PROMPT → WORKFLOW → AGENT
HOW TO DESIGN AN AI COLLEAGUE THAT CAN ACT
The B.L.U.E.P.R.I.N.T. Intelligent Agent Framework v2.4
© 2024–2026 Ahmed Muzammil · CC BY-SA 4.0

START HERE
Before an AI can act for you, someone has to decide where it stops.
A prompt gets you an answer. An agent goes and does things in your systems, so you have to decide what it's allowed to touch.
This deck walks through those decisions one letter at a time, using two example agents: one in a bank and one in a clinic.
WHY THIS EXISTS
A newer model won't fix an agent that was never told what it's for.
What people usually assume
What we see in practice
DO YOU NEED AN AGENT?
1 · Is this a regular job for someone, not a once-in-a-while ask?
If no, a good prompt is enough: you give it what it needs, check the answer and decide what happens next.
Use PINPOINT to write that request well.2 · Would they need to get into your systems or tools themselves, not just tell you what to do?
If no, set up an assistant with one clear job. It answers, and a person acts on what it says.
Use BLUE: the first four letters.3 · Would a checklist fail because some cases are different, so they’d have to decide the steps as they go?
If no, it's a checklist: write the steps down and use ordinary software or a workflow. AI can still do one step inside it.
Three yeses means you're building an agent, and BLUEPRINT is how you describe it. Stop at the first no and use what it points to.
Anthropic, Building Effective Agents (19 December 2024): find the simplest solution possible and only increase complexity when needed, which may mean not building an agentic system at all.
A workflow is a job where you already know the steps. An agent is a job where something has to work out the steps each time. Most work that looks like it needs an agent is a workflow with one or two decisions inside it. A workflow with an AI step that gets into your systems still needs its limits written down: use BLUEPRINT for that step.
Wherever you stop, write down why in one sentence. It's the first question a reviewer will ask.
OUR TWO EXAMPLES
Two agents we'll build together.
In a bank: Priya's meeting-briefing agent
Priya is a relationship manager. Before each client meeting she pulls notes together from several systems.
Her agent will do that for her: read what's relevant and hand her a one-page brief the evening before.
In a clinic: Dana's front-desk agent
Dana runs the front desk at a busy family clinic. Every morning she checks who's booked in, which forms are missing and who hasn't confirmed.
Her agent will prepare that list and draft reminder texts for her to check.
Each letter adds one decision to both agents. By the end you'll have a full description of each, and a template for your own.
FROM AN IDEA TO A DESCRIPTION · PRIYA
Where most ideas start, and where they need to get to.
HOW THE IDEA USUALLY ARRIVES
This covers tone and one thing to avoid. It says nothing about what a finished brief looks like, what the agent can open, or who answers for it.
THE JOB
WHAT IT DOES ALONE, AND WHAT IT ASKS ABOUT
WHAT IT CAN OPEN
FROM AN IDEA TO A DESCRIPTION · DANA
The same step for Dana at the front desk.
HOW THE IDEA USUALLY ARRIVES
Friendly tone and one thing to avoid again. It doesn't say when the list is due, who may send a text to a patient, or what it can see in the booking system.
THE JOB
WHAT IT DOES ALONE, AND WHAT IT ASKS ABOUT
WHAT IT CAN OPEN
WHAT AN AGENT IS
An agent is AI with a goal and permission to act. It picks a step, does it, checks what happened, and keeps going until the job is done or it has to stop.
It still uses prompts. The difference is that it keeps going by itself.
A workflow follows steps someone wrote down in advance. An agent works out its own steps as it goes (after Anthropic, Building Effective Agents, 19 December 2024).
Simon Willison, 18 September 2025: “An LLM agent runs tools in a loop to achieve a goal.” In other words, an AI that keeps using your systems, step by step, until it gets there.
PROMPT OR AGENT?
With a prompt, you decide each step. With an agent, it decides.
That last row is why an agent needs its job and its limits written down.
WHERE BLUEPRINT SITS
Most real systems are a mix. If only one step needs the AI to decide, describe it as a workflow with one smart step.
THE FIRST FOUR LETTERS
BLUE
Start with the least structure that works.
Behaviour & Values
What should it put first, and what does it do when unsure?
Limits
What can it do alone, what must it ask about, and what is off-limits?
Use Case & Outcome
What is the one job, who is it for, and what does finished look like?
Examples
What does a good result look like, and have you tried the awkward cases?
BLUE is enough for an assistant that only answers questions. Once it can open systems or do things, add PRINT.
BLUE · B
B Behaviour & Values
Give it a ranked list of what matters.
Why it mattersRules cover the cases you thought of. Something will come up that you didn't, and the agent needs to know what matters most when it does.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downYour agent's top three priorities, in order. When two clash, which one wins?
BLUE · L
L Limits & Autonomy
Sort everything it might do into three lists.
Why it mattersAn agent will try whatever seems helpful. Without clear lists, it has to guess where your line is. Reading counts too: opening a file is an action.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downThree lists: just do it, ask first, off-limits. If something isn't on any list, it stops and asks.
BLUE · U
U Use Case & Outcome
One job, one person who answers for it, and a clear picture of finished.
Why it mattersAn agent with a vague job drifts into other work. When something goes wrong, nobody is sure whose call it was.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downThe job in one sentence, the named person who answers for it (a person, so no team names), and what finished looks like.
BLUE · E
E Examples & Evaluation
Collect examples of good and bad, then test it on the awkward ones.
Why it mattersAn agent can reach a good answer the wrong way, for example by opening something it shouldn't. It can also answer the same question differently on different days.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downFive test cases, including one it should refuse. Agree the pass mark before you run them, and run each one more than once.
THE NEXT FIVE LETTERS
Add these once the agent can open systems and do things.
Permissions & Identity
What can it open?
Reliability & Recovery
What if something breaks?
Interfaces & Handoffs
Which systems does it use, and when does it hand over to a person?
Navigation & Orchestration
What are the steps it follows, and when does it stop and/or escalate?
Trusted Context & Memory
What does it read and keep?
BLUE describes the job. PRINT describes how it gets done safely inside your systems.
PRINT · P
P Permissions & Identity
Telling it isn't the same as stopping it.
Why it mattersIf we only tell the agent “don't look at other advisers' clients”, it can still try. A block in the system makes it impossible.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downWhose access does it borrow, and what is it blocked from?
PRINT · R
R Reliability & Recovery
Decide what happens when something breaks.
Why it mattersSystems go down, data goes out of date and AI models get updated. An agent that quietly guesses its way past a problem does more harm than one that stops.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downWhen it gets stuck: what it does, who hears about it, and how you undo anything it already did. Keep a full record of every run, like a flight recorder.
PRINT · I
I Interfaces & Handoffs
Before it uses a tool, write down what the tool does and what it can break.
Why it mattersThe agent only knows a tool by its description. If “no record found” and “access blocked” look the same, it may tell Priya a client has no history when it was locked out.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downFor each tool: what it's for, what it changes, and what it says when it fails.
PRINT · N
N Navigation & Orchestration
Fix the steps you already know. Let the agent decide only where it has to.
Why it mattersThe more steps the agent chooses, the harder it is to predict. Most jobs are a fixed path with one or two judgement calls inside.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downThe steps that are always the same, the one or two it decides, and when it should stop.
T · TRUSTED CONTEXT & MEMORY

A messy filing cabinet makes a messy AI.
Imagine the smartest analyst in the world, with a bad filing cabinet. The AI may be brilliant. Its information may not be.
Context engineering
What goes in: duplicated, stale, over-shared and unlabelled files. What comes out: confident, wrong answers.
T · TRUSTED CONTEXT & MEMORY
Check what it reads, and who it reads it for.
Every document and record shows three things
This goes for data as well as documents. Every document or record the agent uses should show who owns it, who's allowed to see it, and whether it's the effective version on that date.
“Effective” means it's the version people are meant to follow today. It isn't an old one that's been replaced, and it isn't a draft that hasn't been approved yet. If there are three travel policies and none is marked effective, the agent will pick one and sound sure about it.
Filter by the person before the agent sees anything
Data has access rules too. A customer's details should only be visible to the relationship manager who looks after that customer. If you already have a system that controls this, such as a customer entitlement system, use it to filter the data for the person using the agent before it reaches the agent.
The same goes for documents. Otherwise a paper meant only for the management committee or the CEO can end up in an answer to an employee who was never meant to see it, or the agent may mention that the paper exists at all.
PRINT · T
T Trusted Context & Memory
Decide what it gets to see before it answers, what it trusts most and what it keeps.
Why it mattersAgents read a lot. Some of it is out of date, some of it contradicts itself, and some of it might be an email saying “ignore your rules”.
Priya's briefing agent (bank)
Dana's front-desk agent (clinic)
Write downIts sources in order of trust, how old is too old, and what it may remember.
MY VIEW ON "HUMAN IN THE LOOP"
A person in the loop only helps if they can say no.
1983: he broke the loop
A rubber stamp isn't a safeguard
People get tired
What works better
Better questions to ask: Where does a person's judgement change the outcome? What happens when they say no? Do they have the time and context to challenge it? Who answers for it when it's wrong? Can you switch it off?
WHY "ASK FIRST" NEEDS CARE
People stop reading approval requests.
People also got worse with practice. After about 50 requests, they caught around 5% of the dangerous ones, down from about 17%.
What this means for your agent: ask people to approve the whole plan, and let the system block anything that should be impossible. These numbers come from AI coding sessions, and we're assuming they carry over to other kinds of approval. Source: claude.com/blog/auto-mode-default-in-claude-code · 7 August 2026 · 1,053 testers.
CHECK YOUR DESCRIPTION
Three questions before you hand it over.
The agent doesn't mark its own homework. A person or a fixed rule does.
The full nine-letter check is in the “For the builders” section at the back.
WHY AGENTS GO WRONG OVER TIME
Too tight, too loose, or slowly drifting.
Overfitting: too tight
The agent learned its examples and past preferences so closely that it stumbles on anything new.
Like a new joiner who memorised last year's answers but never understood the questions.
Causes: too few or too similar examples, and memory that treats one person's habit as a rule. Priya once asked for short summaries; now every client gets three lines, even on a complex loan.
Underfitting: too loose
The rules and examples are too thin or vague, so the agent gives safe, generic answers.
Like a temp told only to “help visitors”.
Causes: vague instructions, missing examples and no clear “done”. Dana's agent greets everyone the same way and can't handle a delivery or a VIP.
Drift: slowly changing
Behaviour that was fine at launch shifts a little at a time.
Causes: AI model updates, memory piling up, changed policies or data, and small edits to the instructions that nobody tested again.
None of these announce themselves. That's why the next page is about testing again whenever anything changes.
KEEPING IT UP TO DATE
When anything changes, test it again.
When the instructions, the tools or the AI model change, treat it as a new version and test it again before it goes live. Corrections from people are evidence for the next review. The agent doesn't rewrite its own rules.
IS IT CLEAR ENOUGH?
Could someone from risk or audit read it and understand:
Then try one real run. Pick a single piece of work the agent did. Can you show what it looked at, what it was allowed to do, which checks ran, what the person saw and changed, who decided, and what went out?
THE FULL JOURNEY
Define the task
Describe the AI colleague
BLUEPRINT is the job description.
EVOLVE
TAKEAWAY
Every agent that works well had someone decide its limits first.
Start with BLUE. Add the PRINT letters once it can open systems and do things. The template is at the back.
FOR THE BUILDERS · GLOSSARY
The words you'll hear, in plain English.
FOR THE BUILDERS · GLOSSARY
More terms.
FOR THE BUILDERS · TEMPLATE
Fill in the BLUEPRINT.
Three checks before you scale it. Know: can you trust what it reads? Do: can you control what it does? Prove: could you rebuild what happened six months from now?
CHEAT SHEET · THE NINE LETTERS IN PLAIN WORDS
Behaviour & Values
Write down what matters most, in order, and which one wins when two clash. It only says a job is done once the system confirms it.
Limits & Autonomy
Reading isn't acting. Place each action on the ladder: suggest, draft, act with a named person's approval, or act alone. If something isn't on the list, it stops and asks.
Use Case & Outcome
One job, for one person, with one named person who answers for it. Note what starts it, who it affects and how the work is done today, so you can tell whether it helped.
Examples & Evaluation
Test a normal day, the awkward cases, someone trying to trick it (for example, hidden instructions in a document), things breaking, and a request it must refuse. Agree the pass mark first. The safety checks must pass every time.
Permissions & Identity
The agent only sees what the person it's working for can see (see Trusted Context, T). It has its own login, never a shared one, so you can always tell what it did.
Reliability & Recovery
If something breaks, it stops safely and there is a plan to undo or put right anything it changed. Keep a simple receipt for every answer, allowed or refused: where it came from, which version, who checked it. Keep records nobody can edit.
Interfaces & Handoffs
For each tool, write down what it's for and what it can change. “Nothing found”, “it broke” and “not allowed” must look different to the agent.
Navigation & Orchestration
Fix the steps you already know. It stops when the job is done, the budget runs out, it's stuck, it's blocked or it's outside its limits, and says where it got to.
Trusted Context & Memory
Give it the right documents and data, and only the version in force on that date. Each one shows who owns it and who may see it. Anything the person can't see is filtered out before it reaches the agent (pages 21 and 22).
FOR THE BUILDERS · CROSS-LETTER CHECKS
The full consistency check.
The agent never sets its own permissions, marks its own work, or decides whether its own safety checks run.
FOR THE BUILDERS · TOOLS AND FLOW
Tool contracts and how the work moves.
- Tool names and descriptions work like prompts. Write them like a brief for a new hire.
- A handoff between agents or people is a PINPOINT brief: purpose, context with evidence, a narrow ask.
- BLUEPRINT doesn't depend on a protocol. MCP and APIs are implementation choices.
- Move right along the path only when the job needs it.
FOR THE BUILDERS · SEVERAL AGENTS
Adding agents adds no authority.
One BLUEPRINT per agent, and one agreed contract for each handoff.
MAKER
CHECKER
- An agent receiving work gets only the permissions granted to its own role.
- The strongest reason for a second agent in banking is maker-checker: one prepares, a separate one verifies.
FOR THE BUILDERS · REPLAY ONE RUN
Can you replay one run?
A supervisor picks one AI-assisted client interaction. Every answer should already be in the record, and each one traces back to a part of the BLUEPRINT. European regulators will look closely at AI in investor-facing work from 2027 (ESMA has made it a Union Strategic Supervisory Priority).
FOR THE BUILDERS · SHARED WITH PINPOINT
Evidence labels
Labels for “I checked this” and “I'm guessing”. In PINPOINT you ask for them in a single request. In BLUEPRINT the agent uses them all the time: Behaviour & Values (B) requires them, and Trusted Context (T) enforces them.
FOR THE BUILDERS · READ
Daily meeting-briefing agent
Low autonomy · read-only · relationship-led financial services
Job
Trigger
Flow
Always do
Ask first
Off-limits
When it fails
Why it matters
FOR THE BUILDERS · PROPOSE
Client follow-up agent
Approval before action · relationship-led financial services
Job
Trigger
Flow
Always do
Ask first
Off-limits
When it fails
Why it matters
FOR THE BUILDERS · ACT
Operations exception agent
Bounded autonomy · relationship-led financial services
