The Agent Design Canvas
A design method for executives & small businesses

The Agent
Design Canvas

Eight principles and one working canvas for designing an AI agent — from a single job to a system that scales. Built to be read in one sweep, before anyone opens a platform.

Synthesis of 7 source guides Read order Principles → Canvas → Anatomy → Autonomy → Patterns → Guardrails → Sequence Rule Design one agent before designing a system
The map

One agent, at a glance

Everything in this guide, arranged so the layout itself shows the relationships. The core sits at the centre; each ring outward is a layer that contains and governs the one inside it, and the work runs as a loop around the middle. Read outward along a spoke to see how a capability is powered, bounded and supported. Read around to follow the cycle. The step-by-step version of the same design lives in Section 02.

Centre · Core

Identity + Job. Who is acting and the result it owns — everything attaches here.

Ring 2 · Capabilities

The components that power the loop: model, prompt, tools, memory, orchestration.

Ring 3 · Bounded autonomy

The boundary that contains the agent: permissions, gates, fallback, kill switch.

Ring 4 · System & assurance

The world it runs in: shared support, audit, live checks, the model service.

AGENT CORE Identity + Job who is acting ·the result it owns SENSE Memoryrecall context Permissionsread scope Approved Data& Knowledge REASON Model + Promptinterpret · plan Rules & Stopconditions Model Service& versioning ACT Toolscall systems Approval Gateson irreversible Records &Audit trail OBSERVE Orchestrationdecide next step Fallback &Kill switch Live Checks& quality INTERFACE input port INTERFACE output port Trigger / request in action on the world — signed with agent identity
System authority

Sets the rules, approvals and change rights the whole circle runs under.

Identity & access

Every agent its own verified identity · least privilege · zero-trust. The dashed spokes show it threading from the core to permissions and audit.

Earned autonomy

The core climbs outward over time: read → draft → supervised → extended.

Measure & prove

result · cost · speed · risk · demand. Change only through recorded evidence.

Core & work loop Capabilities Bounded autonomy System & assurance Identity thread

Hover a quadrant to trace one spoke — the loop phase, the capability it uses, the control that bounds it, and the system service that feeds it all line up radially. Green arrows show the loop turning clockwise; the trigger enters at the top gate, the action exits at the bottom, signed with the agent’s identity.

The circle shows structure — what contains what. The five moves in Section 02 show sequence — what you decide first, next and last. Same design, two readings.

01

Eight design principles

These hold whether you are a solo operator or an enterprise. Roughly three-quarters of multi-agent failures trace back to design and specification, not to the model — so the discipline lives here, before the build.

P1

Define the job first

Name one business result, one trigger, one output, one handoff. If the agent’s title sounds like a real job, the scope is clear enough to build. Precision in the brief is precision in behaviour.

Scope
P2

Bound autonomy on purpose

Freedom to act, inside clear limits. Permissions, approval gates, fallbacks and a kill switch turn a capable agent into a safe one. Separate what it may recommend from what it may change.

Control
P3

Earn autonomy, don’t assume it

Start read-only. Move to draft-only, then supervised action, then wider authority — each step unlocked by a track record, not a good demo. Trust is a rate you measure, not a launch setting.

Trust
P4

Size the blast radius

Before granting a permission, ask what the worst mistake costs and whether it reverses in ten minutes. Irreversible, high-impact actions need a human in the loop — often permanently.

Risk
P5

Prefer the simplest architecture

One disciplined agent beats a fragile crowd. Add a second agent only when one job genuinely splits into two. More agents add coordination failure, not automatic capability.

Structure
P6

Store only what earns its place

For every memory, name the next task it improves. If you can’t, don’t store it. Memory is an operational tool with governance — what’s kept, who updates it, how long it lasts.

Memory
P7

Close every loop to a real system

An output that sits in a chat window creates work; an output that updates a record removes it. When the agent finishes, something must change in a system people already use.

Interface
P8

Prove it, then improve it

Set a baseline before building. Log every action, measure result, cost, speed and risk, and change behaviour only through recorded evidence and approved release. Feedback alone changes nothing.

Assurance

“If this agent takes the wrong action here, what happens and who fixes it?” — the one scoping question that prevents most incidents.

02

The canvas: five moves to design one agent

Read left to right. Each column is one decision with a clear verb and the components it settles. The last column is where a single agent grows into a system — cross it only after the first four are working.

One decision per box · action titles · plain business words · every step ties to a record, a measure or a check.

Move 1 · Define

Set the job & limits

Name the result. Draw the line.
  • Business result & the case it handles
  • Start (trigger) and finish (done)
  • Success metric — one primary
  • The Never List: what it must not do
Sets · System Prompt · Goal
Move 2 · Permit

Give it an identity & limits

Who is acting · what it may do.
  • Own identity — credentials never shared with a human
  • Four permission levels, least first
  • Approval gates on irreversible steps
  • Fallback & kill switch — who can throw it
Sets · Identity · Bounded autonomy
Move 3 · Equip & run

Give what it needs, do the work

Sense → reason → act → observe.
  • Data, rules, tools, people, other agents
  • Source recorded for each dependency
  • Reasoning pattern for the task
  • Memory: only what the next case needs
Sets · Model · Tools · Memory · Loop
Move 4 · Prove

Check the result

Measure. Record. Escalate.
  • Verify output against defined criteria
  • Log action + evidence (audit trail)
  • Trust rate: approved without edits
  • Faults & ideas → system authority
Sets · Orchestration · Assurance
Move 5 · Scale

Group into a system

Only when one job becomes many.
  • Group agents around one result
  • Route & sort incoming work by risk
  • Shared support: identity, data, audit
  • Check the whole: result, cost, demand
Sets · System rules · Multi-agent
The work loop inside Move 3 repeats — read the case, make sense of the evidence, choose the next allowed step, act, observe — until a stopping rule fires: done, unsafe, or unclear.
03

What’s inside: the seven components

Every agent is built from the same parts, like a car — the engine matters, but bad tires or an empty tank stop you just the same. When an agent underperforms, find the broken component before reaching for a more expensive model. Binding all of them to the wider system is one shared layer, and its first pillar is identity.

Component 1

Model

Reads input, reasons, generates output. Use the right model per step, not the best model everywhere.

Breaks asFluent but wrong, vague, or off-task
Component 2

System prompt

Role, context, rules, format, examples, stop condition. Two to five examples lift accuracy 15–40%.

Breaks asSame input, different output; scope drift
Component 3

Tools

Read from and act on real systems via APIs. Start with 3–5, read-only first, add write tools one at a time.

Breaks asConfident claims it can’t verify
Component 4

Memory

Carries context across steps and sessions. Four kinds: task context, stable preferences, history, corrections.

Breaks asRepeats questions; no continuity
Component 5

Decision loop

Drives multi-step work forward and knows when to stop. The difference between a prompt and an agent.

Breaks asAnswers once and stalls
Component 6

Orchestration

Coordinates the parts into a reliable workflow: what runs first, next, last, and what makes it pause or retry.

Breaks asWorks in test, misbehaves in production
Component 7

Interfaces

Connect inputs and outputs to real work. The exit point must change something in a system people use.

Breaks asGreat output that lands nowhere
Wraps all seven

Shared support

Identity & access first — a verified identity per agent, then approved data, safe connections, records & audit, live checks, a versioned model service.

Without itNo system works at scale
ID

Identity is the thread that ties it together

Humans log in with a username and password; an agent doesn’t fit that model, yet it may read email, update the CRM and act across systems on someone’s behalf. So the first question a system must answer is simple: which agent is acting, and what is it allowed to do? Give every agent its own verified identity — never a human’s shared credentials — and that one decision makes the rest of the canvas work.

Verified identity gates permissions (Move 2) connects to tools & systems (zero-trust, least privilege) tags every action (audit & attribution) proves the result (Move 4)
04

The earned-autonomy ladder

Autonomy is a dial, not a switch. Every agent starts at the bottom and climbs only by demonstrated performance. The gate on the right is what must be true before the next rung.

R
Read-onlyMinimal risk
Views records, retrieves data, reads documents. Changes nothing in any system.
→ 2–4 weeks of accurate output, low false positives
D
Draft-onlyLow risk
Creates emails, summaries, records — but a human approves before anything is sent or saved.
→ 30+ days at 85%+ approval, minimal edits
S
Supervised actionMedium risk
Takes actions within defined rules; humans review the higher-risk cases only.
→ Sustained reliability on the routine path
A
Extended authorityGoverned
Acts across a wider scope, still inside permissions, audit and kill-switch controls.
→ Reversibility & blast radius stay acceptable
05

Pick behaviour from named patterns

Inside Move 3, choose proven patterns rather than inventing behaviour. Reach for the reasoning pattern the task shape calls for, add orchestration only when work truly splits, and wrap the whole in safety patterns.

Reason & plan
  • Plan-then-actStrict sequences, compliance, contract review
  • Chain-of-thoughtMulti-step logic and root-cause work
  • ReActLook things up instead of guessing
  • ReflectionCritique and refine its own output
  • Tree-of-thoughtsExplore options when search pays off
Orchestrate
  • RouterSort mixed requests to specialists
  • Supervisor–workerManager breaks a goal into subtasks
  • Parallel fan-outMany views at once, then merge
  • Prompt chainOutput of one step feeds the next
  • A2A meshIndependent agents negotiate directly
Guard & verify
  • Human-in-the-loop gatePause for sign-off on high-consequence acts
  • Act-then-verifyA checker pass before continuing
  • Self-check / LLM-as-judgeInspect output for errors and drift
  • Input sanitisationStrict schemas; block injection
  • Programmable railsTopic, dialogue and execution limits
06

Four mechanisms of bounded autonomy

Instructions are not enough — a strong prompt is only part of the answer. The real controls live in the architecture: what the agent can access, what it can do, and where a human must approve.

Permissions

What the agent can access and change. Grant the minimum; expand only after it proves reliable.

Approval gates

Points where a human must authorise before the agent continues — for anything irreversible, sensitive or first-time.

Fallbacks

What happens when it can’t continue: route to a human, return a flagged partial, or stop and log. Never push forward broken.

Kill switch

Stop it immediately — kill the trigger or pull write access. Every production agent has one, and the team knows how to use it.

The blast radius of an ungoverned agent is not bounded by its intelligence. It is bounded by what it has been permitted to access and act on.

07

The order of work

Design the agent fully before choosing how to build it. Finish the thinking, then pick the simplest tool — no-code, low-code, or full code — that fits.

1

Classify the task

What triggers it, what “done” means, what it must never do.

2

Set the limits

Permissions, approval gates, fallbacks, kill switch.

3

Choose the structure

One agent by default; a group only if the job splits.

4

Select patterns

Reasoning, orchestration and safety primitives.

5

Place the human

Decide which functions to automate and how far.

6

Add guardrails & prove

Baseline, audit trail, failure-mode review, then release.