Workflow Discovery Pack

Workflow Discovery Pack

Read Workflow Discovery Pack: templates guidance with evidence limits, practical checks and editable source material.

Use for one customer or internal workflow. Capture only what is needed for the next decision; expand the packet as evidence and commitments mature.

Inherited story and accountable field roles

Preserve the brief as received before improving or reconciling it. Treat it as an evidence-bearing hypothesis, not current truth.

Inherited claim Exact source and revision Sold, stated, or assumed Commitment or funding status Limitation or unknown
Role Named person or explicit unknown Evidence or authority basis Scope Verified by
Sponsor
Process knower
Operator
Disposition authority
Independent verifier

Do not infer the process knower, operator, disposition authority, or verifier from the sponsor. One person may hold multiple roles only when each role has a separate, recorded basis.

Engagement charter

Field Value
Engagement ID and organization/tenant boundary
Approved evidence store, retention, and deletion rule
Workflow ID
Executive sponsor
Independent sponsor or business-owner backup
Operational owner
Technical owner
Risk owner
Intended users/operators
Receiving service owner
Trigger
Current work product or decision
Initial segment and exclusions
Accepted outcome
Baseline status and evidence Measured / estimated / unmeasured
Target
Measurement window
Verifier
Maximum acceptable failure
Pilot stop conditions
Pilot maximum duration / evidence cutoff / graduation decision date
Continuation mechanism and decision owner External deployment or renewal / internal funding or sponsorship / other

Use the machine-readable workflow-charter template for the governed decision after discovery.

Strategic and operating apertures

Use this as a reconciliation pass, not a fixed interview script. Begin with the widest relevant view that is available, inspect the work and its technical seams, then return unresolved differences to the people who hold the relevant decision rights.

Aperture Goal, blocker, workflow claim, or interaction Exact source and owner Evidence to inspect next Conflict, limitation, or accountable next move
Strategic intent
Operating reality
Customer-visible effect, if any
Candidate value surface Current experience or exposure Proposed observable change Acceptance evidence and verifier Why this candidate now
Customer-visible interaction / internal enablement / cost or capacity / risk and control

Do not require a customer-visible effect when an internal safety, control, or operating decision is more consequential. Do not treat executive sponsorship, interview volume, a survey, or a roadmap as proof of the workflow or permission to proceed.

Field evidence register

Evidence ID Method Case/segment Date Source owner Redaction/classification Observation supported Limitation
Interview / shadow / recording / event log / artifact / metric

Field claim comparison and accountable reframe

Claim ID Class Statement Exact source and owner Scope and limitation Contradiction or missing evidence Currentness
sold / stated / observed / system-enforced / policy-authorized current / contested / superseded

No claim class universally outranks another or grants authority. Observation proves occurrence in scope; system testing proves exact-build behavior; policy remains normative within its revision and scope; a stakeholder assertion proves only that it was stated.

Material conflict Affected boundary or outcome Safe fallback while unresolved Proposed reframe Required disposition and authority Next field move
continue discovery / bounded kickoff / defer / stop find owner / observe case / resolve cited conflict

When field evidence materially changes the brief, use the machine-readable engagement-reframe record to preserve the competing claims, scoped decision, selective downstream impact, and chronology.

Interaction receipts and current readout

Keep source material separate from proposed conclusions. For each meeting, observation, incident review, or artifact handoff, record the interaction ID, engagement boundary, source revisions, proposed additions, named review disposition, appended record IDs, and timestamp in the field-observation log. Preview the changes before append; rejected and deferred proposals remain visible.

The sponsor readout is a projection of these dated receipts and the current governed records. It states the evidence cutoff, current boundary, measured outcome, unresolved consequential decision, next move, and source links. Do not maintain it as an independent status artifact or let it silently override the engagement reframe, value case, release evidence, or service review.

Workflow event log

Sequence Actor Event Input System Decision Evidence Effect Exception
1

Source-of-truth map

Source ID Objects Owner Read/write Classification Freshness SLO Identity API maturity Failure mode

Technical reality check

Use direct evidence to test the stated workflow before assigning readiness or proposing architecture. A questionnaire, interview, diagram, or maturity score organizes hypotheses; it does not prove the target environment is ready.

Surface inspected Exact target / revision / environment Owner Evidence observed Claim supported or contradicted Gap, consequence, and next test
Code or configuration
Data pipeline, source seam, or reconciliation
Identity, permission, tenant, credential, or network boundary
Representative execution, telemetry, failure, or recovery path
Release, rollback, support, or dependency constraint

Mark an uninspected decision-bearing surface unknown. Do not infer readiness from document completeness or replace missing target access with a higher confidence score.

Decision decomposition

Decision Deterministic rule Agent judgment Human accountability Required evidence Postcondition

Context extraction and operational redesign

Interview notes, recordings, documents, and demonstrations are discovery evidence. They do not become prompts, policy, memory, or production authority until the owning role validates them.

| Candidate knowledge | Observed source and owner | Classification / scope | Type | Validate with | Keep, repair, remove, escalate, or abstract | Production destination / invalidation trigger | | --- | --- | --- | --- | --- | --- | | | | Public / internal / confidential / restricted | Fact / rule / judgment / workaround / exception / instruction | | | |

Use this table to distinguish current evidence from rules, professional judgment, and historical scar tissue. Keep customer-specific context with the workflow owner. A reusable product or pattern candidate must be sanitized, recurrence-tested, and routed through the field-learning register; never copy customer data or local policy into a shared artifact.

Exception taxonomy

Exception ID Frequency Severity Detection Current handling Target handling Escalation owner Eval slice

Operational ontology candidates

Kind ID Source Identity key States/relations Policy Action Invariant
Entity

Value-verifiability matrix

Candidate Annual volume Minutes/case Loaded labor/hour Error rate Error cost/event Annualized integration + operating cost Verifier coverage (0–1) Adoption probability (0–1) Residual risk probability (0–1) Priority ratio
gross_annual_value = annual_volume × (
  minutes_per_case / 60 × loaded_labor_cost_per_hour
  + error_rate × error_cost_per_event
)

risk_adjusted_verified_value =
  gross_annual_value
  × verifier_coverage
  × adoption_probability
  × (1 - residual_risk_probability)

priority_ratio = risk_adjusted_verified_value
                 / max(1 currency unit, annualized_integration_and_operating_cost)

Use one declared currency and annual period. Bound every probability or coverage input to 0..1, preserve low/expected/high scenarios, and do not compare ratios built from different units or periods. This ratio ranks qualified candidates; it does not override a hard owner, verifier, authority, safety, or adoption gate. Keep estimated, measured, and realized value separate. Complete the value-case template before a pilot decision.

Assumption and decision log

ID Assumption or decision Evidence Owner Test or review date If false Status
Open / validated / rejected / superseded

Readiness gates

Charter dimension Required evidence Owner Score (0–4) Status
Workflow clarity User, interface, decision, inputs, permitted action, accepted outcome, boundary Operational
Context Owners, source-of-truth rules, revisions, freshness, classification Data
Verifier Independent completion evidence, representative cases, safety slices, isolated graders Evaluation
Integration Typed tool contracts, identity, authorization, duplicate safety, dependency ownership Platform
Adoption User surface, training, review capacity, support, workflow integration Operational
Operations SLOs, alerts, runbooks, rollback, capacity Operations
Risk Threat model, secrets, egress, tenant isolation, prohibited effects, recovery Security/risk

Copy these seven dimensions and their evidence into workflow-charter.readiness. Authenticated authority, accountable ownership, and a lawful data path remain independent hard gates even when the scored dimensions are otherwise strong.

Adoption and ownership plan

Capability Current owner Receiving owner Evidence of readiness Exercise Exit condition
Workflow and policy
Data and tools
Evaluation and release
Support and incident response
Cost and value review
Retirement

Stakeholder value and change contract

Audience Current work and consequence Expected benefit New responsibility or control Evidence of acceptance Owner
Executive or operational sponsor Outcome, risk, or economics Fund, unblock, and decide scope Baseline, guardrails, and value review
Operator or domain expert Better evidence, reduced rework, or faster exception handling Review, correct, escalate, or pause Shadow comparison, training, feedback, and workload measures
Product or applied-AI team Reusable capability or reduced delivery friction Preserve the customer-specific boundary and own shared changes Sanitized recurrence evidence and release decision

Controls: FDE-001, FDE-002, FDE-003, FDE-004, FDE-005, ADP-001, CTX-001, CTX-002.

Cutover plan

Stage Traffic Write capability Entry gate Exit gate Rollback trigger
Offline 0% None Replay fixtures ready Thresholds pass Any prohibited effect
Shadow Mirrored None Observability complete Stable slice metrics Data or policy drift
Canary 1–5% Staged/reversible Human review, support, and rollback ready SLO, adoption, and review thresholds Error-budget burn
Bounded production Named segment Policy-gated Customer owners and runbooks exercised Outcome, adoption, and continuous controls Kill-switch or value threshold