The Three Clocks of AI

I’ve noticed people saying AI transformation is “slow.”

McKinsey’s 2025 State of AI report found that 88% of organizations use AI in at least one function — but only 39% report any enterprise-level EBIT impact. Nearly two-thirds haven’t scaled beyond pilots.

The usual diagnosis: bad data, weak models, poor integration.

But AI systems don’t run on one clock.

They run on three.

⸻

1. The Model Clock

Training, retraining, fine-tuning.

This is the clock everyone watches — and the one improving fastest.

Hours to days.

2. The Platform Clock

Built for safety, reliability, and governance.
Optimized for stability, not learning.

Days to weeks.

3. The Organizational Clock

Balancing AI’s potential with the risk of its opaque,
non-deterministic reasoning. This risk doesn’t scale linearly.

It compounds.

The organizational clock isn’t just slow.
It’s rationally cautious. And rightly so.

Quarters to years.

⸻

By the time an AI system encounters reality —
production, edge cases, actual user behavior —
many of the decisions constraining it were made quarters earlier.

And those decisions are slow to revise because risk is money.

The model adapts in hours.
The platform ships in weeks.
The organization decides in quarters.

Everyone calls this a gap. But it’s not a gap.
It’s three clocks running at the speed they should.

The frustration isn’t that AI is slow.

It’s that we’re measuring AI transformation by the fastest clock
and wondering why the slowest one won’t keep up.

Maybe AI isn’t behind schedule.

Maybe the schedule is wrong.