Accountability Fragmentation · Secondary
Accountability is assigned. Control is not.
How this principle feeds it
people still hold the outcome on paper but act only within permission, so ownership becomes symbolic.
Principles Library / Intent-grounded rules
Intent-grounded rules
Clear Operating Rules Grounded in Intent Enable Autonomous Decisions
Diagnose
AI signs are on by default. Use the AI switch to exclude them.
1. Which of these early warning signs are you seeing?
- Teams follow the process correctly while the outcome it serves gets worse.
- Metrics get hit while the purpose behind them is undermined.
- Similar situations get different decisions depending on who's involved.
- Routine decisions get escalated because the rules give no intent to judge by.
- Broad goals get agreement, then teams execute them in different directions.
- Leaders keep getting surprised by outcomes that technically followed the rules.
- Intent specifications are written once and never revisited as context changes.
- Each incident adds a new rule, and old rules never get retired.
- Agent outputs look plausible and pass the rules but miss what stakeholders meant. (AI)
- When agents miss, teams add rules or improve the agent instead of fixing the goal. (AI)
- The context agents need to read intent correctly lives only in people's heads. (AI)
- Nobody can say what an agent is authorized to do or why its rules exist. (AI)
2. Over the last few months, have you noticed…
For each consequence: Not seen / Sometimes / Persistently / Not sure.
- People and agents stop owning outcomes and do only what's permitted.
- Metrics and controls get chased as targets instead of read as signals.
- Decisions turn inconsistent and political, or stall in escalation.
- Delivery speeds up under rules that have lost their intent, so value stays flat.
- Rework comes from the wrong goal done well, yet gets treated as execution failure.
- Agents repeat mechanical compliance and misread intent at scale. (AI)
3. The condition it becomes
Based on your answers, recognize at most two conditions: No / Partly / Yes, that is us.
Accountability Fragmentation: Accountability is assigned. Control is not.
Local Optimization Bias: Every team hits its targets. The product still misses.
Governance Drag: Work is done. Approvals are not.
Customer Disconnect: Delivery is consistent. Value is not landing.
Intent Drift: The team built exactly what was asked. It is no longer what is needed.
Attribution Failure: We know what happened. Nobody can explain why.
Quality Fragility: Customers find defects before the team does.
Implementation Drift: The agents are building. Nobody is sure it still matches the plan.
Take your Diagnosis to the next level by invoking AI Diagnostic Coaching tool.
What else have you observed? Optional. Shared only if you choose to start coaching.
Start a Diagnosis session: review and edit a draft containing your note, conditions you recognized in Step 3 (including whether they match partly), warning signs and consequence ratings before sending it. No scores or principle files are sent.
Overview
Principle intent
Translate intent into clear rules, policies, boundaries and decision criteria so people and AI systems act autonomously without losing the rules' purpose.
Read the full intent
Too little clarity produces divergent interpretation and constant escalation. Too much prescription replaces judgment with mechanical compliance. For AI systems, precision has to be established before execution begins because agents can't correct a vague goal through conversation the way people do mid-task.
When it holds: what it does for you
Fewer routine escalations
People make routine calls inside clear boundaries without waiting for approval.
Rules keep their purpose
Teams follow rules and hit metrics while the outcome behind them still holds.
Agents aimed correctly
Agents start with the intent and context they need, so their outputs match what people meant.
When it is ignored: from early signs to a lasting condition
Early warning signs · Visible now, cheap to fix
- Teams follow the process correctly while the outcome it serves gets worse.
- Leaders keep getting surprised by outcomes that technically followed the rules.
- Intent specifications are written once and never revisited as context changes.
What happens over time · Accumulates, costly to reverse
- Metrics and controls get chased as targets instead of read as signals.
- Delivery speeds up under rules that have lost their intent, so value stays flat.
- Rework comes from the wrong goal done well, yet gets treated as execution failure.
The condition it becomes
Local Optimization Bias · Primary
Every team hits its targets. The product still misses.
How this principle feeds it
teams optimize their own metrics while the end-to-end outcome they were meant to signal fragments.
Governance Drag · Primary
Work is done. Approvals are not.
How this principle feeds it
routine decisions wait in escalation, so finished work waits for calls that add delay without reducing risk.
Customer Disconnect · Primary
Delivery is consistent. Value is not landing.
How this principle feeds it
work ships on schedule while the outcome the business needed stays missing.
Intent Drift · Primary
The team built exactly what was asked. It is no longer what is needed.
How this principle feeds it
the system keeps executing a goal that no longer reflects the need, and the output stays internally consistent.
Attribution Failure · Secondary
We know what happened. Nobody can explain why.
How this principle feeds it
the miss is recorded but put down to execution, so the decision that caused it stays untraced.
Quality Fragility · Secondary
Customers find defects before the team does.
How this principle feeds it
wrong-goal rework is found after delivery, when it is most expensive, and grows as a share of output.
Implementation Drift · Secondary
The agents are building. Nobody is sure it still matches the plan.
How this principle feeds it
imprecise context lets agents substitute their own readings, each locally correct, which become the new baseline.
Act
Act
Showing all 8 practices. Answer the Diagnose questions and the most relevant come first.
Make decision boundaries explicit and easy to find
Make decision rules, policies and boundaries explicit, and put them where people can find them.
Start by asking your team: What decisions should people be able to make without asking permission?
Avoid · You’ll know it’s working when
Avoid
- Sending every exception up for escalation instead of designing clear decision boundaries
You’ll know it’s working when
Routine decisions get made locally, and similar situations get similar calls.
Explain the intent behind key rules
Explain the intent behind important rules, standards and metrics. Say what each one is there to protect.
Start by asking your team: What outcome or purpose is this rule or specification meant to protect?
Avoid · You’ll know it’s working when
Avoid
- Encoding policies or context into agents without making the assumptions visible or open to debate
You’ll know it’s working when
Teams hit metrics and follow process while the outcome behind them holds, and anyone can say why an agent's rules exist.
Spell out expected outcomes and trade-offs
Define the outcomes and trade-offs you expect, beyond the activities you require.
Start by asking your team: What would tell us the intent specification was wrong, rather than the execution?
Avoid · You’ll know it’s working when
Avoid
You’ll know it’s working when
Leaders stop getting surprised by outcomes that technically followed the rules.
Test important guidance with different teams
Test important guidance with different teams and check whether they read it the same way.
Start by asking your team: Could two reasonable people, or two agents, read this guidance differently and act differently?
Avoid · You’ll know it’s working when
Avoid
- Treating agreement or consensus as proof of shared understanding
You’ll know it’s working when
Agreed goals turn into the same execution across teams.
Review rules by the outcomes they produce
Review rules and rituals on whether they still produce the intended outcome. Adherence alone doesn't count. Change or retire the ones that don't.
Start by asking your team: Where are we complying with a process or metric while missing its intended purpose?
Avoid · You’ll know it’s working when
Avoid
- Treating compliance with a rule or approved specification as proof the outcome is being achieved
You’ll know it’s working when
Metrics move when the purpose behind them moves, and rules that stop serving it get retired.
Treat agent intent specification as formal work
Treat intent specification for agents as a formal activity. Define what a correct outcome looks like, validate it and build the context agents need into delivery infrastructure.
Start by asking your team: What organizational context does an agent have before it starts executing?
Avoid · You’ll know it’s working when
Avoid
- Treating prompt or specification quality as an individual skill instead of an organizational governance job
- Adding more rules after bad agent behavior without checking whether the intent was ever clear
You’ll know it’s working when
Agent outputs match what stakeholders meant, and the context agents need is written into the system.
Review governing intent at the pace of change
Set a review cadence for governing intent that matches how fast the business changes.
Start by asking your team: When was this governing intent last reviewed?
Avoid · You’ll know it’s working when
Avoid
You’ll know it’s working when
Intent specifications get revisited whenever the context changes.
Use clear intent to enable autonomous judgment
Use policies and clear intent to let people and agents use their own judgment. Don't use them to create escalation or to replace judgment with rules.
Start by asking your team: Are our policies enabling autonomy, or just moving decisions into approval chains?
Avoid · You’ll know it’s working when
Avoid
- Assuming more detailed rules always produce better alignment
- Writing rules to eliminate judgment entirely
You’ll know it’s working when
Routine decisions get made without escalation, and people can say what an agent is authorized to do.