1 / 25

Longhand AI

Longhand AIOverview

Choose the work. Set the controls. Define the result. Make the decision. Keep human judgement and exact execution visible.

7 decisionsAbout 70 minutes25 slidesProgress saved

The finish line

One AI Work Decision Card

You will leave with one real workflow classified, measured, bounded, and ready for a build — or an intentional decision not to use AI.

Progress is saved in this browser.

Longhand AI

Longhand AIOverview

Choose the work. Set the controls. Define the result. Make the decision. Keep human judgement and exact execution visible.

7 decisionsAbout 70 minutes25 slidesProgress saved

The finish line

One AI Work Decision Card

You will leave with one real workflow classified, measured, bounded, and ready for a build — or an intentional decision not to use AI.

Progress is saved in this browser.

The seven decisions

Turn AI possibility into one accountable work decision.

Match the work to the right gear

Decide whether the procedure should become a Skill

Turn responsible-AI principles into operating controls

Define the result, baseline, contributions, and remeasurement with TDAR

Make a challengeable Big What, Why, What, How, and What If decision

Separate AI work from deterministic execution

Place human judgement, review, and accountability

The output is not another list of AI ideas. It is one reviewable decision about work that matters, how it should run, and who remains accountable.

The Longhand model

Human judgement drives every gear.

Each gear owns a larger unit of responsibility. The leader still sets direction, defines the boundary, and reviews the result.

Chat

Use when you need a quick answer or idea.

Assistant

Use when you need a document made.

Agents

Use when you need a multi-step task done.

Loops

Use when you need a system that runs itself.

Shift up only when the unit of work requires it. Maximum autonomy is not the goal; useful ownership with visible accountability is.

Gear 1

Chat

One question or idea

Use it when

you need a quick answer or idea.

Example

“What are the three biggest risks in this plan, based on the attached assumptions?”

Human boundary: A person supplies the context and decides which risk matters now.

Gear 2

Assistant

One finished deliverable

Use it when

you need a document made.

Example

“Create a board-ready memo from these approved sources and this reporting template.”

Human boundary: A person owns the audience, claims, tone, and approval to share.

Gear 3

Agents

One multi-step objective

Use it when

you need a multi-step task done.

Example

“Research the competitors, compare the evidence, and produce a review-ready recommendation.”

Human boundary: A person defines scope, permissions, stop rules, and the acceptance decision.

Gear 4

Loops

One recurring outcome

Use it when

you need a system that runs itself.

Example

“Every Monday, gather current signals and produce a linked priority brief.”

Human boundary: A person owns operating policy, exceptions, and changes to what the loop may do.

The future of work

Stop building one giant agent.

General agents bring raw intelligence. Agent Skills add the repeatable procedure that turns a brilliant generalist into a reliable professional teammate.

The Future of Work 2026 · 7 minutes

Watch: The Future of Work 2026

youtube.com/watch?v=TXiaDHn1dUk

A useful computing analogy

Skills are the apps of the AI era.

01 · ModelsThe processorRaw intelligence and reasoning potential.

02 · Runtime + connectionsThe operating systemThe environment, tools, data, permissions, and agent loop.

03 · SkillsThe applicationsSpecialized procedures that package expertise for reliable reuse.

Domain experts can capture how work is really done. The team library compounds, and a new teammate can inherit working knowledge instead of starting from zero.

Step 2 of 7 · 9 min

Build a library of professional procedures instead of one giant agent.

Package the procedure a colleague could reuse

A general model can reason, but repeatable professional work also needs the procedure, sources, tools, examples, and quality rules a good colleague would learn over time.

Do this

2Package the instructions, approved sources, tools or scripts, good examples, Definition of Done, and stop rules.
3Keep large supporting material outside the standing instructions and load it only when the task needs it.
4Apply the Colleague Test: another authorized person should be able to run the Skill and recognize a good result.

A useful Skill carries the procedure

Part	What it holds
Instructions	The sequence, decisions, and boundaries.
Approved sources	The evidence and context it may use.
Tools and scripts	The reliable actions it can run.
Good examples	The quality pattern worth repeating.
Finish and stop rules	What done looks like and when to stop.

Step 3 of 7 · 12 min

Use Govern, Map, Measure, and Manage before giving a system access or authority.

Responsible AI is an operating cycle

Principles do not control a workflow. A reviewable risk-control record makes the authorized purpose, affected people, permissions, tests, accountable humans, stop rules, and escalation boundaries visible before an AI system can create consequences.

Do this

1Govern: name the authorized purpose, prohibited secondary uses, accountable humans, applicable rules, change approval, review cadence, and pause or termination authority.
2Map: inventory data, sources, sensitivity, retention, affected people, downstream effects, permissions, and every action that is read-only, reversible, approval-gated, external, or prohibited.

Responsible AI as an operating cycle

Function	Decision it makes visible	Minimum evidence
Govern	Why may this exist, who owns it, and who can stop it?	Purpose, owner, rules, change control, stop authority
Map	What data, people, permissions, and consequences are in scope?	Data flow, affected parties, actions, authorization
Measure	How will failure be found before reliance?	Cases, thresholds, reviewer tests, retained results
Manage	Which controls operate, and what happens when they fail or change?	Controls, monitoring, incident, regression, retirement

Step 3 of 7 · 12 min

Use Govern, Map, Measure, and Manage before giving a system access or authority.

Tests, controls, and stop rules must operate

Do this

3Measure: define task-specific quality and safety measures, normal, boundary, adversarial, missing-data, and prohibited-action tests, plus thresholds and evidence to retain before testing begins.
4Manage: use layered controls such as minimization, least privilege, deterministic validation, approval gates, monitoring, incident response, regression testing, fallback, and retirement procedures.
5Stop and escalate when authority, consent, facts, source currency, policy, jurisdiction, qualified review, or the ability to contest a consequential decision is unclear.

Remember: This practice record does not establish legal compliance, professional approval, or universal safety. High-consequence work still requires the relevant organizational and qualified professional review.

Step 3 of 7 · Run it

Create the risk-control record

Do this

1Govern: name the authorized purpose, prohibited secondary uses, accountable humans, applicable rules, change approval, review cadence, and pause or termination authority.
2Map: inventory data, sources, sensitivity, retention, affected people, downstream effects, permissions, and every action that is read-only, reversible, approval-gated, external, or prohibited.
3Measure: define task-specific quality and safety measures, normal, boundary, adversarial, missing-data, and prohibited-action tests, plus thresholds and evidence to retain before testing begins.
4Manage: use layered controls such as minimization, least privilege, deterministic validation, approval gates, monitoring, incident response, regression testing, fallback, and retirement procedures.
5Stop and escalate when authority, consent, facts, source currency, policy, jurisdiction, qualified review, or the ability to contest a consequential decision is unclear.

AI risk-control record

Create a reviewable AI risk-control record for this workflow. Use five sections: Govern, Map, Measure, Manage, and High-stakes escalation. Include authorized purpose, prohibited secondary uses, accountable humans, applicable policy or explicit unknowns, data/source/authorization/sensitivity/retention, affected people, action permissions, normal/boundary/adversarial/missing-data/prohibited-action tests, predeclared thresholds, human-review criteria, layered controls, monitoring, incident response, change triggers, retirement, stop authority, and qualified review. Separate observed evidence from assumptions. Mark unknown authority or permission as a blocker; do not invent compliance conclusions.

Workflow: [paste]
Known policies and permissions: [paste or say unknown]

What you leave with

Authorized-purpose and accountability record

Data, people, and permission map

Test, control, incident, and escalation plan

Step 4 of 7 · 12 min

Create a measurable result record before choosing a solution.

TDAR starts with the result, not the tool

An AI initiative can produce impressive activity and no defensible return. TDAR makes the result, comparison point, division of work, and remeasurement decision explicit before implementation creates sunk cost.

Do this

1Result: name the business or service outcome as a number, rate, time, quality measure, or observable state — not as 'use AI.'
2Baseline: record the current value or explicit unknown, source, period, sample, unit, exclusions, and at least one quality guardrail.
3AI contribution: state the bounded part of the workflow the system may perform or assist, including authorized inputs and prohibited actions.

TDAR field test

Field	Review question	Reject when
Result	What changes for the work or customer?	It only measures tool use or output volume.
Baseline	Compared with what, from which source and period?	The number is guessed or the conditions are not comparable.
AI contribution	What bounded work does the system perform?	The claim is that AI improves everything.
Human contribution	Which judgment and accountability stay human?	Review is vague or occurs only after harm.
Remeasurement	When, on what sample, using what rule?	Success will be declared after seeing the result.

Step 4 of 7 · 12 min

Create a measurable result record before choosing a solution.

Separate AI contribution from human accountability

Do this

4Human contribution: state the judgment, verification, approval, relationship, accountability, and exception handling that remain human.
5Remeasurement: set the date, sample, owner, comparison method, and continue, revise, and stop rules before the experiment begins.

Remember: Do not invent a historical baseline. When evidence is unavailable, write 'baseline unknown' and define the collection window required before making a result claim.

Step 4 of 7 · Run it

Create the TDAR Results Brief

Do this

1Result: name the business or service outcome as a number, rate, time, quality measure, or observable state — not as 'use AI.'
2Baseline: record the current value or explicit unknown, source, period, sample, unit, exclusions, and at least one quality guardrail.
3AI contribution: state the bounded part of the workflow the system may perform or assist, including authorized inputs and prohibited actions.
4Human contribution: state the judgment, verification, approval, relationship, accountability, and exception handling that remain human.
5Remeasurement: set the date, sample, owner, comparison method, and continue, revise, and stop rules before the experiment begins.

TDAR Results Brief

# TDAR Results Brief

## Workflow and decision
[What recurring work is being considered, and what decision will this experiment inform?]

## Measurable result
[Measure, target, unit, population, and time period]

## Baseline
[Current value or 'unknown'; source; period; sample; exclusions; guardrail]

## Bounded AI contribution
[Authorized inputs, work performed, outputs, and prohibited actions]

## Accountable human contribution
[Judgments, reviews, approvals, and owner of consequences]

## Remeasurement
[Date, sample, comparison method, continue/revise/stop rules]

## Evidence still missing
[Facts that must be collected or verified]

## Reflection
[Which assumption is most likely to change the decision?]

What you leave with

Five-field TDAR Results Brief

Baseline and quality guardrail

Dated continue, revise, and stop rules

The Decision Making Process

DMP turns possibility into a decision.

The Big What

Name the exact decision, boundary, and timing before debating the answer.

Why

Result and evidence

What evidence, urgency, and assumptions make this worth attention?

What

Bounded workflow

Which users, inputs, outputs, boundaries, non-goals, and finish line are in scope?

How

Credible operating model

Which people, AI, data, checks, permissions, effort, and review make it work?

What If

Decision-changing uncertainty

Which signals, alternatives, owners, and stop rules could change the choice?

Watch the complete Decision Making Process

Step 5 of 7 · 12 min

Expose the choice, evidence, scope, operating model, uncertainty, and alternatives.

Bound the What and make the How credible

A roadmap item can hide five different disagreements: the actual decision, purpose, scope, execution, and risk. The DMP turns those disagreements into reviewable fields instead of compressing them into a yes or no vote.

Do this

1Big What: write the consequential decision that must be made now — build, run a bounded test, defer, reject, or choose another path.
2Why: state the TDAR result, baseline evidence, urgency, observed facts, assumptions, and what happens if no change is made.
3What: define the user, trigger, workflow, authorized inputs, output, downstream consumer, boundaries, non-goals, and finish line.
4How: describe the smallest credible operating model, including people, AI surface, data, deterministic checks, permissions, full effort, review, exceptions, and evidence capture.

DMP review questions

Part	Question	Common failure
Big What	What decision must be made?	A topic or aspiration replaces the choice.
Why	What result and evidence justify attention?	Urgency or enthusiasm replaces evidence.
What	What exactly is in and out?	The proposal expands whenever challenged.
How	What is the smallest credible operating model?	The design ignores permissions, people, or review work.
What If	What could fail or change the choice?	Risk is reduced to a generic disclaimer.

Step 5 of 7 · 12 min

Expose the choice, evidence, scope, operating model, uncertainty, and alternatives.

What If turns uncertainty into decision rules

Do this

5What If: identify decision-changing uncertainties, early signals, continue, revise, pause, and stop rules, accountable owners, and credible current-process or non-AI alternatives.

Remember: The DMP is a Longhand-developed decision record. Its value comes from transparent evidence, explicit alternatives, and a decision another reviewer can challenge — not from making the current recommendation look inevitable.

Step 5 of 7 · Run it

Create the DMP Decision Record

Do this

1Big What: write the consequential decision that must be made now — build, run a bounded test, defer, reject, or choose another path.
2Why: state the TDAR result, baseline evidence, urgency, observed facts, assumptions, and what happens if no change is made.
3What: define the user, trigger, workflow, authorized inputs, output, downstream consumer, boundaries, non-goals, and finish line.
4How: describe the smallest credible operating model, including people, AI surface, data, deterministic checks, permissions, full effort, review, exceptions, and evidence capture.
5What If: identify decision-changing uncertainties, early signals, continue, revise, pause, and stop rules, accountable owners, and credible current-process or non-AI alternatives.

DMP Decision Record

# DMP Decision Record

## Big What
[Exact decision, boundary, and timing]

## Why
[TDAR result, baseline evidence, urgency, observed facts, assumptions]

## What
[User, trigger, workflow, inputs, outputs, finish line, non-goals]

## How
[People, AI surface, data, deterministic checks, permissions, review, effort, failure handling]

## What If
[Decision-changing uncertainties, signals, continue/revise/pause/stop rules, owners]

## Alternatives compared
[Current process, non-AI option, lower-complexity AI option, proposed option]

## Decision and rationale
[Build, bounded test, defer, reject, or alternative]

## Evidence that would reverse this decision
[Specific evidence]

## Reconsideration trigger
[Date or event]

What you leave with

Five-part DMP decision record

Alternative comparison

Decision-reversing evidence and reconsideration trigger

10 / 80 / 10

Human judgement belongs before, during, and after the run.

10%

Human-in

Set the result, context, approved data, permissions, and decision gates.

80%

AI heavy lifting

Research, extract, compare, draft, process, and prepare the reviewable result.

10%

Human-after

Check the Definition of Done, resolve exceptions, approve, and feed learning back into the Skill.

Put approval at the point where judgement occurs. One sign-off after a long run may reveal that something is wrong without revealing where it went wrong.

Step 6 of 7 · 9 min

Separate interpretation, exact execution, and accountable decisions.

Final review is a boundary, not a magic eraser

AI is probabilistic. Some work benefits from interpretation; other work must produce the same exact result, account for every record, and be safe to rerun. Strong workflows use both deliberately.

Do this

4Use deterministic software when the same inputs must produce the same output and every record must be accounted for.
5Let AI suggest an ambiguous mapping; let a person approve it; let deterministic software execute it.
6Put review where judgement occurs, with stop rules before an error can spread.

Remember: A final review is an approval boundary, not a substitute for schema checks, transactions, idempotency, reconciliation, or other deterministic controls.

Three different jobs in one safe system

AI can help	Deterministic systems own	Humans decide
Interpret messy headers	Validate schema and types	Approve ambiguous mappings
Suggest classifications	Deduplicate with explicit rules	Resolve exceptions
Explain rejected rows	Import transactionally	Authorize the real run
Draft remediation	Reconcile exact counts	Accept business risk

Worked boundary · data import

Import data on deterministic rails.

AI can help interpret ambiguity. It should not silently own the transaction, identity rules, or proof that every record landed where it belongs.

1Deterministic
Validate
Schema, types, required fields, and exact row count.
2AI
Suggest
Propose mappings only for ambiguous fields, with evidence.
3Human
Approve
Accept, correct, or reject ambiguous mappings and exceptions.
4Deterministic
Dry run
Apply explicit rules without changing production data.
5Deterministic
Import
Use an idempotent transaction that is safe to retry.
6Deterministic
Reconcile
Account for every imported, rejected, and quarantined row.

input rows = imported + rejected + quarantined

Step 7 of 7 · Run it

Make one real workflow reviewable

Do this

1Complete TDAR: measurable result, defensible baseline, bounded AI contribution, accountable human contribution, and a dated remeasurement rule.
2Complete the DMP: Big What, Why, What, How, What If, alternatives, and the evidence that would reverse the decision.
3Record the responsible-AI purpose, data and permission boundary, tests, human review, stop authority, incident response, change rule, and qualified escalation.
4Choose the gear and reusable Skill, mark the AI, deterministic, and human lanes, then add the Definition of Done and accountable owner.

AI Work Decision Card

AI WORK DECISION CARD

WORKFLOW
Name the real job or recurring outcome:

TDAR — RESULT FIRST
Business result:
Current baseline, source, period, sample, and guardrail:
AI contribution:
Human contribution:
Re-measure date, owner, and decision rule:

DMP
Big What — the exact decision, boundary, and timing:
Why — result, evidence, urgency, and assumptions:
What — workflow, users, inputs, outputs, boundaries, and finish line:
How — people, AI, data, deterministic checks, effort, and permissions:
What If — uncertainties, alternatives, signals, and stop rules:

RESPONSIBLE AI CONTROLS
Authorized purpose and prohibited uses:
Data, affected people, permissions, and retention:
Tests, thresholds, and evidence retained:
Human review and stop authority:
Incident, change, retirement, and qualified-escalation rules:

RIGHT GEAR
Chat / Assistant / Agents / Loops:
Why this gear fits:

REUSABLE SKILL
Should this become a Skill? Yes / No
Instructions, approved sources, tools, examples, and quality rules it needs:

WORK BOUNDARIES
AI may:
Deterministic software must:
A person must decide or approve:

FINISH LINE
Definition of Done:
Stop rules and exceptions:
Accountable owner:

The overview is complete

You know what to build — and what not to.

Keep the card with the work. Use the audit to identify the next capability gap, then choose the training that helps you shift up without hiding the judgement or control the result requires.

Assess

Take the AI Tune-Up

Rate the evidence you have today and open the exact training for each red or yellow gap.

Open training

Set up

AI Agent Starter

Choose a home base, set the data boundary, connect work, and save a reusable workflow.

Open training

Build

Build your first Skill

Turn the chosen job, workflow, toolbox, quality criteria, and stop rules into a reusable Skill.

Open training

Longhand AIOverview

One AI Work Decision Card

Longhand AIOverview

One AI Work Decision Card

Turn AI possibility into one accountable work decision.

Human judgement drives every gear.

Chat

Assistant

Agents

Loops

Chat

Assistant

Agents

Loops

Stop building one giant agent.

Skills are the apps of the AI era.

Package the procedure a colleague could reuse

Responsible AI is an operating cycle

Tests, controls, and stop rules must operate

Create the risk-control record

TDAR starts with the result, not the tool

Separate AI contribution from human accountability

Create the TDAR Results Brief

DMP turns possibility into a decision.

Result and evidence

Bounded workflow

Credible operating model

Decision-changing uncertainty

Bound the What and make the How credible

What If turns uncertainty into decision rules

Create the DMP Decision Record

Human judgement belongs before, during, and after the run.

Human-in

AI heavy lifting

Human-after

Final review is a boundary, not a magic eraser

Import data on deterministic rails.

Validate

Suggest

Approve

Dry run

Import

Reconcile

Make one real workflow reviewable

You know what to build — and what not to.

Take the AI Tune-Up

AI Agent Starter

Build your first Skill