The 5,000-hour problem.

Somewhere right now, a partner is presenting a staffing grid. Four workstreams, twelve consultants, eighteen months. Five thousand hours to fix a revenue engine. The room nods, because that's what these projects have always cost.

The grid made sense when humans did all of the work. Someone had to read every field in the CRM, trace every broken handoff, reconcile the forecast against the ledger, document what they found, and build the fix. That's months of senior time wrapped in years of junior time, and the billable hour was how you metered it.

Then the grunt work stopped requiring humans.

Agentic AI — not a chatbot, but agents that read a system's metadata end to end, map every field and dependency, surface every orphaned automation and silent failure — does the discovery layer of a transformation in days. Not approximately. We watch it happen: the diagnostic that justified the first thousand hours of a traditional engagement now takes a well-built agent pipeline and one senior operator about a week. Across the engagement, the compression runs somewhere near eight to one.

A pyramid of analysts has to eat. A firm that leveraged junior headcount for fifty years cannot sell you an eighth of the hours, because the hours are the product. So the staffing grids stay big, the agents get used quietly to pad margins instead of cut timelines, and the client pays 2019 prices for 2026 work.

Wooden wheels moved carts for three thousand years. The alloy wheel with a rubber tire is the same shape. Nothing else about it is the same — not the speed, not the load, not the roads it opened. The firms selling you 5,000 hours are selling you a beautifully crafted wooden wheel, with a straight face, in 2026.

Here's what the alloy version looks like in practice. Findings in week one, not a kickoff deck. Fixed scope priced on the outcome, because when the diagnostic takes days there's nothing to meter. Senior operators only — the people who'd have reviewed the juniors' work now just do the work, with agents underneath them. And you own everything that gets built, because there is no incentive to make the engine need us forever.

It changes staffing too, and almost nobody is talking about that part. A Salesforce admin trained on agentic multipliers is not ten percent faster than one who isn't — they produce like a pod. When we place platform people, that training is the spec, not a bonus. One seat, screened hands-on by consultants who do the work, producing what used to take a team.

The hour had a good run. It was the right unit when effort and progress were the same thing. They aren't anymore.