My AI Predictions That Were Wrong (And What I Learned)

It's December, which means it's time for year-end reflection. I made some predictions about AI and professional services back in 2023. Most of them were partially right and partially wrong. Let me be honest about where I missed and what that teaches us.

Prediction 1: "AI Will Compress Project Timelines by 40–50%"

What Actually Happened: Time compression happened, but it was closer to 15–25% and only in specific parts of projects.

I was wrong because I underestimated how much of professional work is about judgment calls, client interaction, and opinion synthesis—not raw information processing. A strategy project doesn't take six months because the analysis is hard. It takes six months because understanding what matters to the client and selling them on the answer takes time.

AI made the analysis faster. But it didn't accelerate the client management and consensus-building parts. Those are still as slow as they've always been.

What I Learned: Don't confuse "AI does this task faster" with "AI makes the overall project faster." The constraint is often not where you think it is.

Prediction 2: "Firms Will Cut Prices Because AI Will Reduce Their Costs"

What Actually Happened: Almost nobody cut prices. Margins improved for early adopters, or they used the gains to do higher-quality work.

I was wrong because I underestimated how inelastic professional services pricing is. If a firm tells clients they're cutting rates because of AI, the implication is "our service was overpriced before, and now we can deliver the same thing cheaper." That's not a message any firm wants to send. Instead, they used AI to either increase profit or improve quality. Both are better business moves.

What I Learned: When you get more efficient, the rational economic move isn't always to pass the savings to customers. You have choices, and most successful firms chose differently than economics textbooks would predict.

Prediction 3: "Adoption Will Be Fast Because the ROI Is Obvious"

What Actually Happened: Adoption is still slower than I expected. Even with clear ROI, change management is hard.

I was wrong because I underestimated organizational friction. Even if a firm demonstrates "AI saves us 5 hours per week," getting all 200 people to actually use the tool and build it into their workflow takes months. Some people resist. Some people try it and decide they prefer their old way. Some people never see the ROI apply to their specific work. It's much messier than "clear ROI → rapid adoption."

What I Learned: There's a difference between technical readiness and organizational readiness. Both matter, and most organizations underestimate the second one.

Prediction 4: "The Firms That Move First Will Have Permanent Advantage"

What Actually Happened: Early movers did gain some advantage, but it's smaller than I expected. The real advantage goes to firms that moved at the right pace—not fastest, not slowest, but thoughtful.

I was wrong because I assumed that being early mattered more than being right. In practice, a firm that implemented AI carelessly in 2023 has spent two years dealing with governance problems, user adoption issues, and technical debt. A firm that waited until 2024 and implemented thoughtfully might already be catching up.

What I Learned: Speed can be an asset or a liability depending on execution quality. Sometimes moving second, having learned from first movers' mistakes, is a better position.

Prediction 5: "Regulatory Clarity Will Come Quickly"

What Actually Happened: Regulatory guidance is coming slower than I expected. We're still in a gray zone on a lot of questions.

I was wrong because regulators move slowly by nature, and AI is complicated. I expected professional services regulators to move faster because the stakes are high. But they're also cautious, and they want to get it right. The guidance that has come out is mostly conservative and prescriptive, not clarifying. Firms are still largely navigating on their own.

What I Learned: Don't expect regulators to solve ambiguity quickly. Be prepared to make governance decisions in the absence of clear guidance, and document your thinking.

Prediction 6: "Data Quality Will Become the Key Bottleneck"

What Actually Happened: This one I got mostly right. Data quality matters enormously. But the flip side I didn't fully predict: so does organizational buy-in on data governance. A firm with bad data can still use AI effectively if they're very careful about validation. A firm with good data but no governance discipline can still mess it up.

What I Learned: Technical readiness and organizational readiness are coupled. You can't solve one without the other.

Prediction 7: "AI Will Enable Smaller Firms to Compete With Larger Ones"

What Actually Happened: This is still an open question. Small firms have easier change management. Large firms have more resources. The outcome is probably somewhere in the middle—small firms move faster on AI, large firms catch up with money and scale. The competitive gap might get narrower, but it's not closing as fast as I thought.

What I Learned: Scale advantages are durable. AI changes the curve but doesn't eliminate it.

What These Wrong Predictions Tell Us

If I had to summarize the pattern, it's this: I underestimated how much professional services work is social, organizational, and political, and overestimated how much is purely technical and process-driven.

I also overestimated the speed of change. Technology adoption is slower than technologists think it should be. But that slowness often leads to more sustainable outcomes.

What I'm Still Confident About

Despite all the wrong predictions, I still believe:

• AI will be central to professional services in five years. Firms without serious AI capability will be at significant disadvantage.
• The winners will be the ones who solve the governance and organizational pieces, not the ones who have the fanciest models.
• Firms that start now (even if others started earlier) can still catch up because the game is still early.
• This is going to matter a lot. Not in 2023 way where everything changes overnight. But in the sense that in 2030, we'll look back at 2025 and see it as the critical inflection point where firms made choices that determined their competitive position for a decade.

The Meta-Learning

Being wrong about some predictions while being roughly right about others teaches you something: long-term direction often matters more than short-term timing. I got the direction right (AI will be important). I got the timing messier (faster in some ways, slower in others). That's actually useful information. It suggests the thing to do is make moves that work regardless of timing uncertainty.

If you're betting on AI mattering and want to build resilience to being wrong about timing, you invest in foundational things (data quality, governance, talent) rather than on specific timelines.