top of page

My second AI wave. The first one froze.

This is my second AI wave.

I remember how the first one ended, because I watched it from a particular seat. In the nineties, I deployed and integrated expert systems and decision support: the first commercial AI. I wasn't writing the inference engines. I sat between the machine and the business: gathering requirements, explaining the task to the experts, testing what came back. The seat, as it turned out, with the best view of the cause of death.

The idea was elegant: sit with the experts, extract what they know, encode it as rules, and the machine reasons like your best specialist. In the demos, it always worked.


Then it died, and the industry politely called the funeral "AI winter".

The killer had two hands.


The first was the rule bottleneck. Every rule had to be pulled out of an expert's head, one interview at a time, and experts couldn't articulate half of what they knew. Every extracted rule was brittle; every edge case bred three more.


The second was worse: the answers never added up to a world. Each rule worked only in its own context, and even a narrow domain needed a practically endless supply of questions. Once, tracing one expert's answer to its root, we arrived at The Adventures of Tom Sawyer, read in childhood. That was our honest joint conclusion. Knowledge wasn't a list of answers. It was a web, and the web had no bottom.


I spent the decades between the waves as an enterprise architect in banks.

When LLMs arrived, I did not recognise the moment. The first models answered hard questions badly, and I'd seen fluent demos die before. My question wasn't "how good is it?" It was: can this thing hold a world? Rules never could, at any price.


That's the real difference this time. The machine learns patterns from everything we've written. It arrives with a world, imperfect but present, instead of demanding one rule at a time.


But nobody from the first wave forgets: the knowledge still has to come from your experts. The bottleneck didn't die. It moved. Where it moved to is one of the central arguments of Architecture of Intellect. The book is in my profile.

Anyone else here shipped an expert system? Tell me what your winter looked like.

Related Posts

See All
Waves, not tides

The question your enterprise was built to answer is dead. The winning strategy is no longer "build and scale". It's "sense and respond".

 
 
 
The company that pays people to spend on AI

Minimum AI spend quotas to force usage, seven-figure token bills, no return. Why knowledge workers adopt tools that pay back, not tools ordered, and who holds the mandate to rebuild the work.

 
 
 
Why I wrote this book

This week, a board will approve an eight-figure AI budget. The evidence: a vendor deck and a thirty-minute demo. I've watched companies buy AI twice. In the nineties it was called expert systems.

 
 
 

Comments

Rated 0 out of 5 stars.
No ratings yet

Add a rating

Get the next argument first. Articles, campaign posts and book news. No more than one email a week.

bottom of page