AI is the Compression Algorithm for Experience

In 2016, Andy Jassy — then CEO of AWS, now CEO of Amazon — made a statement that became a kind of consulting industry koan: "There is no compression algorithm for experience."

He was right. Until he wasn't.

What Jassy Was Actually Saying

The original insight was sound. Experience isn't a file you can zip and transfer. It's the scar tissue from a thousand bad calls. It's knowing which alarm to ignore at 2am and which one to wake the whole team up for. You couldn't shortcut it. You had to earn it — year by year, client by client, production incident by production incident.

That framing shaped how professional services got priced. Senior people were expensive because they held compressed knowledge that junior people hadn't earned yet. The business model made sense: sell access to experience, charge accordingly, staff juniors where you can get away with it.

The Model Didn't Update. We Wrote a New One.

AI didn't iterate on that model. It replaced it.

A senior engineer with 20 years of production experience can now encode that judgment — into prompts, into agents, into the architecture of a system. The junior who once needed three years of ramp-up time to become useful now operates with guardrails built from decisions made long before they joined. Not because they earned it. Because the context was given to them.

That's the compression. Not of the experience itself — but of the cost of accessing it.

"Fast at Being Wrong" Is the Real Risk

Here's where the nuance lives, and where most of the hype breaks down.

AI doesn't replace experience, it amplifies it. A senior engineer with an agent is 10x more effective. A junior with an agent is often just fast at being wrong — and confident about it. The difference isn't the tool. It's the judgment behind the tool.

This is why the staffing model matters. When we say AI-first at Millwater Consulting, we mean senior engineers using agents they built, tested, and trust. Every tool we recommend, we run ourselves. Every agent we ship to a client, we ran internally first. The AI amplifies the experience. The experience still has to be there.

No junior bench. No learning on your time. No surprise ramp-up tax three weeks into an engagement.

At Millwater Consulting, we’ve built AI that codifies over 100 years of shared experience. And surfaces it contextually, when it’s needed. Right time, right place.

The Value Has Shifted

The old model priced labor. The new model prices judgment.

What you're paying for isn't the hours someone sits in your codebase. It's the quality of the decisions being made about your codebase — architecture, security posture, what to build and what to kill. That judgment doesn't get cheaper because we use agents. It gets faster, more leveraged, and more consistent.

The ramp-up tax disappears. The "we'll escalate when needed" staffing game disappears. You get the senior perspective from the first commit, not after a six-week discovery phase where someone is still reading your README.

What This Means Practically

The right question isn't "how many people?" It's "how much judgment, and how fast?" Headcount was a proxy for experience. It's an unreliable one now. Context is the new moat. AI trained on your codebase, your incident history, your architecture decisions - outperforms generic AI every time. Firms that compound that context across engagements will widen the gap on firms that start fresh each time. Output is the only metric. Not hours logged. Not standups attended. Does the system work? Is the judgment sound? If yes, the work is done. If not, no amount of process saves it.

Where We Land

Andy Jassy was right in 2016. There was no compression algorithm for experience — not a good one anyway. AI isn't a perfect compression either. But it's close enough to matter, and it gets better every quarter.

The consulting firms still clinging to headcount as a proxy for value are selling a model that no longer holds. The ones that understand what actually changed — that the labor of synthesis is cheap now, and judgment is the only scarce resource — will build something worth having.

We built Millwater Consulting on that premise. Senior engineers from day one, running the same agents we'd ship to you, trained on experience that took decades to earn. The ground truth is that AI compressed the delivery. The judgment behind it didn't come cheap.

That's the model. We don't plan to update it.

— Rishi Prasad, Lead Full Stack Developer, Millwater Consulting https://millwater.consulting