What Is Our Work Worth in Times of AI?

Jun 3, 2026

In our last issue, we argued that HR, not IT, has to lead enterprise AI. Today, we follow the money one layer down, to what AI is doing to the price of the work itself.

💸 The Unit That Broke

Software is where everyone expected the disruption first, and it shows. Anthropic is past forty-five billion dollars a year in recurring revenue with fewer than five thousand people, three to five times more productive than the giants that defined the last two decades.

But the deeper shift is not how the work gets built, it is how it gets sold. For fifty years, software charged two ways: per user, you buy the seat, and per use, you pay every time you turn the key.

The new breed does neither.

AI-native firms give the software away and charge for the outcome: a law firm no longer buys contract-review software; it pays for the reviewed contract.

That same logic is now hitting professional services, which sold time for a century. When an agent drafts in minutes what once was billed for hours, the link between hours and value snaps.

AI-native firms automate the entry-level work and price the result, undercutting incumbents who must also retrain their people and rewrite how the whole firm is paid and organized, all while under fire. This is what an industrial revolution does: it redefines the unit of value itself.

The clearest way to see the shift is at the level of a single deliverable. Here is the comparison, built from our analysis of efficiencies measured or reported by 14,800 individuals in the US, Europe, and ANZ as of May 2026 and public labor data:

🔄 Offshore Outsourcing replaced by Insourcing plus Agentic Software

For twenty years, the smart move was to push work out: offshore it, hand it to a giant outsourcing center, bill by the hour or the head.

AI just flipped that logic. As agents absorb the repetitive, high-volume work, keeping it in-house becomes cheaper than outsourcing, and the pricing shifts from hours to results.

Two examples make it real.

• Outsourced call centers bill per person, per hour, win or lose; the new AI call centers charge per resolved conversation, so you pay only when the problem is actually fixed.
• Accounting and Tax: an agent now takes a return from raw documents to filing-ready and charges per return, not the dozen-plus hours a junior used to bill; a senior still signs off, but the work underneath runs at machine speed. One of the twenty-five thousand AI companies we track, Basis, reached a billion-dollar valuation doing exactly this across accounting, tax, and audit.

What stays human is the judgment. As Richard Susskind, co-author of The Future of the Professions, puts it, the client itself is now a competitor: they run the first-pass diagnosis with AI and only call you for the deep expert, to go one layer down. The headline view is free now. You get paid for the layer beneath it.

The market is showing some clear trends

In our datasets, AI-linked price concessions in professional services cluster around 10-20 percent, with a median near 14 percent. Public evidence supports this direction, including a reported 14 percent reduction in audit fees in a recent KPMG/Grant Thornton negotiation. Fortune 500 clients are increasingly using these efficiency assumptions as an anchor to demand shared productivity gains.

But the firms most exposed are not the ones whose clients push hardest. They are the ones who cannot even see, internally, where the saved hours went.

​
​

🪜 Machine efficiency is destroying the bench

The apprenticeship layer is under stress. Junior work like the first-draft memo, the research deck, and the document review was never really about the output. It was reps at the gym, the thousand repetitions that quietly build the judgment that makes a senior worth their rate. AI does the reps now. The work still ships, clients keep paying, and the erosion stays invisible right up until the day the senior bench is supposed to take over, and cannot.

​Professor Ethan Mollick at Wharton points to the term his colleagues there coined for the mechanism: cognitive surrender. You let the machine do the thinking, so the thinking never happens. The pyramid that organized this industry, a few partners over wide layers of juniors, is what built that judgment automatically. As it dissolves from the bottom up, firms are debating what replaces it: an “obelisk” with fewer layers, or an “hourglass” pinched in the middle as AI absorbs mid-level work. Either way, the rung where judgment used to form is the one disappearing.

🧭 What To Do

There is no clean answer, and the advice is not the same for everyone.

🔗 Resources

📣 Feel free to use this data in your communications, citing "AI4SP" and linking to AI4SP.org.

📬 If this email was forwarded to you and you'd like to receive our bi-weekly AI insights directly, click here to subscribe: https://ai4sp.org/60​