The Reward for Competence Is Often More Work

When AI makes work faster, the first instinct is to assign more of it.

If a tool allows someone to produce a report in one hour instead of four, the first instinct is to celebrate the three hours saved. It’s a real gain. But ask what happens to those three hours. Do they become space for better judgment? Verification? Learning, mentoring, documentation, recovery? Or do they become three more reports?

I have been thinking about this since a meeting yesterday.

The discussion was about AI tools and how they could help teams work faster. Much of the time-consuming work, such as drafting, summarizing, reviewing, and preparing reports, becomes faster. In the public sector, where teams carry large mandates with limited staff, this sounds like a gift.

Then someone just said it: if people become faster and better at their work, the organization may simply ask them to do more.

This is one tension raised in many conversations about AI and productivity. We talk as if efficiency creates relief. It does not always. Sometimes it creates capacity. Then, the system looks at that capacity and says, “Wonderful. We have room for more.”

I am not immune to this. With the agents I have built, the temptation is constant. One system leads to another idea, which leads to another prototype, which leads to a small fleet of tools I did not plan to build. The speed makes it easy. At some point, I have to stop, step back, and inject some discipline, because the rabbit hole of creating and creating is its own version of the same problem. Capacity expands. The question of what it is actually for still has to be asked.

This is the pattern, at every scale.

The reward for competence is often more work.

Many people know this without needing a framework. The reliable person, sometimes the office pabibo, gets invited to more committees. The colleague who writes well gets asked to review more documents. The team that delivers early becomes the team everyone calls during a crisis. The person who knows how to fix the spreadsheet becomes the spreadsheet person for the next five years. This is how reputations are built, and it is also how people end up overloaded.

I am reminded of Jevons paradox. From energy economics: when a technology makes a resource more efficient to use, total consumption can increase rather than decrease. More efficient engines made coal cheaper to use, which encouraged wider use across industries. The unit cost fell, so demand rose.

Human attention is not coal, though some Mondays make the comparison tempting.

When AI lowers the cost of producing certain kinds of work, organizations may demand more of that work. More summaries, dashboards, decks, more “quick asks”... which are never really as quick as advertised. Consumption rises because the price has dropped.

We see this at ECAIR. We operate as an AI-first organization; and, where appropriate, first drafts are written with AI. For standard templated documents, the workflow is straightforward. For higher-stakes cases, the document type determines how much human involvement is required before it moves forward.

What does not change is everything that comes after the draft. Someone still has to check whether the output is correct, know what is important, understand the context, the politics, the consequences, the edge cases, and the people affected by the decision. AI can help produce the artifact. Accountability still lands on a human desk… usually, the desk of a competent person. And competence, as it turns out, attracts more work.

I learned a low-stakes version of this as an undergraduate.

We had a small research team, and every week we gave updates to our research leads. We were productive… sometimes, very productive. But we also learned to pace ourselves. We would not always present everything we had. We kept some bala for the next meeting. Enough to show progress. Not enough to reset expectations.

This was, of course, our brilliant strategy. I am now convinced our research advisors knew exactly what we were doing and simply let us enjoy the illusion. Very generous of them.

But underneath our very sophisticated student strategy was a real instinct. We understood, even without naming it, that if we showed everything at once, the bar for the next meeting would move. More output would become the new normal. The system would learn how much we could produce at full sprint and treat that as our walking speed. Competence can trap you this way.

Later, in Singapore, I saw a healthier model. Our scientific director was explicit about the deal: if we met our deliverables ahead of time, we could use the remaining time for research driven by curiosity and personal interest. Finishing early did not mean being loaded with more work. The reward for competence was autonomy. It worked because finishing early did not just benefit the individual; it kept the system from running too hot.

Complex systems need slack. A system running at full capacity all the time becomes brittle. In organizations, the signs are tired teams, rushed judgment, poor documentation, shallow analysis, and a strange dependence on the same few reliable people. Everyone knows who they are. They get more things, then urgent things, then things that require “just a quick look.” The organization becomes dependent on a few people’s ability to absorb overflow.

This is risky for two reasons. People burn out. And the system never learns how much invisible work was being held together by competence, goodwill, and caffeine.

If AI helps a teacher prepare materials faster, some of the saved time should go into better feedback, some into rest, some into doing nothing for a few minutes, which remains an underrated technology. If AI helps a government team summarize public feedback faster, some of the saved time should go into checking whether the summary captured minority views or signals that cut against the consensus. Faster processing should not mean thinner listening.

The question I’d ask is not “How much more can we now assign?” But, “What should this new capacity be for?”

Some of it can go to output. Organizations have mandates, deadlines, and real constraints. I am not arguing for everyone to disappear into a hammock of self-actualization; though, to be clear, I am personally open to field-testing this model.

The point is proportion. If every efficiency gain is immediately consumed by additional tasks, AI will intensify work rather than improve it. People will produce more but think less. They will respond faster but recover less. The dashboards will look healthy, and nobody will be able to explain why everyone is tired.

Some saved time should become buffer. Some should become thinking time, quality control, or rest. And the work that does not show up in the artifact—prompting, reviewing, verifying, editing, contextualizing, coordinating, owning the final output—is still work. If leaders only see the finished artifact, they will underestimate the human effort still required.

For managers: watch where work accumulates. The people holding things together are often the ones nobody worries about because they never seem to be struggling.

I say this as someone who does exactly the thing I am cautioning against. I rely more on the people I can rely on. It is genuinely hard to catch yourself doing it. What I try to do, when I remember, is check in. A simple “if things are getting too much, let me know” helps. Better still is noticing before people have to say it.

I hope my team remembers this sentence when the time comes. 😄

Another thing that helps: milestones, not as pressure, but as a guide. If someone finishes before a scheduled milestone, that window belongs to them. Then the next milestone arrives, and the work picks up again. It is a small structure, but it protects slack without having to negotiate for it every time.

For institutions, the task is to decide what the saved capacity is actually for. Quality, learning, service delivery, recovery, innovation, documentation, public trust. The answer cannot always be volume.

AI is already making many forms of work faster. The real test is what institutions do after the work gets faster.