AI Doesn't Have a Model Problem. It Has a Leverage Problem.
Every AI team I talk to is chasing the same improvements.
A bigger model.
A better prompt.
A larger context window.
The newest release from OpenAI, Anthropic, or Google.
Sometimes those upgrades help.
Usually… less than people hope.
Because for an entire class of AI problems, the model isn’t where the biggest leverage lives.
In an earlier post, I argued that AI work naturally separates into two jobs.
Understanding what someone wants.
Deciding what to do about it.
AI is exceptionally good at the first.
Business rules should almost always own the second.
That distinction has held up remarkably well across every implementation I’ve looked at since.
But I’ve realized there’s another pattern hiding underneath it.
Not another kind of AI.
Another place where leverage lives.
Take Generative Engine Optimization.
You publish an article.
A week later someone asks ChatGPT:
“Who are the leaders in AI Operations?”
Does your company appear?
There isn’t a rule that guarantees it.
There isn’t even a meaningful probability you can calculate.
The models evolve.
Retrieval changes.
Ranking changes.
Deterministic. Probabilistic. Possibilistic. Same operating system, three altitudes.
Even the companies building these systems can’t fully explain why one source appears and another doesn’t.
Yet everyone knows there are things that improve your chances.
Better information architecture.
Consistent entities.
Structured metadata.
Canonical URLs.
Fresh, factual content.
None of those guarantee success.
But ignoring them almost guarantees failure.
That’s a very different kind of system.
You don’t control the outcome.
You influence it.
I stumbled into something mathematicians described decades ago.
They call it Possibility Theory.
Probability asks,
“What are the odds?”
Possibility asks,
“What conditions make this outcome more achievable?”
You don’t force the result.
You increase the possibility that it happens.
And suddenly I started seeing this pattern everywhere.
Search.
Recommendations.
Personalization.
Agent routing.
Content discovery.
API selection.
They behave less like software executing instructions…
…and more like ecosystems responding to conditions.
Here’s where I think many teams go wrong.
When these systems disappoint, they optimize the visible layer.
They switch models.
Rewrite prompts.
Fine-tune.
Change vendors.
They keep working on the AI.
When the biggest leverage usually lives somewhere else.
The deterministic substrate.
The boring stuff.
Taxonomies.
Metadata.
Version control.
Knowledge organization.
Naming conventions.
Templates.
Audit trails.
Almost nobody demos these at an AI conference.
Ironically…
They’re often the reason the AI succeeds.
You increase possibility by maximizing certainty upstream.
That’s the operator principle I’ve kept coming back to this year.
The cleaner your foundation becomes…
You increase possibility by maximizing certainty upstream.
the more opportunities the AI has to produce value.
Not because the model suddenly became smarter.
Because the environment became easier to reason about.
This doesn’t replace the AI Operating System framework.
It strengthens it.
Understanding remains probabilistic.
Decision Policy remains deterministic.
Human Review still defines the trust boundary.
But many of the business outcomes we’re chasing…
discoverability…
recommendations…
adoption…
relevance…
…aren’t deterministic or probabilistic at all.
They’re possibilistic.
And that changes the question.
Instead of asking,
“Which model should we use?”
The better operators ask,
“What foundation can we improve that makes success more likely?”
That’s where the leverage lives.
Most organizations think they’re investing in AI.
The best operators are quietly investing in the systems underneath it.
Models create possibilities.
Systems determine which ones become reality.
Drafted with AI. Refined with care.