A practical working resource
Record what is ready, what needs work and what should prevent a live start.
Use the guide, editable worksheet and fictional example together. Download the ZIP, open the HTML guide in a browser and work in a copy of the CSV template.
What is included
- Printable readiness framework and fictional review
- Editable checklist with evidence, owners and actions
- Three prompts for gaps, failure handling and a review record
Who it is for
Workflow owners, delivery leads and business sponsors preparing or reviewing a bounded AI pilot.
How to use it
Complete the record with the people who perform and own the work. Keep evidence and unresolved questions visible. Use the prompts to challenge a draft, then have the relevant owners confirm the decision.
All worked examples are fictional. The material helps structure planning and review; it does not promise a particular outcome.
Using this resource
Decide whether a pilot can start
Use this checklist after choosing a specific workflow and before exposing live work to a proposed AI system. Its purpose is to reveal what needs to be ready, who will make it ready and which unresolved issues prevent a useful experiment. The checklist applies whether an internal team or an external provider will build the solution. It does not certify a product or replace your organization’s approval process.
Bring the workflow owner, sponsor, delivery lead and systems or data owners together. Walk through the proposed user experience from input to final decision. Include a wrong answer, missing data, a system outage and a request outside scope. Record “ready”, “needs work”, “unknown” or “not applicable”, with evidence and an accountable owner. A label without a reason is difficult to review later.
Check the essentials
| Area | Evidence to have before starting |
|---|---|
| Purpose and scope | A single workflow, intended users, boundaries, exclusions and a specific decision the pilot will inform. |
| Ownership | A sponsor who can approve the trial, a workflow owner, delivery responsibility and an escalation contact. |
| Current performance | A baseline with defined measures, observation period and known gaps. A plan to establish missing measures. |
| Data and access | Approved input sources, access permissions, suitable sample preparation and any restrictions on storage or sharing. |
| Evaluation | Representative ordinary and difficult cases, expected outcomes, reviewer instructions and acceptance criteria. |
| Human control | Review checkpoints, clear authority for actions, exception handling and a practical manual fallback. |
| Operation and cost | Bounded users or volume, support arrangements, usage visibility, cost ceiling and responsibility for stopping work. |
| Decision and exit | A review date, expand/revise/stop criteria, evidence to collect and a plan for data, access and handoff at the end. |
Some gaps prevent live operation even when other items are ready. If nobody can approve data access, investigate without that data. If a failure cannot be detected or contained, reduce the scope or keep the experiment in an offline evaluation. Record the reason and the work needed to reconsider it.
Fictional worked example
An example distributor wants a limited pilot that suggests categories for internal service requests. A coordinator will approve every assignment. The workflow owner and delivery lead agree on a narrow group of request types. The sample and manual review process still need preparation.
| Item | Status | Action |
|---|---|---|
| Scope and human review | Ready | Record approved categories and the coordinator’s confirmation step. |
| Data access | Unknown | Systems owner to confirm permitted fields and sample handling before use. |
| Baseline | Needs work | Observe current routing and correction work using the agreed measure. |
| Evaluation cases | Needs work | Prepare ordinary, ambiguous and out-of-scope cases with expected reviewer decisions. |
| Fallback | Ready | Continue the existing manual route when suggestions are unavailable or unsuitable. |
| Usage and cost limit | Needs work | Delivery lead and sponsor to agree visibility and a stopping threshold before live trial. |
Decision: continue preparation; do not begin the live trial yet. No percentage score is needed. The unresolved access and evaluation items determine the immediate action. The example does not imply a result or an expected level of accuracy.
Blank readiness record
| Field | Working notes |
|---|---|
| Pilot, version and review date | ________________________________________________ |
| Sponsor, workflow owner and delivery lead | ________________________________________________ |
| Readiness item and status | ________________________________________________ |
| Evidence reviewed and remaining question | ________________________________________________ |
| Action owner and due date | ________________________________________________ |
| Blocking item and condition for resolution | ________________________________________________ |
| Decision and next review | ________________________________________________ |
The template CSV includes each readiness area and room for evidence, actions, owners and dates. Keep a separate copy for each review so decisions can be traced to the information available at the time. “Not applicable” needs a reason. Do not count an unknown item as ready or average a critical unresolved issue into a positive total.
Use the review to move work forward
Close the session with a short decision: prepare further, approve a bounded start, revise the scope or stop. Name the actions needed for the next decision and the person who will check they are complete. The accompanying prompts help test whether a checklist entry has useful evidence and whether failure handling is clear. Ask the people responsible for the workflow to confirm the result.
More from Authentia
Explore Authentia resources or contact Authentia to discuss the next decision.