Strategic Decision Graph

How can the next consequential business decision start from what the last one taught you?

A Strategic Decision Graph records how context became commitment. It connects the purpose, problem, evidence, options, authority, action, outcome, and lesson behind a decision.

Most business systems record the resulting state. A CRM may record a 20% renewal discount. A Strategic Decision Graph records why the exception was allowed, who approved it, what the business expected, what happened to the customer, and whether the next renewal should follow that precedent.

The goal is not to automate judgment. It is to help people and authorised AI make better-supported decisions and learn honestly from their consequences.

State Record And Decision Trace

System	What it records	Question it answers
CRM, ERP, or ticket system	Customer, transaction, case, or current state	What is true now?
Data warehouse	Historical events and analytical state	What changed?
Policy system	Rules and permitted constraints	What should usually happen?
Strategic Decision Graph	Purpose, evidence, options, authority, action, and outcome	Why did we choose this, and should we do it again?

A state record tells you where the business landed. A decision trace lets you judge whether the route should become precedent.

How Judgment Compounds

Purpose → Problem → Question → Decision → Action → Evidence
   ↑                                               ↓
   └──── wiser judgment ← evaluated precedent ← review

The graph compounds only when an observed outcome changes future judgment. Capturing more decisions without reviewing them creates an archive, not wisdom.

Wisdom means choosing the right action, at the right time, in the right way, for the right reasons, toward the best available outcome for the named beneficiary.

Judgment test	What the trace retains
Right action	Options, chosen lever, prediction, and accepted loss
Right time	Trigger, decision window, urgency, and review point
Right way	Method, permissions, constraints, and human gate
Right reasons	Purpose, evidence, principles, assumptions, and dissent
Best available outcome	Beneficiary, trade-offs, uncertainty, and observed variance

Judge the decision from what was knowable at the time. Judge the outcome separately so luck does not masquerade as wisdom.

The Smallest Trustworthy Trace

Each consequential decision needs only enough information to explain and review it later.

Field	Question	Evidence
Purpose	Whose valued future required a decision?	Beneficiary, setpoint, protected value
Problem	What gap required action?	Starting state, consequence, uncertainty
Options	What credible routes were compared?	Alternatives, predictions, accepted losses
Timing	Why act now, later, or not at all?	Trigger, decision window, review point
Authority	Who was allowed to decide?	Human owner, consent, escalation rule
Action	What commitment changed the state?	Method, constraints, approval, handoff
Outcome	What happened for the beneficiary?	Result and variance against the prediction
Precedent	What may the next judgment reuse?	Accepted, rejected, or review-required lesson

Retain supplied evidence, policy, authority, and rationale. Do not expose private deliberation, secret reasoning, or model chain-of-thought.

Worked Example

A renewal agent proposes a 20% discount when policy normally permits 10%.

The CRM records the final discount. The decision trace also links:

• recent service incidents and churn risk;
• the policy version in force;
• a comparable approved exception;
• the finance approver and consent boundary;
• the prediction made before the discount;
• the customer's later renewal and product usage.

At review, the business can ask whether the exception protected the relationship, merely delayed churn, or set a harmful precedent. The answer can change the next renewal decision.

When A Graph Is Worth Building

Start with one workflow where several of these conditions hold:

• people repeatedly rebuild the same context;
• exceptions are common and precedent changes the answer;
• the decision crosses several systems or teams;
• approval happens outside the system that records the result;
• errors carry material financial, safety, compliance, or trust costs;
• a human glue role exists because no application sees the whole decision.

Routine deterministic work may need automation, but it rarely needs a decision graph. Start where “it depends” is honest and the reason matters.

Business Value Is A Hypothesis

A Strategic Decision Graph may improve:

• decision speed by retrieving relevant precedent;
• decision quality by comparing evidence, exceptions, and outcomes;
• coordination cost by reducing repeated cross-system synthesis;
• governed autonomy by making authority and escalation visible;
• trust and auditability by preserving why an action was permitted.

Foundation Capital describes context graphs as a possible new system-of-record category. A Strategic Decision Graph narrows that thesis to decision lineage: the trace connecting purpose, judgment, authority, action, and outcome.

This is an informed market thesis, not proof of demand or defensibility. Proprietary precedent, cross-system context, learning effects, and switching costs become advantages only if retained traces improve decisions and remain useful across people, tools, and time. Lock-in without portable evidence is a risk, not customer value.

Build One Before Building A Platform

Use a Decision Journal to capture one live choice:

1. Name the beneficiary and valuable outcome.
2. Record the facts available before the decision.
3. Compare credible options and accepted losses.
4. Name the human authority, consent boundary, and stop condition.
5. Record the prediction, action, and review point.
6. Compare the outcome with the starting state.
7. Mark the trace as precedent, counterexample, or unresolved evidence.

Proof of done: another authorised person can explain why the decision was reasonable from what was knowable at the time, separate judgment quality from outcome luck, and know whether to reuse or escalate it without inventing missing context.

Measure Value, Not Graph Volume

Before treating the graph as an asset, compare it with a frozen starting state:

• Did comparable decisions take less time?
• Did preventable reversals, escalations, or repeated exceptions fall?
• Did the beneficiary outcome improve?
• Can an auditor reconstruct evidence and authority at commit time?
• Can a new team member apply precedent without copying an old answer blindly?
• Can the organisation export its decision evidence without losing meaning?

Revenue, retention, and workflow volume may show commercial demand. They do not by themselves prove better judgment or beneficiary value.

Human Authority And Failure Modes

AI can retrieve, compare, predict, and critique. It cannot own human values, consent, moral accountability, or irreversible commitments.

Watch for:

• bad precedent — repeated decisions can compound error;
• false causality — an outcome after a decision does not prove causation;
• surveillance — decision lineage must not become indiscriminate monitoring;
• authority drift — precedent cannot silently replace consent or judgment;
• context debt — stale evidence and unversioned policies make traces unsafe;
• graph theatre — nodes and edges that change no decision create no value.

Changes my mind: comparable exception-heavy workflows improve as reliably without decision traces, or retained traces fail to improve speed, quality, auditability, or governed autonomy after mature review.

Context

• depends-on Purpose — define the beneficiary and protected values before retaining organisational judgment.
• depends-on Business Strategy — choose the position and accepted loss before deciding which traces deserve investment.
• applies-to Decision Making — frame and review each consequential choice.
• applies-to Decision Journal — learn the method with one decision before building infrastructure.
• pairs-with Context Graphs — make relevant evidence queryable across agents, entities, and time.
• proved-by Performance — test whether precedent improves beneficiary outcomes rather than graph volume.

Source

• Foundation Capital, “AI's trillion-dollar opportunity: Context graphs”, 22 December 2025.

Questions

Can you explain the difference between a state record and a decision trace?

• Which repeated judgment carries the highest coordination or exception cost?
• Which system sees its full context at commit time?
• What outcome would prove that retaining precedent improves the next decision?

Next question: which judgment call should become your first trustworthy decision trace?