A service can be running while the people using it still have unresolved work. These support assets help a team define an acceptable outcome, investigate without changing the service and check how it would recover.
Each task pairs a human brief with an artificial intelligence (AI) prompt template, required inputs, expected outputs and a fictional worked example. Adapt the record to your sources and existing authority. The download contains the complete template; the descriptions below help you choose one.
Acceptance criteria
Define observable outcomes and the checks needed to accept a service or support change.
State whose outcome matters and under which conditions. Describe the expected behaviour, failure response and observation needed. Name the person who accepts the result and keep targets separate from actual observations.
Human brief and AI prompt template — Markdown
Example. For the fictional service, an accepted handover must leave one current owner, retain pending work and return a receipt to an authorised caller. The checks have not run, so the sample has no actual results. An availability target remains for the service owner to agree.
Read-only diagnosis
Establish the observed service position without changing it during investigation.
Collect the relevant records using authorised read access. Preserve timestamps, environment and source identity. Summarise symptoms and affected work, distinguish live observations from historical or simulated records, and identify useful next checks.
Human brief and AI prompt template — Markdown
Example. For the fictional lost response, support reads the request record, case version and receipt store. The example describes what to collect; no live incident has occurred and no booking is retried.
Causal uncertainty
Assess possible causes while retaining missing information and competing explanations.
Compare each plausible explanation with observations that support or contradict it. Consider timing, dependency direction and shared causes. Choose the next observation that could distinguish explanations rather than forcing a definitive root cause.
Human brief and AI prompt template — Markdown
Example. A fictional missing booking could reflect failed acceptance, failed notification after acceptance or incomplete records after restoration. The sample leaves the cause open and proposes checking the durable receipt and restored-record history.
Runbook validation
A runbook records the steps for a known service task. Check whether that procedure is suitable for the actual target and current conditions.
Identify the runbook edition and target. Check prerequisites, permissions, expected effects and what might have changed. Establish stop conditions, verification and recovery, including business actions that a technical rollback would not undo.
Human brief and AI prompt template — Markdown
Example. A fictional restore runbook brings back data from before a handover. Validation identifies the risk of losing the receipt while staff have already moved the work. The proposed procedure must reconcile those actions before resumption.
Honest measures
Use measures that describe the outcome and cost relevant to a decision.
Define the observation before interpreting it. Record the unit, population, start/end events, period, exclusions, source and owner. Include failure, review and rework where they affect the outcome, and explain what a figure cannot establish.
Human brief and AI prompt template — Markdown
Example. A fictional cancellation measure counts resolved learner cases as a proportion of all in-scope cases during a stated period. It reports outstanding cases and repeat contacts separately. A cost-per-resolution measure includes relevant work on unresolved cases, review and rework. Tokens are the units of text a language model processes or produces. A fast response from an application programming interface (API), or a low token count, does not establish that the learner’s case was resolved.
Connected guidance
Supporting a service connects these tasks to people, records, response and learning. Metrics explains how to define and interpret observations.
AI-supported architecture describes how these activities can use AI assistance. Browse the other IT practice collections when a task crosses governance, delivery and support.
Methodology configuration
Use the terms and records your team already understands. Keep their meanings, source versions and decision owners clear. Methodology configuration explains how to map the activities and reviews to your own approach.
Edition and reuse
Version 1.0 · September 2026 · © 2005–2026 Adrian Sutherland.
Portal-authored downloads use CC BY 4.0. Keep the author, source edition and licence with adaptations, and identify changes. Read the edition note and use the asset index — JSON for the complete list of tasks and download addresses.