Why AI isn't living up to the promise
Every person on our team was getting results. The company wasn't.
Last month, I realized we’d hit a plateau with AI.
Everyone had developed their own workflows.
Their own prompts.
Their own tools.
We were getting real results. But those results weren’t spreading across the team. One person figured something out, but it stayed with them. The next person solved the same problem from scratch two weeks later.
We were using AI. We just weren’t using it as a team.
The labs almost guarantee this.
ChatGPT, Claude, Gemini are all designed as personal tools.
One conversation.
One session.
One user.
You can’t fault the individual for using the tool the way it was built. But the result, at the company level, is a collection of solo players.
No shared learning.
No compounding.
Every session starting from zero.
And, I think we’re pretty advanced in our “solo” use. We’re using Claude Code, Codex, and some pretty cool techniques. But there was no way to compound that work.
When we stopped and tried to build something cohesive… something that worked at the company level, not just the individual level… we hit the context wall immediately.
The data we needed to make AI work for us as a team was everywhere and nowhere. Spread across systems. Different formats. Hard to access. Harder for AI to actually use.
So we had to slow down to speed up.
Intelligence is no longer the scarce resource.
This is the insight that changed how I’m thinking about AI for professional services firms.
In the old world, intelligence was scarce. You hired experts who had it. You hired professionals, usually with some sort of certification in their domain.
Now AI provides intelligence cheaply. The bottleneck has shifted. The real constraint now is context. And context lives in your data and in your experience.
The companies that will win with AI are not necessarily using better tools or the newest model. They’re building a better data layer. They’re treating data the way they’ve always treated talent: as a strategic asset that needs to be organized, maintained, and made accessible.
We spent the last month doing exactly that.
My team mapped the data we actually need on a regular basis — marketing and sales performance, conversion metrics, project throughput, quality measures.
Then we built tools (with AI) to extract that data from our various systems every day. Consolidate it. Clean it. Make it usable.
Here’s just one thing we found when we did (and we’re just getting started).
We have significantly more leads this year. The sales team is fully booked. On the surface: a good problem.
Except. Our conversion rate is down.
When we let AI look for correlations across the consolidated data, it found something I wouldn’t have found on my own.
The average time between a lead’s first conversation with us and their second had shifted. Not collapsed. Shifted.
From 7 days to 11 days.
Four extra days. Leads sitting. Getting colder.
Conversion rate dropped 5%.
Without the data layer, we’d have done what most firms do: told the sales team to follow up faster. Close harder. Get better.
But the root cause wasn’t the sales team. It was a capacity problem… more leads than our scheduling could handle at the same pace.
The fix was structural, not behavioral.
We would have blamed the wrong thing.
We would have fixed the wrong thing.
Multiply that kind of insight across every part of your business.
Here’s what I think is coming:
In 18 months, the firms that have built a data layer will have a structural advantage over those that haven’t. Not because they’ll have access to better AI, everyone has access to the same models. But because their AI will know their business. Deeply. Across time.
The firm without a data layer gives their AI a blank notepad at the start of every session. Whatever context exists lives in one person’s head, and it resets when the session ends.
The firm with a data layer hands their AI a full history: every campaign result, every conversion rate, every bottleneck, every pattern across months and years. Their AI finds correlations that no individual would catch manually.
Same tools. Very different results.
That gap is already opening. And once it’s open, it’s hard to close.
Two firms.
One is a collection of solo players.
Good people.
Skilled.
Each getting results individually.
Nobody building on what the person next to them learned last week.
The other is a data company.
Context shared.
Learning compounding.
Insights appearing that would never surface from individual effort alone.
The solo players aren’t failing. They just hit a ceiling they can’t see.
Are you past it?
Steve “abundant intelligence” Gordon
P.S. The hardest part of building the data layer isn’t the tools. It’s stopping long enough to decide what data you actually need. Most firms skip that step — and then wonder why their AI keeps giving them generic answers. That decision is the whole step.
P.P.S. When you’re ready, our team is standing by to help you in four ways:
Become a Published Author — We’re taking 7 entrepreneurs who want a book published before Fall. We’ll extract your expertise into a book in 90 days and help you launch it to land clients within weeks. [Learn more →]
Join me on Substack: The ONLY place to find the full back catalog of Freedom Firm Insider issues and more exclusive insights to grow your firm. substack.stevegordon.io
Where’s Steve:
July 9-10 - Ed Slott’s 2-Day IRA Workshop - New York City
July 15-16 - Strategic Coach 10X Ambition - Chicago
July 30-Aug 1 - Garrett Planning Network Annual Retreat - Minneapolis
If you’re at any of those events, come see me.