Every time a kind of work gets cheaper to produce, companies produce more of it. AI is doing that to reports, emails, decks and analysis right now. Making them costs almost nothing, so the volume climbs, and the limit moves to the people who have to read the output, judge it and act on it. A leader who misses that shift ends up with managers buried in output they never asked for and still have to read.
Cheaper AI output raises the volume of work
Coal showed it first. Steam engines had become far more efficient, and Britain was burning more coal, not less, because cheaper power found more uses. William Stanley Jevons pointed this out in 1865, the pattern now carries his name, and it shows up wherever a cost drops.
Email is the cleanest example I know. In retail marketing, I have yet to see anyone cut send volume and keep revenue without changing the price. People respond to more, so the math always favors sending more, and the only lever that holds revenue at a lower volume is price. That is why everyone's inbox keeps filling up, whatever anyone says about it.
Inside a company the same math runs on your own people. A competitor summary that once took a week now takes a minute, so every department starts commissioning its own, and with first drafts free, decks start showing up for meetings that never had one. Each one may be fine on its own, and all of them land on managers whose hours have not changed.
Your team absorbs everything AI produces
Most of what a business produces internally is consumed by the business itself. Think of a sales manager's Monday: the pipeline report, the forecast, two dashboards, a research summary and a draft proposal to review. Each is an input to a decision that manager has to make that week, and when the inputs are good, the decisions get better and so does the next round of work. That loop is how good companies improve.
AI speeds up the producing side of that loop and does nothing for the absorbing side. Researchers at Stanford and BetterUp gave the result a name in Harvard Business Review in September 2025: workslop, AI output that looks finished but does not move the work forward, so the person receiving it has to fix it or redo it. That is the absorbing side of the loop breaking. The cost does not disappear. It moves from the person who made the draft to the person who has to use it.
Budget what your team can absorb, not what AI can produce
The question for a leader is how much the team can take in and act on. A few rules I would put in place:
- Give every recurring output a named reader and a decision it feeds. If nobody can say what they decide from a report, stop making it.
- Retire one recurring report, meeting deck or update for every new one AI makes possible.
- Hold the person who sends AI output responsible for checking it first, so the cost of a bad draft stays with whoever made it.
- Count decisions made and acted on, not documents produced.
Where more volume is the right answer
Customer-facing volume is a different case. If more sends, more posts or more outreach bring in more revenue, as email has shown for years, then cheap production is exactly what you want, and price is your control. The rules above are for the work your own people have to read.
Count what lands on one manager's desk next week
Pick one manager and count every report, dashboard, deck and AI draft that reaches them next week. Then ask which of them changed a decision. The gap between those two numbers is work nobody needed, and it is the first place to cut.
For the finance version of this problem, see why finance teams end up checking AI more than using it, and for what to measure instead of output, scoring AI on growth. You can read about the AI and growth work I do with leadership teams. If your managers are reading more and deciding less, reach out.
John Wyatt · John Wyatt's page · All posts · RSS