Why AI "Disruption" FUD Is Mostly Wrong—And Why That's Worse

Last month a client asked me to assess whether AI would "disrupt" their business. I gave them the answer they didn't want: the disruption crowd is mostly wrong.

Then I explained why that's worse.

Only 8.6% of enterprises have AI agents in production right now. That's not the sign of an industry on fire. But if you think that means you can relax, you've misread the pattern. I've watched this movie before—through the internet wave, mobile, blockchain, and now AI. The script always promises revolution. What actually arrives is something more dangerous: gradual absorption by the people already in power.

The castle doesn't fall. The castle just gets new plumbing.

The Disruption Story Never Plays Out the Way It's Sold

Remember when the internet was going to kill banks? When mobile was going to destroy retail? When blockchain was going to eliminate every middleman on earth?

The revolutionaries got the technology right and the sociology wrong. JPMorgan didn't disappear—it became the largest fintech company in the world. Walmart didn't get Amazoned out of existence—it built the second-largest e-commerce operation in America. The incumbents didn't die. They absorbed just enough of the new thing to keep the walls standing.

That's exactly what's happening with AI right now. Every major enterprise vendor is bolting AI into their existing platforms as fast as their engineering teams can ship. Microsoft isn't getting disrupted by OpenAI—it owns 49% of it and is embedding it into Office. Salesforce isn't being replaced by an AI startup—it's becoming Einstein GPT. Oracle, SAP, Workday, ServiceNow: same story, different press release.

The boring truth? Incumbents with distribution beat startups with better technology almost every time. Especially when the customer base is risk-averse professionals who need audit trails, compliance frameworks, and vendor insurance.

If you're a CPA or auditor waiting for AI to blow up your industry from the outside, you're watching the wrong movie.

What Actually Happens Is Quieter and Worse

Here's what I'm seeing in client engagements right now:

Margins compressing 2-3% a year. Junior roles not getting backfilled. Entire departments becoming "centers of excellence" staffed by three people and a fleet of copilots. Clients expecting the same deliverable in half the time because "you have AI now, right?"

Nobody gets fired the day the railroad arrives. The town just slowly empties out.

I was talking to a tax partner last week who casually mentioned they didn't hire their usual four first-years this season. Not because of layoffs. Not because of a hiring freeze. Just because the workload math changed. They couldn't articulate exactly why they needed fewer people—just that the leverage ratio felt different now. Three years from now, those four jobs won't be "lost to AI." They'll just never have existed.

That's not disruption. That's erosion. And erosion doesn't make the news.

The Pattern You Need to Recognize

If you've been in professional services for more than a decade, you've seen this before. Not with AI specifically, but with every technology wave that promised transformation and delivered optimization instead.

Excel didn't eliminate accountants. It eliminated roomfuls of bookkeepers and changed what "accountant" meant. Email didn't kill law firms. It killed every legal secretary who only typed correspondence. Cloud didn't destroy IT departments—it cut headcount by 40% and renamed the survivors "cloud architects."

The professionals who got hurt weren't the ones who ignored the technology. They were the ones who assumed their job title was a defense. The ones who survived were the ones who noticed which 10% of their work was becoming 5% easier every quarter and made a deliberate move up the value chain.

I'm not talking about "upskilling" or "embracing lifelong learning" or any of that motivational-poster garbage. I'm talking about cold pattern recognition: What part of my work is becoming automatable? What client problem does that let me solve that I couldn't before? And what do I need to learn in the next six months to be the person who owns that new problem?

The Uncomfortable Question You Need to Ask

Here's the part where I don't give you the answer.

What's the one thing in your practice getting 5% easier to automate every quarter?

Not "Could AI theoretically do this someday?" Not "What's the hot new tool everyone's talking about?" But what specific task in your actual work is becoming incrementally, measurably easier to delegate to software every 90 days?

If you can't name it, you're already behind. If you can name it but you're still doing it the same way you did two years ago, you're watching the drought and calling it a dry spell.

The professionals who thrive in the next decade won't be the ones who panicked about disruption. They'll be the ones who noticed the slow shift and moved before the job description changed around them.

What to Do Monday Morning

Stop waiting for the meteor. Start tracking the drought.

Here's your assignment: Open your time tracking from last quarter. Find the three tasks you bill for that took 10% less time than they did a year ago. Not because you got faster—because the tooling got better, the template got smarter, or the client started accepting a different format.

Those are your leading indicators. The question isn't whether those tasks disappear. The question is what you're going to own instead.

If you need help thinking through what that looks like for your practice, I'm happy to compare notes. I've spent the last year helping clients figure out which side of this shift they want to be on. The answer is never "ignore AI." But it's also never "panic and retrain as a prompt engineer."

It's usually something much more boring: "Do more of this one thing you're already good at, and stop pretending the thing that's becoming a commodity is still your value proposition."

What's your answer?