Version 1.0 · July 2026
“They themselves are makers of themselves.” James Allen, 1903

Published by Teal Agility. Free to read, free to share.

Preface

We wrote this Guide for a simple reason. Technology can now perform real parts of real jobs, and that has changed the most important question at work: no longer whether the machine can do something (increasingly, it can), but whether you should hand it over, and how you stay in charge of the outcome when you do.

The word “augmented” simply means extended: Augmented Work is work in which technology extends what you can do, while you stay in charge of it.

If you have ever asked a system to draft a message and then fixed it before sending, you have already practiced augmented work without the name. This Guide is not tool training; it teaches the part the tutorials skip, the part that stays valuable as the tools change: the judgment.

And if part of you is afraid of what these systems mean for your work, you are in good company; so are many of the people who build them.

This Guide was written to replace that fear with something sturdier: the ability to decide.

A map tells you which road to take; a compass gives you direction and trusts you to choose the path. This Guide is a compass. It prescribes no tools, no vendors, no steps, because those change constantly; it offers only the guidance that does not change: how to keep human judgment in charge of work that machines increasingly help perform.

Technology does not make the professional; it reveals them. A careless person with a powerful tool produces careless work faster. A rigorous person becomes more rigorous still. This Guide exists so that what the tools reveal in you is worth revealing.

If you get value from this Guide, consider yourself the person it was written for, whatever your job title, industry, or level of experience.

Every term this Guide uses in a special sense is explained where it first appears, and collected in Terminology, near the end.

Definition of Augmented Work

Augmented Work is the practice of performing knowledge work with systems while a human deliberately directs the work, judges the results, and remains accountable for the outcome

It comprises the following four judgments working in tandem:

1. Deciding the handover
2. Verifying in proportion to stakes
3. Governing the return
4. Directing the dividend

A person who exercises these judgments is a practitioner. Every practitioner exercises the first three; the fourth belongs to practitioners whose decisions govern other people’s work. The judgments are collectively referred to as augmented practice.

Practitioners may refer to them by any names they choose, provided they are exercised as described in this Guide.

Why Practice Augmented Work?

Most work exists to create value. Machine output, by itself, is not value; it becomes value only when human judgment directs it, verifies it, and answers for it.

The aim of Augmented Work is therefore better value through better judgment.

Better judgment means striving for the right balance of three elements:

• Combined capability: The practitioner and the system, working together, achieve better results than either alone.
• Accountability: a human can always answer for what was produced and how.
• Return: the time and capacity gained are deliberately directed, not silently absorbed.

Practitioners should aim for a sustainable balance among these three, because each fails without the others:

• Capability without accountability produces impressive work nobody stands behind.
• Accountability without capability wastes the technology.
• Capability and accountability without a governed return produce only busier people.

Because these judgments apply to virtually any knowledge work, Augmented Work is limited to no industry, role, or seniority; it can be practiced by one person or an entire organization.

The Human Foundation

The practitioner, not the system, is the foundation of augmented practice.

Systems generate output. Everything below comes from the practitioner, and no system supplies it:

• Intent: deciding what is worth doing and why.
• Conscience: knowing right from wrong and acting on it at a cost.
• Care: the empathy and trust shown to other people.
• Taste: recognizing what is genuinely good, not merely plausible.
• Meaning: knowing what the work and its return time are for.

Hand over the task; keep the skill:

• Judgment is a skill, and skills fade without use. The loss is discovered only on the day the system is unavailable; this is quiet deskilling.
• A practitioner who can neither do a task nor check it another way can no longer honestly sign off on its output. So, keep practicing the core craft unassisted, some of the time (the shop owner still counts the drawer the register totaled; the designer still sketches).

Sustainable practice is part of the practice. Augmentation that increases output while degrading the practitioner’s attention, thinking time, or rest has failed, whatever the numbers say; the pace of work is part of the quality of judgment.

This Guide prescribes no particular division between assisted and unassisted work. How much to retain, and which skills, is the practitioner’s call, made deliberately rather than by drift.

Augmented Work Judgments

1. Deciding the Handover

The purpose of the handover decision is to consciously place every task in human hands, in system hands, or in both.

For any task, a practitioner decides among three options:

• Retain the task and perform it unassisted, whether by choice or because this kind of task must never go to a system at all.
• Share the task, with the system contributing and the human directing and finishing.
• Delegate the task to a system, within explicit limits.

Sharing and delegating may also create new work that neither human nor system could produce alone (for example, a small shop that now thanks every customer personally, something there was never time for before); practitioners should treat such created work as a result to pursue, not an accident.

A practitioner may delegate any suitable task to a system. Regardless, the practitioner remains accountable for the outcome. Delegation enables focus; delegation without accountability is abandonment.

A handover is a decision, not a default:

• Work drifting to a system because the system was open is not a handover; it is an accident with the same consequences.
• Any acceptable standing handovers, tasks that always go to a system, should be made explicit rather than left implicit.

The handover decision must consider, at minimum, all of the following:

• the stakes of the task;
• whether its output can be verified at reasonable cost;
• whether the information involved may enter the system at all; and
• whether the decision affects a person’s life.

No minimum or maximum share of work must be delegated; the right share is whatever the practitioner can remain accountable for.

2. Verifying in Proportion to Stakes

The purpose of verification is to turn system output into something that deserves a human’s signature.

Verification enables trust; trust without verification is gambling.

How much checking is enough:

• A practitioner must match the depth of checking to the stakes: the more damage a wrong output could cause, the more carefully it must be checked (for example, an everyday email the tool drafted may only need a quick read before you send it; the balance your banking app shows should be cross-checked against your own records before it enters your annual tax report).
• Trust is earned one task at a time and one system at a time. A tool that proved reliable at one job has not proved itself at every job; trusting the tool as a whole is the mistake. This is calibrated trust.
• Earned trust does not last forever. Systems change quietly, often without announcement, so trust should be refreshed from time to time, not only when a change is noticed, and always when the stakes rise.
• When a practitioner does not yet know how much damage a wrong output could cause, they should treat the stakes as high until they know.

Where proper checking is impossible, or would cost more than the task is worth, the judgment moves upstream, back to the handover decision itself:

• The handover itself must be reconsidered.
• Until it is, unverified output must be treated as a claim, not a fact.
• Impossibility of checking is never a license to trust; it is a signal that this task needed different limits, a different expert in the loop, or no handover at all.

There are no specific guidelines for how verification should look. Reading, testing, recalculating, comparing sources, and consulting a qualified colleague are all forms of verification.

Verification begins at tool choice. Prefer systems that make checking easy: ones that show their sources, admit their uncertainty, and reveal their assumptions.

A system that hides how it reached an answer raises the cost of every check, and that cost belongs in the handover decision.

Output arrives as well as departs; much of what reaches a practitioner came from other people’s systems. The same judgment applies in reverse:

• Output received is still output; the depth of checking still follows the stakes.
• Polish is not proof. Fluency now costs nothing; well-written no longer means well-founded.
• Provenance may be asked for, not only owed; asking how a result was made and checked is a professional question, not an accusation.
• Pressure is a signal. Urgency, strong emotion, or too-good-to-be-true: each is a reason to slow down, not to hurry.

3. Governing the Return

The purpose of governing the return is to ensure that what augmentation gives back is directed by a decision rather than absorbed by default.

Returned time enables growth, but without a decision, it just means more work ahead. The practitioner’s obligations:

• Identify the return: work out roughly how much time or capacity was actually gained.
• Name it: say where the saved hours are going, out loud or in writing. Left unnamed, returned time is silently backfilled, absorbed by more of the same work with nobody having chosen that.
• Direct it on purpose: toward deeper work, learning, other people, or life outside work.

An employed practitioner does not fully control the return; their organization does. Their part is to see it clearly and negotiate for it knowingly; governing it across a team or organization belongs to the fourth judgment, below.

Reviewing the return can occur continuously, at regular intervals, or both; this Guide prescribes no schedule. What it requires is that the review happens, because a return nobody examines is a return somebody else has already spent.

4. Directing the Dividend

The fourth judgment belongs to leaders. A leader, in this Guide, is a practitioner whose decisions govern other people’s work; leadership does not replace the first three judgments; it adds this one.

The purpose of directing the dividend is to decide, honestly and on purpose, what an organization does with the capacity that augmentation creates.

When systems make a team more capable, the organization receives a dividend: extra capacity, speed, and hours. The leadership judgment is what to do with it. The options:

• reduce people to reduce cost;
• produce more with the same people; or
• reinvest in the people themselves: their growth, their work, and time genuinely returned to their lives.

This Guide does not prescribe the outcome.

A restructuring can be sound, and survival can require hard decisions.

What it binds is the process.

A leader practicing Augmented Work must be able to show all of the following:

• that alternatives to reducing people were genuinely explored, with a record;
• that people affected were treated with dignity and given real help through the change;
• an honest account of where the dividend actually went;
• that returned time was governed on purpose, not silently backfilled;
• that limits for the systems were set before they were put to use, not after something went wrong; and
• that capability grew in the people, not only in the software.

A leader who restructured with integrity can meet this standard. A leader who reinvested thoughtlessly can fail it. The standard judges how leaders decide and how they treat people, not whether their decisions were comfortable.

Practitioners who do not lead still have use for this judgment: it describes what they may reasonably ask of those who lead them.

When Work Itself Is Redefined

Many people, including highly skilled ones, carry a quiet fear: that these systems will take their job. The fear deserves a straight answer, not a slogan.

A job is a bundle of tasks, not a single thing. Systems take over tasks; what happens to jobs is decided by people. When cash machines arrived, bank tellers were widely expected to disappear. Instead, branches became cheaper to run, more of them opened, and tellers’ work moved toward what machines could not do. The job was not deleted; it was rebuilt. Nothing guaranteed that outcome; people chose it.

A practitioner should periodically apply the handover judgment to their own role as a whole:

• which of its tasks are moving to systems;
• which are becoming shared;
• which remain human; and
• what new work the combination now makes possible.

A role examined this way is being redesigned by the person who holds it. A role never examined is being redesigned by someone else.

This Guide does not promise that no work will be lost; that would not be honest. What it holds is this: the future of a role is decided task by task; the person best placed to make those decisions is the one holding the role; and what an organization does with the capability its people gain is a judgment (Directing the Dividend), never a law of nature.

People met earlier industrial revolutions unprepared and learned their new place after the fact.

This Guide exists so that, this time, working people meet the change ready: deciding, task by task, what their work becomes.

Practicing Together

Few practitioners work alone. Where people work together, the rules of augmented practice should be shared, not improvised privately by each person. At minimum, those who work together should make explicit:

• which kinds of work may be handed over, and which never;
• what information may enter which systems; and
• how work made with systems is disclosed to one another.

Making these agreements explicit is not bureaucracy; it is kindness. It spares each person from guessing alone, and it turns scattered private judgments into a practice a team can trust.

This Guide prescribes no format for such agreements; a single page everyone has read beats a policy nobody has.

What a Practitioner Answers For

A practitioner is answerable, at minimum, for all of the following.

Each is a duty of the person, not a feature of the tool.

• Fairness. Systems learn from the past, and the past was not fair to everyone. Before relying on a system whose output affects people, a practitioner must be able to state who it could treat unfairly and how that unfairness would be detected. Inability to answer either question means the system is not ready to be relied upon.
• Privacy. Information about other people must not be entered into a system without the authority to do so. A client’s file, a colleague’s message, a patient’s record: such decisions are rarely the practitioner’s alone to make. Uncertainty about that authority is itself the instruction: stop and check first.
• Human decisions about humans. Wherever an outcome changes a person’s life, including hiring, lending, grading, and diagnosing, a human must genuinely own the decision. Reading a system’s recommendation about a person is not the same as considering the person.
• Honesty. Systems are used to do work, never to fake work, fake evidence, or fake a human being. Presenting unverified output as verified fact is a lazier kind of deception.
• Disclosure. A practitioner must be able to state what they did, what systems did, and how the result was checked; this account is the work’s provenance. A practitioner who cannot state what a system contributed must verify or redo the work rather than guess. When others are talking to a system, being assessed by one, or reading its words, they deserve to know; when in doubt, disclose.
• Accountability. Accountability can’t be delegated; responsibility remains with the practitioner, who selects the tool, decides on handover, and sets the required level of verification. The outcome, desired or not, belongs to them. Where an organization sets the conditions of work, it answers for those conditions.

The Boundary

Augmented Work has one edge, and it is worth marking precisely.

The moment a person stops judging and merely approves, augmentation ends, and automation begins.

Automation (work performed by systems with no human exercising judgment inside it) is sometimes appropriate.

It is also a different thing, with different rules, and it is outside what this Guide describes. Approval that has become a formality is not judgment; a signature without verification behind it marks the boundary already crossed.

Readers may know the phrase “human in the loop“. A loop can contain a human and still contain no judgment.

Placing a person at a checkpoint is an arrangement; this Guide describes what that person must bring, and keep, for their presence to mean anything.

Human in the loop indicates where the human sits. Augmented Work says what the human must still be able to do when they get there.

Terminology

• Augmentation dividend: the capacity, speed, or hours an organization gains when work is augmented; the organizational form of the return.
• Augmented practice: the exercise of the four judgments, working in tandem.
• Calibrated trust: trust in a system that is earned one task at a time, matched to the stakes, and checked fresh when stakes or systems change.
• Handover: the deliberate decision to retain a task, share it with a system, or delegate it to a system.
• Human in the loop: a common arrangement in which a person reviews or approves a system’s output at set points. The arrangement alone guarantees nothing; augmented practice is what makes the human in the loop meaningful.
• Judgment: the human act of weighing what matters and deciding, especially where no rule or tool gives the answer.
• Knowledge work: work whose main material is information and decisions rather than physical objects. It happens wherever people think, write, plan, design, create, analyze, or decide, whatever the industry.
• Output: whatever a system gives back: text, a number, an image, a design, a recommendation, or an action.
• Practitioner: any person who performs knowledge work with the help of intelligent systems and exercises the judgments described in this Guide.
• Provenance: the account of what the practitioner did, what systems did, and how the result was checked.
• Quiet deskilling: the gradual, unnoticed loss of an ability caused by no longer practicing it, discovered when the system that replaced it is unavailable.
• Return (the): the time or capacity gained when work is augmented; its organizational form is the augmentation dividend.
• Silent backfill: the absorption of returned time by additional work without a deliberate decision.
• Stakes: the consequences of an output being wrong, for the practitioner, for others, or for the organization (for example, a typo in a rough draft is low stakes; a wrong total in your tax report is high).
• Standing handover: a task category explicitly agreed to always go to a system.
• System: any technology capable of performing parts of knowledge work (for example, chat assistants that write and answer questions, tools that generate images and designs, or apps that total, sort, and fill in numbers). The Guide uses this word, so it remains true regardless of which technologies exist.
• Value: a potential or realized benefit for anyone the work touches, including the customer, the organization, the practitioner, and the wider community.

Endnote

One can, and likely should, add other principles, methods, and techniques to augmented practice, from any tradition and any profession. Still, the four judgments, the duties a practitioner answers for, and the spirit of human accountability must be preserved.

Start anywhere. The next time a system hands you an output, ask what it would take for that result to deserve your signature. That is the whole practice, begun.

Technology will keep changing. The judgment this Guide describes will not expire with any version of any tool, because it never lived in the tools. It lives in you. Keep your hand firmly upon the helm of judgment and let the time the machines return to you be spent on the things only a human life can hold: family, craft, thought, and rest.

History

The oldest source in this Guide predates every computer.

In 1903, James Allen wrote that people are makers of themselves through what they choose to cultivate; Augmented Work applies that claim to an age in which what we cultivate includes our machines.

The augmentation tradition proper began in the early 1960s, when J.C.R. Licklider described human and machine working in symbiosis, and Douglas Engelbart outlined a program for augmenting human intellect.

This Guide continues that tradition into an era when their question finally became urgent for everyone who works.

Acknowledgments

This Guide is open for review by practitioners and readers from different countries, industries, and walks of life. Contributions, corrections, and challenges are welcome at hello@tealagility.com; a guide earns trust through the people who put it to work.

We thank the following people for improving this Guide through their review:

• Nuzli Hernawan (Jakarta, Indonesia), for pressing the Guide to address how systems should support the people who verify them.

References

• Allen, James. As a Man Thinketh. 1903.
• Licklider, J.C.R. “Man-Computer Symbiosis.” IRE Transactions on Human Factors in Electronics, 1960.
• Engelbart, Douglas C. Augmenting Human Intellect: A Conceptual Framework. Stanford Research Institute, 1962.
• Bainbridge, Lisanne. “Ironies of Automation.” Automatica, 1983.
• Jesuthasan, Ravin, and John W. Boudreau. Reinventing Jobs: A 4-Step Approach for Applying Automation to Work. Harvard Business Review Press, 2018.
• Gasser, Urs, and Viktor Mayer-Schönberger. Guardrails: Guiding Human Decisions in the Age of AI. Princeton University Press, 2024.
• Laloux, Frederic. Reinventing Organizations. Nelson Parker, 2014.

© Teal Agility.

This publication is offered for license under the Attribution ShareAlike license of Creative Commons, accessible at http://creativecommons.org/licenses/by-sa/4.0/legalcode and described in summary form at http://creativecommons.org/licenses/by-sa/4.0/.

By using this Guide, you acknowledge that you have read and agree to be bound by the terms of the Attribution ShareAlike license of Creative Commons.