Fit diagnosis and product classification are different
Decisagent Fit v0.4 applies an explicit ruleset to user-supplied context and returns a provisional operating-design recommendation. It does not empirically classify a user's organization or deployment. Product maturity classification uses accepted, exact-scope evidence under Methodology v0.2. Candidate research never enters Fit or the public read model.
Both systems keep useful AI participation separate from permissible action authority. Neither collapses the result into a single numerical score.
The accepted-intelligence pipeline
Research, extraction, and deterministic derivation may propose. They may not accept claims, classifications, or relationships. Public pages consume only the generated human-accepted projection.
Each assessment is scoped to a product, component, configuration, deployment, or research artifact. Evidence from one scope cannot silently raise another. Every criterion is recorded as SUPPORTED, CONTRADICTED, UNKNOWN, CONFLICTING, or NOT APPLICABLE.
Cumulative DA-0–DA-5 capability levels
Higher is not inherently better. An irreversible or high-error-cost decision may responsibly use a lower-authority design even when more automation is technically possible. Methodology v0.2 historically names DA-3 “Governed Decision Agent”; its actual threshold is narrower: traceable decision support plus at least one explicit governance signal. It does not establish complete operational governance.
| Level | Evidenced capability | Cumulative threshold | External action | Outcome loop |
|---|---|---|---|---|
| DA-0 | Useful analysis without a first-class external decision | No cumulative Decision Agent threshold | No | No |
| DA-1 | Named decision plus evidence assembly | Named decision + decision-specific evidence assembly | No | No |
| DA-2 | Explicit alternatives evaluated into a recommendation or justified abstention | DA-1 + explicit alternatives + evaluated alternatives + recommendation/abstention | No | No |
| DA-3 | Traceable decision support | DA-2 + decision trace + at least one explicit objective, constraint, or uncertainty signal | Not established; a service/API may execute decision computation | No |
| DA-4 | Decision plus bounded external-action capability | DA-3 + external action + exact mode/scope + policy-bound preauthorization or action-time human approval | Capability evidenced at scope; operational release still requires Assurance | Not required |
| DA-5 | Closed outcome and learning loop | DA-4 + decision identity + observed outcome + attribution + expected-vs-actual + governed later-decision update | Yes | Yes |
A human may accept a level below the evidence ceiling with an explicit reason. The ordinary review path cannot accept a level above the ceiling. DA-4 classification establishes action capability at the assessed scope, not implementation assurance. Releasing external action additionally requires the operational gates and accountable authority record in a Decision Assurance Case.
All 21 Methodology v0.2 criteria
Each criterion below states what is assessed, observable evidence, a common false positive, and its maturity effect. The glossary is descriptive; the versioned code rules remain authoritative.
01 · DECISION
Named decision
Definition: The assessed scope represents a named, externally meaningful decision rather than only an internal task or tool choice.
Why it matters: Without a stable decision object, evidence, alternatives, authority, and outcomes cannot be tied to the choice that matters.
Observable evidence: A defined decision statement, accountable owner, trigger, boundary, and decision-specific inputs or output.
Common false positive: Internal tool routing, prompt planning, or workflow branching is not automatically the named external decision.
Effect on maturity: Required with evidence assembly for DA-1 and every higher level.
02 · EVIDENCE
Evidence assembly
Definition: The scope assembles information relevant to the named decision with enough structure to support the decision process.
Why it matters: Decision support must work from decision evidence, not merely generate fluent text.
Observable evidence: Decision-specific retrieval, joins, contextual records, effective-time handling, or documented input contracts.
Common false positive: Generic retrieval or summarization does not establish decision evidence assembly.
Effect on maturity: Required with a named decision for DA-1 and every higher level.
03 · EVIDENCE
Gap detection
Definition: The scope identifies missing, conflicting, stale, or insufficient evidence that affects the decision.
Why it matters: Known gaps support abstention and prevent absence from being interpreted as negative evidence.
Observable evidence: Decision-specific completeness checks, conflict flags, freshness gates, or explicit insufficiency states.
Common false positive: A generic data-quality warning is not decision-specific gap detection.
Effect on maturity: Strengthens safe abstention but is not a formal level prerequisite in Methodology v0.2.
04 · DECISION
Explicit alternatives
Definition: Externally meaningful courses of action are represented as first-class alternatives.
Why it matters: A decision requires real choice, including defer, abstain, or escalate when appropriate.
Observable evidence: Named options or action candidates tied to the external decision boundary.
Common false positive: Workflow steps, tools, prompts, or internal branches are not necessarily decision alternatives.
Effect on maturity: Required for DA-2 and every higher level.
05 · DECISION
Alternatives evaluated
Definition: The scope compares explicit alternatives rather than merely listing them.
Why it matters: Evaluation makes tradeoffs, constraints, and the basis of a recommendation inspectable.
Observable evidence: Rules, scores, optimization, scenarios, or structured comparison applied to the named alternatives.
Common false positive: An option list without comparative evaluation does not satisfy this criterion.
Effect on maturity: Required for DA-2 and every higher level.
06 · DECISION
Recommendation produced
Definition: Evaluation produces a recommendation, proposal, decision, or justified abstention.
Why it matters: The result must resolve the decision-support step rather than stop at information presentation.
Observable evidence: A selected alternative, ranked proposal, decision result, or decision-specific abstention with rationale.
Common false positive: A score, summary, answer, or dashboard signal is not necessarily a recommendation.
Effect on maturity: Required for DA-2 and every higher level.
07 · GOVERNANCE
Explicit objectives
Definition: At least one decision objective is represented explicitly.
Why it matters: Objectives expose what the decision seeks to improve and support review of tradeoffs.
Observable evidence: Named target measures, optimization goals, priorities, or objective hierarchy used by the decision.
Common false positive: A KPI mentioned in marketing copy does not prove an explicit decision objective.
Effect on maturity: One of three explicit governance signals; at least one objective, constraint, or uncertainty criterion is required for DA-3.
08 · GOVERNANCE
Explicit constraints
Definition: Hard boundaries or eligibility rules are represented explicitly in the decision process.
Why it matters: Constraints distinguish permissible alternatives from attractive but unauthorized or infeasible ones.
Observable evidence: Policy rules, limits, eligibility, capacity, contractual, safety, consent, or regulatory gates.
Common false positive: A prompt instruction or prose guideline is not proof of an enforced constraint.
Effect on maturity: One of three explicit governance signals; at least one objective, constraint, or uncertainty criterion is required for DA-3.
09 · GOVERNANCE
Explicit uncertainty
Definition: Material uncertainty is represented as part of the decision rather than hidden behind a point answer.
Why it matters: Uncertainty can change the preferred alternative, required review, or abstention posture.
Observable evidence: Ranges, confidence with calibrated semantics, scenarios, sensitivity, missing-evidence states, or abstention thresholds.
Common false positive: A generic confidence score or disclaimer is not explicit decision uncertainty.
Effect on maturity: One of three explicit governance signals; at least one objective, constraint, or uncertainty criterion is required for DA-3.
10 · GOVERNANCE
Decision trace
Definition: The scope produces a reproducible record of the decision basis and result.
Why it matters: A trace supports review, challenge, incident response, evaluation, and later comparison.
Observable evidence: Decision ID, time, alternatives, evidence references, objectives, constraints, uncertainty, result, and relevant policy/model versions.
Common false positive: A transcript, debug log, chain-of-thought, or generic audit event is not a reproducible Decision Record.
Effect on maturity: Required for DA-3 and every higher level.
11 · EXECUTION
Decision-service execution
Definition: The scope can execute decision computation or expose a decision result through a service or API.
Why it matters: This distinguishes operationalized decision logic from static analysis while preserving the action boundary.
Observable evidence: A callable decision service, rules runtime, decision endpoint, or documented runtime execution path.
Common false positive: A decision-service API response does not prove a consequential external operational action.
Effect on maturity: Useful execution evidence, but it is not by itself a DA-4 prerequisite or proof of external action.
12 · EXECUTION
External operational action
Definition: The exact assessed scope can cause a consequential effect outside the decision computation itself.
Why it matters: External effect creates the material DA-3 to DA-4 safety and authority boundary.
Observable evidence: A documented connector, transaction, communication, operational change, order, or other external effect caused by the assessed scope.
Common false positive: Vendor language that a system 'can act' without exact mode, scope, and effect does not establish external action.
Effect on maturity: Required for DA-4 and every higher level.
13 · AUTHORITY
Policy-bound execution
Definition: External action is pre-authorized by an explicit policy with scope, conditions, and limits.
Why it matters: Standing authority must be narrower than general system permission and auditable at action time.
Observable evidence: Named permitted action, object scope, thresholds, conditions, duration, exception, trace, and stop controls.
Common false positive: A broad role permission or agent credential is not a bounded execution policy.
Effect on maturity: One of two permissible DA-4 authority paths; policy-bound execution or action-time human approval is required.
14 · AUTHORITY
Action-time human approval
Definition: A recorded human approval authorizes the specific consequential action at action time.
Why it matters: Human review is meaningful only if it can approve, reject, or alter the actual action before effect.
Observable evidence: An approval state tied to the action proposal, approver role, timestamp, limits, and execution record.
Common false positive: Prior generic approval or an after-the-fact notification is not action-time authorization.
Effect on maturity: One of two permissible DA-4 authority paths; action-time approval or policy-bound execution is required.
15 · AUTHORITY
Execution mode and scope
Definition: The acting mode, permitted actions, target scope, and operational limits are explicit for the assessed scope.
Why it matters: A product may support advisory and effectful modes; classification cannot inherit the broadest possible mode.
Observable evidence: Exact configuration or deployment documentation identifying mode, action, resource, channel, limits, and authority boundary.
Common false positive: Broad product capability or another component's mode cannot establish this scope's execution mode.
Effect on maturity: Required with external action for DA-4 and every higher level.
16 · AUTHORITY
Override or exception path
Definition: A defined person or control can stop, override, revoke, or route an exception.
Why it matters: Consequential systems need a practical path when evidence, policy, or operation departs from assumptions.
Observable evidence: Named stop/revocation control, exception owner, escalation route, and recorded resolution.
Common false positive: An undefined 'kill switch' or generic administrator role is not an operational override path.
Effect on maturity: Important safety evidence but not a formal level prerequisite in Methodology v0.2.
17 · OUTCOME
Originating decision identity
Definition: The downstream record retains the stable identity of the decision that caused or proposed the action.
Why it matters: Outcome learning is impossible if later results cannot be joined to the originating decision.
Observable evidence: A decision ID propagated through authorization, execution, and outcome records.
Common false positive: A session, conversation, workflow, or transaction ID is insufficient unless it identifies the originating decision.
Effect on maturity: One of five cumulative outcome/learning prerequisites required for DA-5.
18 · OUTCOME
Downstream outcome observed
Definition: A downstream result of the decision is later observed.
Why it matters: Completion telemetry is not proof that the decision achieved or harmed the intended outcome.
Observable evidence: A measured business, operational, service, risk, or other decision-relevant result at a defined horizon.
Common false positive: System telemetry, model accuracy, click success, or process completion is not necessarily the downstream outcome.
Effect on maturity: One of five cumulative outcome/learning prerequisites required for DA-5.
19 · OUTCOME
Outcome attributed to decision
Definition: The system has an explicit basis for relating the observed outcome to the originating decision.
Why it matters: Many later changes can influence a result; temporal sequence alone can mislead.
Observable evidence: A comparator, holdout, causal design, cohort logic, or bounded attribution posture linked to the decision.
Common false positive: An outcome occurring after a decision does not establish attribution.
Effect on maturity: One of five cumulative outcome/learning prerequisites required for DA-5.
20 · OUTCOME
Expected versus actual
Definition: The expected outcome recorded at decision time is compared with the later observed result.
Why it matters: Predeclared expectations prevent hindsight from rewriting what the decision was supposed to achieve.
Observable evidence: Versioned expected range or target, observed result, variance, and review at the declared horizon.
Common false positive: A retrospective dashboard without a decision-time expectation is not expected-versus-actual comparison.
Effect on maturity: One of five cumulative outcome/learning prerequisites required for DA-5.
21 · LEARNING
Governed update to later decisions
Definition: Outcome evaluation can change a later decision policy only through a defined, reviewable governance process.
Why it matters: Learning must not silently expand authority or alter policy from noisy feedback.
Observable evidence: A review cadence, change proposal, accountable approver, versioned policy update, and rollback or monitoring plan.
Common false positive: Generic retraining, online learning, or 'continuous improvement' is not a governed later-decision update.
Effect on maturity: One of five cumulative outcome/learning prerequisites required for DA-5.
A decision service is not external action
A DA-3 system may compute or expose a governed decision through an API. DA-4 additionally requires evidence that the exact assessed scope causes a consequential external effect, the execution mode and scope are explicit, and the action is either tightly pre-authorized by policy or approved by a human at action time.
Recommendation ≠ authorization ≠ decision-service execution ≠ external operational action ≠ successful outcome.
DA-5 requires the whole outcome loop
Monitoring, retraining, adaptive models, and generic “continuous learning” do not satisfy DA-5. All five outcome/learning criteria are cumulative: the originating decision identity, observed downstream outcome, attribution, expected-versus-actual comparison, and a governed update to a later decision process.
Evidence posture, human review, and correction
Evidence posture records who supplied or verified a claim: vendor claimed, vendor documented, customer reported, third-party research, or independently tested. These labels are not interchangeable. INDEPENDENTLY TESTED requires a linked Decisagent evaluation record.
Accepted revisions are immutable and dependency-bound. Material new evidence can invalidate a dependent classification until it is reviewed again. UNKNOWN and CONFLICTING remain visible rather than being silently filled.
Apply the methodology to your decision Report evidence or interpretation changes →