time unveiling a new “AI layer,” “AI-driven” direction, or “GenAI-powered” product.
It’s an understandable impulse.
The speed of change in AI right now borders on manic. At the start of this year:
Google Cloud AI Agents didn’t exist.
Legal AI tools still underwhelmed.
DeepSeek wasn’t even part of the mainstream zeitgeist.
But have we swung too far?
The Cult of the Perpetual AI Pivot
I’ve watched companies abandon three years of disciplined B2B focus to chase a consumer play with an LLM front end.
I’ve seen others bolt their AI onto markets they barely understand, just to claim the “early entrant” badge.
Are these over-pivots working? Most are not.
Pivoting is hard even when it’s done right. My current company has been in a 12-month pivot. It took six months of focused, sometimes painful, work before we started seeing the beginnings of a payoff.
“AI over-pivoting” collapses under its own weight.
Pivoting is hard because a successful pivot isn’t just about changing your product — it’s about re-engineering everything:
marketing positioning
sales playbooks and training
operations and workflows
customer support
sometimes even your financial models and KPIs
And that’s why “AI over-pivoting” collapses under its own weight.
The Bubble We Pretend Not to See
All this exuberance feels eerily familiar. I’ve been posting about signals of an emerging AI bubble. Anecdotal, yes.
So now, I’m shamelessly leaning on the insights from Economist Paul Krugman—infinitely more data driven than I—in his recent Substack essay “Technology Bubbles: Causes and Consequences” (October 2025). (Pay the price and subscribe. It’s worth it!)
With deep analysis to support his thesis, Mr. Krugman compares our current GenAI euphoria to Britain’s railway mania of the 1840s and the telecom bubble of the 1990s — both of which ended in massive losses and a trail of broken companies.
My favorite line from Krugman’s essay:
“Asset bubbles are natural Ponzi schemes.”
Early investors make out like bandits. Late entrants are left holding the bag.
And the operators — the founders grinding to keep up with the hype while trying to build real businesses — often become collateral damage.
The AI Ponzi Mindset
To be clear, I’m not suggesting every AI investment is intentionally trying to defraud us. (We can leave that to the crypto crowd 😊!)
But the human psyche doesn’t seem to be able to avoid a good bubble bursting.
Krugman again: the biggest bubbles in history shared one significant element — the race for monopoly rents.
“in order to capture monopoly rents you want to get there first and build the greatest capacity, thereby deterring a competitor from entering the market.”
I’d argue that the race to create the infrastructure for data centers is to the AI bubble as fiber was to the telecom bubble, and railway tracks were to the railroad bubble.
And that’s why we’re seeing these “exuberant” valuations for AI chip companies (ahem, Nvidia), data center providers, and real estate investment trusts.
Because it’s a race to cornering the market.
McKinsey reports that companies will invest almost $7 trillion in capital expenditures on data center infrastructure globally (assuming no bubble bursting pull back).
If We Know We Are Heading for a Cliff, Why Aren’t We Slowing Down?
So why don’t founders and investors slow down?
Because the incentives are upside-down.
If you’re a tech leader and you overinvest in AI and it crashes — well, “everyone did it.” You don’t stand out as the “loser”.
But if you don’t chase this big payoff—and end up missing The Next Big Thing … You’re out of a job.
Today, like all bubbles before this, everyone is running with the herd.
Be a Snow Leopard, Not a Buffalo
Buffaloes move in herds.
Snow leopards chart their own independent paths.
Today, like all bubbles before this, everyone is running with the herd.
I’m not saying ignore AI. Far from it. AI is here to stay — whether your business is or not.
But founders who survive hype cycles (like the one we are currently in), and build enduring companies, do one thing differently: they are more circumspect in how they evolve their business to a Gen AI model.
They adopt new technology because it solves a real problem, not because it checks an investor deck box.
They experiment quietly before they announce loudly.
They focus on operational lift before narrative lift.
And while they may not be the ones that hit the headlines in the next 2 years with tens of billions of dollars in valuations, they will be, IMHO, the ones that are still showing steady growth after the “AIpocalypse”, where we have no electricity or water to run our washing machines, but we’ll still be able to generate yet another deepfake Taylor Swift video.
The survivors won’t be the ones that sold AI-fast.
They’ll be the ones that built AI-smart.
An AI Pivot Checklist
If you’re assessing your own AI strategy, ask:
Is our AI feature solving a real customer problem, or just showcasing capability?
Can we measure the incremental value of our additional AI functionality — or are we measuring marketing hype?
Does this AI pivot strengthen our core business model, or distract from it?
Have we adapted the rest of the organization — not just the product — to this new direction?
If you can answer yes to all four, you’re leading the curve.
If not — maybe it’s time to slow down, breathe, and think like a snow leopard.
