Process

One controlled engagement shell, four delivery modules.

The work should remain inspectable whether the outcome is a low-confidence conclusion, an inconvenient finding, or a working automation pilot. Method confidence, result confidence, and production readiness are deliberately kept separate.

Method overview

Five stages, with an integrity gate at each handoff.

The exact work changes by offer. The control logic does not: diagnose the decision, freeze the boundary, preserve source and configuration identity, challenge the output, test failure behaviour, and release a versioned package with explicit limitations.

Core rule: a polished summary or functioning demo never replaces the authoritative record. Every material conclusion or automated action should retain a route back to its scope, source, configuration, version, test state, and human review where applicable.

Diagnose and qualify

For the free one-workflow diagnostic: decide whether paid design work is justified, without free written maps, registers, or solution design.

Confirm one named workflow, owner, intended users, and consequence
Test whether evidence and access are sufficient to assess responsibly
Keep the free written outcome short: proceed, not a fit, or insufficient access
Defer maps, gap registers, control design, and build scoping to paid work

Freeze scope, evidence, and operating boundary

Create the authoritative engagement record before substantive analysis or implementation begins.

Define the assessment object or workflow, start and end points, audience, and intended use
List included systems, sources, users, deliverables, exclusions, assumptions, and fallback
Set evidence cut-off, review rounds, acceptance criteria, change control, confidentiality, and release profile
Assign stable engagement, source, finding, configuration, test, and deliverable identifiers where applicable

Apply the appropriate module

Use a stable control discipline while changing the object of work.

Workflow review: map actors, systems, records, transformations, controls, exceptions, and reconstruction risk
Controlled automation: separate deterministic, AI-assisted, human, and prohibited steps; implement the bounded pilot and retained operating record
Claims: separate claim meaning, candidate evidence, support verdict, buyer consequence, and remediation
Research: define the question, source protocol, competing interpretations, uncertainty, and decision consequence

Internal challenge, failure testing, and release gate

Test the work before polished presentation or a successful happy-path demo creates false confidence.

Confirm every material conclusion or automated action has a traceable basis
Keep evidence, observed fact, inference, generated output, human judgment, recommendation, and limitation distinct
Test missing, malformed, conflicting, duplicate, stale, unauthorized, and out-of-scope conditions where relevant
Confirm human review, override, escalation, rollback, and manual fallback behaviour
Check redaction, permissions, versions, accessibility, security responsibilities, and public-release status

Controlled delivery, handover, and decision workshop

Release the package or pilot as a governed record, then convert it into action with the people responsible for operation and the decision.

Present the assessment disposition, research conclusion, or pilot outcome
Walk through the evidence boundary, workflow operation, failure states, conditions, and limitations
Assign owners for immediate actions, maintenance, monitoring, and longer evidence work
Record closeout lessons, permission state, reusable components, and reassessment triggers

Controlled language

Different questions require different kinds of conclusions

Operational readiness, pilot acceptance, claim support, evidence completeness, assessment confidence, and CAL run validity are related but not interchangeable.

Operational disposition

Can the decision or workflow proceed?

Ready, conditionally ready, not ready, insufficient evidence, accepted with limitations, requires remediation, or do not rely.

Claim verdict

What does the evidence support?

Supported, partially supported, reasonable inference, unsupported, contradicted, or not checkable.

Run or pilot validity

Was the analytical or implementation result reliable?

Valid or accepted, valid with limitations, review required, invalid, incomplete, or exploratory only.

A supported claim does not automatically make a product buyer-ready. A completed CAL run does not automatically make the run valid. A functioning automation demo does not automatically make the workflow production-ready.

Deliverable architecture

Three layers for three kinds of reader

Executives should not need to read the full register, but reviewers and operators should be able to inspect how the conclusion or workflow was produced.

Decision layer

Assessment disposition, pilot outcome, or decision brief; material findings; gating conditions; prioritized actions; confidence; and completeness.

Working layer

Workflow maps, target designs, test cases, claim registers, source and evidence registers, finding and risk registers, matrices, and recommendations.

Evidence and operating layer

Method, source references, configurations, reviewed material, assumptions, failure results, overrides, limitations, versions, ownership, and evidence cut-off.

Review and release control

Final documents and pilots are controlled releases, not silently changing demos.

Material changes require a version increment and a change summary. Public examples require an explicit synthetic, redacted, research, or permissioned release profile. Implemented workflows also need named ownership, known limitations, maintenance responsibility, and reassessment triggers.

Release checks

IDs, versions, status, evidence cut-off, method or configuration version present
Claims, findings, outputs, and tests resolve to controlled sources or inputs
Contradictions, exceptions, inferences, failures, and limitations remain visible
Human review, fallback, ownership, synthetic status, and relationship context are explicit
Accessibility, print, link, redaction, and handover checks complete

Start with the workflow or decision, not a predetermined tool.

The free one-workflow diagnostic decides whether the design sprint is justified: proceed, not a fit, or insufficient access. Free written maps and solution design stay out of scope.

Request the diagnostic Compare the offers