Trust and quality notes
- Last updated
- September 9, 2026
Faster tools do not automatically make a sustainable working life. They can shorten one task while raising the number of decisions, conversations, and unfinished threads competing for attention. If your days already feel scattered, adding more automated work may increase the strain rather than relieve it.
Jonny Miller argues that this human constraint will become central as agentic workflows spread. In his original essay [1], he says the limiting factor will not be technical capacity but our nervous systems. His practical response is not to reject automation. It is to train the ability to notice strain, change our state, and process emotion before pressure accumulates.
Automation can raise the pace without reducing the load
Miller begins with costs he believes are already visible: mental exhaustion from intensive AI use, surrendering too much cognition to tools, and less human contact at work. His concern is about cumulative load. A tool may complete an assignment quickly, but the person still has to frame the request, judge the result, switch contexts, resolve ambiguity, and decide what happens next.
That distinction matters for anyone new to AI. Automation can reduce manual effort while increasing cognitive throughput. You may handle more inputs in the same afternoon, but your capacity to absorb them has not expanded at the same rate. The result can be a strange mix of speed and depletion.
Miller does not claim the pace will reverse. He assumes it is unlikely to slow and asks how people can remain capable within it. His answer has three parts: interoception, self-regulation, and emotional fluidity.
My interpretation is that automation changes the definition of work readiness. Knowing how to use a tool is only one layer. A person must also know when their judgment is degrading, how to recover a workable state, and how to keep unresolved frustration from leaking into the next decision.
Notice overload before it becomes a crash
Miller’s first practice is interoception, the ability to sense what is happening inside your body. He uses the metaphor of noticing when you are reaching the limit of an internal token window. The point is to detect stress when it is still manageable, rather than after it has become an implosion.
In ordinary work, early signals might include compulsive tab switching, shallow breathing, an inability to read a paragraph twice, or the urge to answer every new notification immediately. Those examples are interpretive, not Miller’s checklist. His stated principle is earlier awareness. If you notice strain sooner, you have a better chance of preventing impaired decisions.
This is a useful correction to the way productivity is often measured. A completed task tells you nothing about the condition in which it was completed. If a fast morning leaves you unable to think clearly in the afternoon, the apparent gain may have displaced rather than removed a cost.
A simple practice follows from Miller’s argument: create moments in which you can actually notice your state. Before starting another automated run or opening another thread, pause long enough to ask whether you are focused, agitated, numb, or depleted. The aim is not perfect self-knowledge. It is catching a trend before it controls the rest of the day.
Learn to change gears deliberately
The second capacity is self-regulation. Miller describes it as deliberately shifting upward into alertness or downward into relaxation. In a high-throughput environment, both directions matter. Some tasks require energy and attention. After intense work, the ability to wind down matters just as much.
He also connects regulation with staying grounded rather than collapsing into binary thinking. Under pressure, complex questions can start to look like urgent yes-or-no choices. A generated answer may further encourage premature certainty because it arrives polished and confident. Regulation creates enough internal space to examine the answer instead of simply reacting to it.
My interpretation is that the most valuable pause in an automated workflow may occur after the output appears. That is the moment when speed can turn into haste. A regulated person can ask whether the result fits the situation, what remains uncertain, and whether another human needs to be involved.
This does not require turning every workday into a wellness program. It means treating state changes as part of competent execution. If an intense session leaves your mind racing, recovery is not separate from the work. It helps preserve the judgment needed for tomorrow’s work.
Process emotion instead of carrying debt
Miller’s third practice is emotional fluidity. He argues that people need to notice and express a fuller range of emotions at work so disappointment and frustration can be processed in the moment. Otherwise, feelings become what he calls emotional debt.
Automation does not remove the emotional texture of work. A system can fail unexpectedly. A colleague can disagree with an output you considered strong. A role can change in ways that bring grief, relief, fear, or excitement. Faster production may even create more moments of feedback and correction.
The source’s argument is that bottling these reactions creates accumulation. My interpretation is that teams using more automation will need better language for the human experience surrounding it. “The workflow succeeded” and “this change is hard on me” can both be true. Treating the second statement as irrelevant does not make it disappear.
Emotional fluidity also protects relationships. Unprocessed irritation often appears later as abrupt messages, rigid decisions, or withdrawal. Recognizing the feeling earlier gives a person more choice about how to communicate it.
A human capacity worth designing around
Miller’s picture of humanity after automation is not one of passive users watching machines work. It is a picture of people managing far more cognitive motion and needing stronger internal skills to do so without breaking down. Technical systems may scale quickly. Human attention, emotion, and recovery still require care.
That suggests a practical design principle: do not evaluate an automated workflow only by what it produces. Look at what it asks of the people around it. Does it multiply alerts? Does it invite constant switching? Is there a clear stopping point? Can someone challenge the result without feeling that they are slowing everything down? These questions are my application of Miller’s thesis, not claims made directly in his essay.
The humane use of automation is not merely removing tasks. It is building a pace at which people can still sense, judge, relate, and recover. Miller’s three capacities offer a grounded place to begin.
If you want help choosing automation that supports the people doing the work, start with one process at Agentic Workers.