Two days ago we described two hires given the same 100-piece job. One worked through it in ten days with a chat window. The other spent two days building a pipeline and finished in five. That post asked who gets paid. This one asks the next question, the one the agency's owner asks the moment they see the second timesheet: can the builder teach the pipeline to everyone else?
Of course they can. The question is what happens to the builder when they do. This is still a thought experiment; no agency, client or person in it is real. The surveys, the economics and the one company policy cited below are real, and each is labelled with who ran it.
- 01Sharing the method is good for the firm and, unrewarded, a net loss for the builder.The builder gives up a private advantage, and the shared method becomes the expected pace for everyone, including them. Unless something is paid for the transfer, the rational move is to keep it quiet.
- 02From a third to over half of AI users already hide their use, and some say why.Four surveys since 2024 put the share at 32% to 57%. Ivanti's respondents named a secret advantage; Slack's named looking like a cheat or less competent. Only 13% in Microsoft's 2026 index say they are rewarded for redesigning work even when results fall short.
- 03This is old economics with a new tool.Knowledge hiding has a research literature of its own, and the ratchet effect explains why a worker who reveals how fast a job can be done expects the standard to move. Taylor wrote it down in 1911.
- 04Firms can pay for distillation, and one has published how.Omnisend's 2% to 4% raise scores how widely a workflow is adopted by others. Authorship credit, time-boxed exclusivity and a share of saved hours are the other three levers, each with a side-effect.
01 — The scenarioThe dilemma in one paragraph
The agency wants the pipeline. If every content person runs it, the firm produces twice the output for the same payroll, and the question in yesterday's post about clients asking for a share of AI savings becomes easier to answer. So the owner asks the builder to run a session, write it up, and help the others set it up. The builder says yes, because it is a reasonable ask from an employer.
Three things then happen to the builder. The edge that made them the fastest person in the building is gone, because everyone has it. The baseline expectation for everyone, including the builder, is now five days rather than ten, so the slack they used to have is gone too. And the next method they build will be asked for sooner, because the firm now knows they build methods. The firm gained a process. The builder gained a reputation and lost a position. If the pay is the same as before, the transfer was free to the firm and expensive to the person who made it.
That is the whole dilemma. Everything below is about whether anyone pays for the transfer, and what happens when nobody does.
02 — The evidencePeople already hide AI use, and the surveys say why
We are not guessing at the builder's response. Four surveys since 2024 have asked workers whether they conceal their AI use, and a fifth asked whether reinventing work is rewarded. The answers are consistent across vendors, countries and years. All five are self-report surveys, four of them commissioned or run by companies that sell workplace software, so read the figures as direction rather than precision. Our comparison of 20 workplace AI studies explains how we tier this kind of evidence.
| Survey | Share | What the respondents said |
|---|---|---|
| Microsoft and LinkedIn Work Trend Index, May 2024 (31,000 people, 31 countries; vendor-commissioned) | 52% | of people who use AI at work are reluctant to admit using it for their most important tasks; 53% worry it makes them look replaceable |
| Slack Workforce Lab, Fall 2024 Workforce Index (17,372 desk workers, 15 countries; vendor-commissioned) | 48% | would be uncomfortable admitting to their manager that they used AI for common tasks; the reasons given are that it feels like cheating (47%) and fear of looking less competent (46%) or lazy (46%) |
| University of Melbourne with KPMG, Trust, attitudes and use of AI, April 2025 (48,000+ people, 47 countries; academic-led, KPMG-funded) | 57% | of employees say they hide their use of AI and present AI-generated work as their own |
| Ivanti 2025 Technology at Work (6,000+ office workers; vendor-run) | 32% | of generative-AI users keep their use secret from their employer; 36% of those say they like the secret advantage and 30% fear their job may be cut |
| Microsoft Work Trend Index, May 2026 (20,000 AI-using knowledge workers, 10 markets; vendor-commissioned) | 13% | of AI users say they are rewarded for reinventing work with AI even if results aren't met; 45% say it feels safer to focus on current goals than to redesign work with AI |
Two rows carry the argument. Ivanti's respondents were asked why they keep it secret, and the most common answer was that they like the advantage. That is the builder in our scenario, before anyone asks them to share. And Microsoft's 2026 index, the Work Trend Index, finds that the small group it calls frontier professionals share tips and agents with colleagues far more than other users do (61% against 36%) and are twice as likely to say they are rewarded for reinventing work regardless of outcome (26% against 11%). Microsoft presents that as a trait of the group. It reads just as well the other way: the people who are paid to share, share.
03 — The economicsWhy hiding the method is rational, and has been for a century
Organisational researchers have a name for what the builder is tempted to do. Connelly, Zweig, Webster and Trougakos defined knowledge hiding in a 2012 paper in the Journal of Organizational Behavior as an intentional attempt to withhold knowledge that someone else has asked for, and sorted it into three forms: evasive hiding, where the person gives incorrect information or a misleading promise of a full answer later; playing dumb; and rationalised hiding, where the person gives a reason the knowledge cannot be shared. "I would have to clean up the prompts first" is evasive hiding. "It only works on my machine" is rationalised hiding. Neither is dishonesty exactly. Both are what a person does when sharing costs them something.
Labour economists have a second name for the cost. The ratchet effect describes what happens when a worker who shows how fast a job can be done sees the standard raised to match. Gibbons showed in 1987 that when neither the firm nor the worker can commit to future behaviour, no pay scheme gets the worker to stop restricting output. The previous post quoted Frederick Taylor's 1911 observation that workers deliberately kept employers ignorant of how fast work could be done, and his explanation was the same: they had seen the piece rate cut after working harder. The chat window and the pipeline are new. The incentive to keep the pipeline quiet is not.
The practical reading for a manager is that the builder's reluctance is not a character flaw to be managed. It is a price signal. The firm is asking for something valuable and offering nothing, and the quiet no is the market's answer.
04 — The leversFour ways a firm can pay for the method being shared
Only one of these four has a published company policy behind it. The other three are the shapes we see proposed, each with the incentive it creates written beside it, because every one of them teaches the next builder something.
A raise scored on adoption by others
Omnisend, a 250-person company, gives a 2% to 4% raise to standout AI users, judged on time and cost saved, on outcome impact, and on how widely the workflow they built is adopted by colleagues. Reviewed quarterly, with the head of AI operations expecting no more than 60% to qualify in the first round. The incentive: build things other people can run. The side-effect: a person can be paid for a method that spreads and never quite works.
Authorship credit
The pipeline is documented under the builder's name, cited when it is used, and listed in their review. Costs nothing and is the minimum. The incentive: reputation. The side-effect: reputation does not pay rent, and a builder who is asked for the third method with only credit to show for the first two notices.
Time-boxed exclusivity
The builder runs the method alone for a stated period and is measured on output at the old baseline during it. After that it is shared. The incentive: build early, because the head start is the reward. The side-effect: a quarter of the firm running slower than it could, and a deadline that invites evasive hiding at the end.
A share of the saved hours
The firm measures hours the method saves across the team over a window and pays the builder a stated fraction, as gainsharing plans have long done for factory teams. The incentive: build methods that save the most hours for the most people. The side-effect: the measurement is the hard part, and a fraction of nothing is what the builder gets if the saved hours are never counted.
The salary bump gives people that extra bit of motivation.Bernard Meyer, head of AI operations at Omnisend, quoted by Fortune, March 30, 2026
Omnisend's design deserves a closer look because it answers the dilemma directly. The third criterion, as Fortune reported it, is widespread adoption of the workflow the person developed. A builder at that company who keeps the method private forfeits part of the raise. A builder who spreads it is paid for the spread. The example Fortune gives is a sales team whose 24-hour follow-up rate went from about 20% to close to 100%, which is the kind of number a method produces when it is run by everyone rather than by one person. We have not verified the follow-up figure beyond the article; it is Omnisend's account.
None of the four levers is free of side-effects, and a firm will usually combine two. Credit plus a share of saved hours pays the builder in both currencies. Exclusivity plus an adoption-scored raise gives a head start and then pays to end it. The one combination that does not work is credit alone with a raised baseline, which is the default a firm chooses by not choosing.
05 — The checklistThe manager's checklist: make sharing the promotion path, not the tax
Each row is a situation a manager will recognise and the move that keeps the next method from going quiet.
"If you build a way of working that others can use, we will pay for it, and here is how. If you have one now, we would rather buy it from you than find it on your last day."
The failure mode this avoids is not the builder leaving with the method, though that happens. It is the builder staying, sharing nothing further, and the firm never knowing what the second and third methods would have been. Our AI transformation practice designs the measurement and the pay rule together, because one without the other is the default many firms already have.
06 — ConclusionThe transfer is not free, and the builder already knows it
Write the pay rule for a shared method before you ask for one, and move the builder's baseline last
Firms that pay for distillation get more of it. Firms that ask for it free get the surveys' numbers: a third to over half of their AI users keeping quiet about it.