Your AI Strategy Is Quietly Destroying Your Firm’s Value

Like many Founders, I’ve spent hundreds of hours researching, using, testing, and talking to my peers about AI the last 12-18 months.

In those hundreds of hours, I’ve come to notice one fork in the road that’s quickly become critical in whether AI is creating value or destroying it: AI is either being used to cut costs, or it’s being used to compound expertise.

Which way you decide to go at that fork in the road may have an existential impact on your firm’s value to your clients, and subsequently your existence as a firm.

I wanted to learn more about the difference between these two paths and found that there is already a wealth of research supporting the notion that compounding expertise is the most likely path to success.

The Trap You’re Probably Already In

When AI gets framed as a cost-reduction tool, three things happen: middle managers get defensive, leadership ambition shrinks to incremental savings, and the firm slowly hollows out the human capabilities. The judgement, contextual reasoning and narrative development that clients actually pay you for all disappear the moment the experts walk out the front door.

Harvard Business School researchers call this “the efficiency trap.” Every leader making real progress in their study on effective AI usage had already moved past efficiency framing and on to compounding expertise. The ones still stuck on cost reduction were producing pilots, not transformation.

The trap is subtle. It feels responsible. It looks like good management.

It is quietly destroying the asset you’re trying to build.

Why AI Doesn’t Flatten Expertise

There’s a comforting story going around: AI will let anyone do anything. Junior staff will perform like seniors. Generalists will replace specialists. Headcount becomes optional.

The data says no.

A controlled experiment at UK fintech IG Group tested this directly. Researchers gave AI tools to three groups: domain experts, adjacent specialists, and outsiders from unrelated functions.

On easy tasks, AI nearly closed the gap. Outsiders performed at expert level.

On expertise-intensive tasks, the gap held. Outsiders with AI scored almost the same as outsiders without it. They couldn’t tell good output from mediocre output. They had no judgment to apply.

The researchers call this “the AI wall.” AI lifts you to the ceiling of your existing domain knowledge. It does not break through it.

The implication for your firm is direct.

If you reduce junior hiring because AI lets seniors do more, you stop developing the judgment that makes seniors effective. Your apprenticeship pipeline collapses. Five years out, you have no one ready to evaluate AI output at senior level.

You will have optimized yourself out of a valuable firm.

What Actually Drives Adoption

Here’s the third finding, and it surprises most founders.

Microsoft Research tested what predicts heavy AI use across hundreds of information workers. They measured leadership communication, formal training, infrastructure, and peer influence.

Peer influence won. Decisively.

Heavy AI use scaled with peer influence at roughly twice the rate of leadership communication. In fact, top-down messaging had no detectable direct effect once peer dynamics were accounted for.

Translation for founders: your job is not to mandate AI use. Your job is to make peer learning visible.

Lunch conversations beat training videos. Slack channels beat all-hands emails. A senior person showing a prompt that worked beats a buried instructional document.

This is not soft culture work. It is the actual mechanism by which AI value gets unlocked.

The Three Findings, Connected

Read alone, each finding is interesting.

Read together, they tell a single story:

1. The efficiency frame destroys long-term firm value.
2. AI cannot manufacture expertise it doesn’t already have access to.
3. Adoption scales through human relationships, not technology rollouts.

The common thread is human expertise. The firms that win are the ones treating AI as expertise infrastructure; protecting the people who hold judgment, building the peer cultures that transmit it, and resisting the temptation to gut the apprenticeship pipeline for short-term margin.

The firms that lose are the ones using AI to do the same work with fewer people.

Both firms will report productivity gains in the first year. Only one will be worth more in year three.

Three Things to Do This Quarter

Stop framing AI internally as cost reduction. Frame it as capability expansion. The words you use shape what your team protects and what they cut.

Audit your junior hiring assumptions. If you’ve quietly reduced hiring because AI lets seniors do more, you’re eating your own seed corn. The firms that win the AI decade will be the ones that keep investing in the people who develop judgment.

Make AI learning visible. Create one channel where your team shares what’s working: prompts, failures, weird wins. Model it yourself. Visible curiosity from the founder is worth more than a training budget.

The firms thinking critically about staying relevant in the next five years should pay particular attention. Clients do not pay premiums for cost-reduction stories. They pay premiums for defensible expertise, scalable systems, and proof that the firm can grow without the founder.

AI can help you build all three.

It can also help you destroy all three.

Choose deliberately.