AI Doesn't Have to Beat Your Best. It Just Has to Shrink the Pyramid.

Everyone in finance is asking whether AI can beat the best human in the room. Wrong question. Wrong benchmark.

The Wall Street Journal just covered a new study casting doubt on AI's ability to time the stock market. Short version: it's worse at it than the marketing suggests. Cue the relief in a hundred investment committees. CFOs forwarding the article with "See? Told you." Partners breathing easier about that AI strategy they've been slow-rolling.

Here's what that relief misses: Beating the market was never where the disruption starts.

The Question Finance Leaders Are Getting Wrong

I spend my days with finance leaders, and they all ask the same prestige question: "Can AI outperform our best analyst?"

It's comfortable, because the answer is still no. The quants are safe. The rainmakers are safe. The partners who closed the impossible deal are safe.

The question that actually reprices your org chart is cheaper and uglier: How much of the analytical work gets compressed before AI beats anyone's best?

AI doesn't have to make your smartest person obsolete. It has to be good enough that you need fewer of everyone else. Not zero. Fewer. That's the disruption, and it's already here.

We've Watched This Movie Before

Excel already taught this exact lesson — we've just forgotten the plot.

When spreadsheets arrived in the 1980s, they didn't replace the CFO. They didn't eliminate financial analysis. They hollowed out the rows of analysts underneath and quietly changed who held leverage on the finance team.

Before Excel, building a financial model meant rooms full of people with calculators and ledgers. Junior analysts spent weeks manually updating projections, re-keying numbers, checking arithmetic. The CFO needed that army because the mechanical work was enormous.

Excel compressed all that effort into hours. Suddenly one senior analyst could do what used to require six people. The top got more powerful — armed with better tools, faster iteration, cleaner models. The layer doing the manual build-out got smaller. Much smaller.

Nobody fired the CFO. They just stopped hiring the next five junior analysts.

The Disruption You're Not Benchmarking Against

This is where finance leaders are aiming their telescope at the wrong part of the sky.

You're testing whether AI can pick better stocks than your star portfolio manager. Whether it can spot fraud faster than your most experienced auditor. Whether it can structure a deal better than the partner who's done 200 of them.

Stop benchmarking AI against your superstar. Point it at first-pass analysis, draft memos, variance explanations, reviewable output. The median Tuesday, not the career-best trade.

That's audit work. That's tax research. That's FP&A grunt work. That's compliance documentation. That's the associate staying late to build the third version of the sensitivity analysis because the partner wants to see different assumptions.

Right now, today, AI can:

• Draft the initial variance explanation when actuals miss budget

• Pull comparables and generate the first-pass valuation range

• Summarize 200 pages of financial statements into the 12 things that actually matter

• Generate the regulatory disclosure language for the fourth time this quarter

• Build the Excel model from the memo (yes, really)

Is any of that perfect? No. Does it beat your best person? Also no.

Does it mean you need fewer people between the partner and the work product? Yes. And that's the entire game.

The New Org Chart Nobody's Drawing Yet

I was working with a mid-sized accounting firm last month. Managing partner asked me: "Should we be training our interns on AI tools?"

I gave him the uncomfortable answer: "You should be asking whether you'll need the same number of interns."

Silence. Then: "We've always had first-years do the cash-to-accrual conversions, chase down the missing invoices, build the workpapers..."

Right. You've always needed bodies for that work because it took bodies. The math is changing.

Not because AI is brilliant. Because AI is good enough that one senior person can review and correct its output faster than they can build it from scratch. And when that flip happens, you don't need three juniors feeding work up to one senior. You need one senior and the machine.

The pyramid gets shorter. Flatter. The people who stay aren't the fastest forecasters or the most efficient workpaper builders. They're the ones who know which draft is actually right — the judgment the machine still can't fake.

Can the AI tell you that the variance explanation makes no operational sense even though the math checks out? That the valuation looks textbook-perfect but is missing the one intangible that actually drives this industry? That the client is technically compliant but walking into a disaster the rules don't see coming?

Not yet. Maybe not ever. But if that's not your skillset, I'd be worried.

What To Do Monday Morning (The Uncomfortable Version)

Here's where I'm supposed to give you the comfortable action items. "Experiment with AI tools!" "Upskill your team!" "Embrace the future!"

I'm going to ask harder questions instead, because you need to sit with the tension, not paper over it:

Where in your organization is AI already good enough that you need fewer people? Not "where could AI theoretically help someday" — where is it good enough today that the headcount math changes?

What does "senior" mean when the leverage ratio flips? If one experienced person can direct AI through work that used to require a team, what are you hiring for? What are you training people to do?

Who on your team has the judgment that matters when everything else gets commoditized? The technical skills get easier to replicate. The judgment — knowing when the model is lying, when the client needs a different answer than they asked for, when the rule misses the risk — that's the new moat.

And the most uncomfortable one: Are you building AI into the right layers of your org chart, or the comfortable ones?

Most firms I see are giving AI tools to senior people to make them more efficient. That's fine. That's safe. That's not where the disruption happens.

The disruption happens when you realize the entire layer of people feeding work to those senior people gets compressed. And if you're not actively managing that transition, it's going to manage you.

The Real Benchmark

The study the WSJ covered is real. AI probably can't beat your best stock picker. Your best forensic accountant. Your best M&A advisor.

But I've watched technology disruption cycles rewrite entire industries four times now. The disruption never starts by beating the best. It starts by making the best person need less help.

Excel didn't replace financial analysis. It just changed how many analysts you needed per dollar of work. AI won't replace professional judgment. It'll just change how many professionals you need per judgment call.

That's not hype. That's not fear-mongering. It's pattern recognition from someone who's watched the movie before.

Your smartest person is safe. The pyramid underneath them? That's the conversation you should be having this week.

This week's action: Pull your org chart. Find the work that's reviewable, repeatable, knowledge-based but not judgment-based. That's where the math changes first. If you're not pressure-testing those roles against what AI can do today, you're benchmarking against the wrong standard.

What are you discovering when you point AI at the median work instead of the exceptional work? I'd genuinely like to know — the pattern is still forming, and the people doing the work see it clearer than the people writing about it.