Why the Next Productivity Revolution Will Be Measured in Judgment, Not Just Hours

The strange moment we are in

What if the biggest productivity breakthrough in a profession is not that people work faster, but that the system finally learns to distinguish between work that creates value and work that merely consumes effort?

That question is becoming impossible to ignore. On one side, new tools promise something almost magical: describe the work, and an agent will do it. On the other side, some of the most labor intensive and value critical organizations in the economy are seeing something far less magical: productivity rising, reimbursement falling, compensation lagging, and subsidy costs climbing. The tension is not just about technology or money. It is about whether modern institutions can translate more human effort into more economic value, or whether we are simply getting better at producing work that the system still underprices.

This is the real story hiding inside the current productivity conversation. We tend to talk as if automation is about replacing tasks and compensation is about paying for tasks. But the deeper issue is coordination. When the work becomes harder to measure than the effort required to do it, incentives break. When incentives break, more output does not necessarily mean more reward. And when that happens, the promise of productivity becomes a trap: people are asked to do more while the system admits that what they do is worth less.

The productivity illusion: more output, less relief

At first glance, the numbers sound familiar, even reassuring. Productivity rises. Compensation rises too, though more slowly. But reimbursement declines. Costs remain stubbornly high. That combination is revealing because it shows that a gain in efficiency does not automatically become a gain in organizational health.

Think of a hospital practice like a restaurant where the kitchen gets faster every year, but the menu prices fall, ingredient costs stay high, and the staff still needs to be paid. Faster service does not solve the business model if the market refuses to recognize the value of the output. The kitchen can become more efficient, but the economics can still worsen.

That is exactly what happens when work relative value units per full time equivalent climb while revenue per unit falls. The system is asking for more throughput from clinicians, but the return on each unit of effort does not keep pace. The result is a widening work pay gap: more work, thinner margins, heavier subsidy, and a growing sense that success is being measured in the wrong currency.

This is why productivity metrics can be deceptive. They often measure activity, not meaning. A clinician can see more patients, complete more documentation, and bill more units, yet the organization can still be under strain. The apparent rise in productivity may simply reflect a system that is squeezing harder rather than thinking better.

The dangerous assumption in modern institutions is that if people are busier, value must be increasing. In reality, busyness is often what value looks like when measurement is broken.

Agents change the cost of thinking, not just the cost of labor

Now enter the agentic promise: describe the work, and the agent builds it. That sounds like a convenience story, but it is actually a redesign story. The shift is not merely that software can complete tasks faster. It is that the cost of translating intention into execution is dropping.

That matters because many organizations are not constrained by a lack of effort. They are constrained by the friction between what people know needs to happen and what actually gets done. Every department has this gap. Someone knows a workflow should be automated, a draft should be generated, a schedule should be optimized, a report should be assembled, or a follow up should be triggered. But building the solution traditionally requires time, technical skill, and coordination. Agents compress that gap.

In practical terms, this means a manager can say, “When a discharge summary is complete, draft the follow up message, update the task board, and flag any missing labs.” An operations lead can say, “When demand exceeds threshold X, rebalance staffing and notify the team.” A clinician can say, “Summarize the last three encounters, identify unresolved issues, and prepare the note template.” The point is not that the agent replaces judgment. The point is that it externalizes repetitive execution so judgment can move upstream.

That distinction is crucial. In high stakes environments, the bottleneck is rarely the existence of a task. It is the cost of making the system remember, route, check, and repeat the task reliably. Agents are interesting because they turn process design into a more conversational activity. Instead of building workflows as a permanent artifact first, you can describe the work and test the workflow in motion. In other words, the interface to automation begins to look more like management itself.

The hidden connection: both stories are about underpriced coordination

These two developments seem unrelated at first. One is about AI agents that save hours every week. The other is about physicians working more while reimbursement lags. But they are actually two sides of the same economic problem: coordination is undervalued until it breaks.

A physician’s labor is not just a sequence of billable actions. It is coordination under uncertainty. Diagnosing, deciding, documenting, communicating, and following up are all part of the same value chain. When reimbursement models reward fragments of that chain unevenly, the system may pay for visible activity while ignoring invisible orchestration. The physician becomes more productive in the operational sense, but the institution struggles to monetize the full contribution.

Agents expose the same flaw from the opposite direction. They can automate the visible fragments, the repetitive tasks that make work look expensive. But the truly valuable part of many jobs is not the repeatable step itself. It is knowing which steps matter, in what order, for which exception, under which constraints. That is coordination intelligence. And if organizations confuse automation with value creation, they will automate the easy parts and continue underinvesting in the harder ones.

This suggests a more useful framework: every role has two layers. The first is execution, the set of routine tasks that can be described, delegated, and repeated. The second is judgment, the capacity to decide what should happen when the situation is ambiguous, incomplete, or changing. Most institutions have historically measured and paid for execution more easily than judgment. But agents are shifting the economics of execution, which should force a rethink of what is scarce.

When execution becomes cheap, judgment becomes the premium asset.

That is the deeper connection between an agent platform and the physician work pay gap. Both reveal that the old metric of value, hours worked or tasks completed, is becoming less reliable. The more a profession depends on interpreting messy reality, the less its true contribution can be captured by simple throughput measures. The more software handles repetitive execution, the more human work must be evaluated by the quality of decisions, exceptions resolved, and downstream outcomes improved.

A new mental model: the three kinds of work

To make this practical, it helps to separate work into three categories.

1. Repetitive work: steps that can be specified, repeated, and checked. Examples include drafting routine messages, routing forms, summarizing records, scheduling follow ups, and updating systems.

2. Coordinative work: work that makes other work happen. Examples include handoffs, exception handling, escalation paths, prioritization, and cross team alignment. This is the glue layer that most organizations underestimate.

3. Judgment work: work that requires interpretation, tradeoffs, and accountability. Examples include diagnosing atypical cases, setting policy, resolving conflicting signals, deciding what matters now, and knowing when not to automate.

Many institutions mistakenly pay attention almost exclusively to repetitive work because it is easiest to observe. But repetitive work is also the cheapest to automate. Coordinative work is harder to see, yet it often determines whether the organization actually functions. Judgment work is the rarest and most expensive, but it is also the source of resilience.

This matters for healthcare, but it applies everywhere. If a practice pays for repetitive volume while neglecting coordinative load, the people who keep the system coherent get squeezed. If a company adopts agents to eliminate repetitive work but does not redesign roles, it risks hollowing out the very layer that keeps quality from collapsing. Automation should not simply remove labor. It should reveal where human value really sits.

A useful test is this: if an agent can handle a task, is that task truly the core of the role, or just the visible residue of a broken process? In many cases, the most annoying part of work is not the thing that matters most. It is the thing that survives because no one has had time to redesign the system.

What leaders should do next

The temptation in a moment like this is to chase one of two extremes. One is the technophile fantasy: deploy agents everywhere and assume hours saved equals value created. The other is the defensive stance: focus only on reimbursement and compensation and treat automation as a distraction. Both miss the point.

The productive path is to use agents as a diagnostic tool. If a workflow can be described clearly enough for an agent to execute it, then the organization should ask three questions. First, should that workflow exist at all in its current form? Second, what judgment is being buried inside what looks like routine work? Third, if the routine part becomes cheaper, how should the human role be redefined so the savings translate into better outcomes rather than just higher expectations?

In healthcare, that might mean using agents to pre draft notes, surface missing data, and prepare follow up instructions, while explicitly protecting clinician time for diagnosis, patient communication, and exception management. In other industries, it might mean automating status updates and routine reporting so managers can spend more time on decisions, coaching, and system redesign. The goal is not to make people disappear into the software. The goal is to make them more human at work by stripping away the parts of work that only masquerade as value.

There is also a compensation lesson here. If productivity rises but pay lags, organizations should stop pretending the old distribution of value still works. They need to ask whether compensation is aligned with the parts of work that are becoming scarcer, not the parts that are becoming cheaper. In an agent rich environment, the premium is less likely to belong to the person who can do the most tasks and more likely to belong to the person who can define the right task, detect the exception, and improve the system.

Key Takeaways

• Stop measuring work only by volume. Volume can rise even when value capture falls. Track whether more output is actually improving margins, outcomes, or decision quality.
• Use agents to expose hidden friction. If a workflow is easy to describe to an agent, it is probably also easy to audit, simplify, or eliminate.
• Separate execution from judgment. Protect the time and incentives for high value decision making, especially in roles where routine tasks dominate the calendar.
• Redesign compensation around scarce contribution. Pay attention to coordination, exception handling, and downstream impact, not just visible throughput.
• Treat automation as a management tool, not only a labor saving tool. The best use of agents may be to reveal where the real bottlenecks and false costs live.

The real question

The future of work is not mainly about whether machines can do more tasks. It is about whether institutions can finally see the difference between doing more and creating more value.

That is why the promise of agents and the widening work pay gap belong in the same conversation. One shows that execution is becoming cheaper to produce. The other shows that effort is still too often underrewarded even as output rises. Together, they point to a more uncomfortable and more useful conclusion: productivity is no longer a sufficient proxy for value.

The organizations that thrive next will not be the ones that simply automate the most. They will be the ones that learn to pay for judgment, redesign around coordination, and use agents to make hidden work visible. In that world, the central skill is not just getting work done. It is knowing which work should matter, who should do it, and what the system should finally stop pretending is the same as value.