Template

Closing Report Template

Show what changed at work, what remains uncertain and what to do next.

Choose the scope of your report

Use this template after practice or a pilot to propose continuing, changing or stopping a way of working with AI. Completing lessons shows learning progress, not workplace impact.

For individual practice, report one task, its results and your next step. You do not need a company rollout plan. For a team or company pilot, also name the business owner, reviewer, data permissions, operating costs and person authorized to approve the next step. Practice does not authorize changes to shared systems.

Include only information needed for the decision and permitted for the report's audience.

1. Define the decision

Field	What to record
Task and scope	People, dates, cases and boundaries
Intended improvement	Target measure and quality conditions agreed before testing
Method	Comparable inputs, timing method, quality checks and reviewer
Proposal	Continue within a stated scope, change the method or stop
Limits	Missing evidence and other changes that could affect results

Cover the tasks actually tested. Do not invent five use cases or require a failure to make the report look credible.

2. Separate observations from forecasts

Task	Before AI	With AI	Evidence type	Source and check
[Task]	[Time, cases, quality]	[Same measures]	Measured / self-estimated / assumed	[Record, date, reviewer]

Measured means supported by observations or records. Self-estimated means reported by a participant without independent timing or verification. Assumed means used for a forecast, not observed.

Include checking, corrections, failed attempts and human intervention in elapsed work time. Report setup time and tool costs separately. Explain differences in case difficulty or workload.

Do not combine these evidence types into a headline presented as measured. A separate forecast may use measured inputs and assumptions if you show the formula, range and sensitivity. Time saved is not automatically cash saved; a before-and-after comparison alone does not establish causation.

3. Worked example: a shorter run is not yet a successful pilot

Illustrative figures from E1 practice data, not an observed MasterOS run:

Measure	Baseline	AI-assisted trial
Cases	4	4
Recorded work time	40 minutes	32 minutes
Accepted outputs	4/4	3/4
Other resources	Record baseline costs separately	API: 0.24 USD; setup/testing: 30 minutes
Independent practice	Not supplied	2/3 participants completed without help

The recorded time is 20% lower: (40 - 32) / 40. But E04 stopped without an accepted output while the access outcome remained under investigation. The two totals do not represent equally completed work. Do not claim a 20% productivity gain, ROI or cash saving.

Proposal: keep the affected route paused, verify what data was returned, correct a defect if confirmed and retest. The business owner decides whether to resume after the checks. This report is a proposal, not approval.

4. Record use, quality and unresolved issues

Evidence to collect

Completion count and criteria, separately from actual workplace use.
Who used the method, how often and how use was confirmed.
Skill/workflow version, owner, permitted scope and source versions where relevant.
Output quality, review effort, corrections, tool costs and incidents.
Confirmed causes versus questions still under investigation; workload, staffing or policy changes that could affect comparisons.

An incident-free trial may still have limited coverage. State what was tested rather than inventing a failure or claiming all risks are controlled.

5. Copy and complete

Task / individual or team scope:
Testing period and cases:
Target measure and quality conditions:
Before / with AI, using the same units:
Evidence type, source and reviewer:
Review/correction time; failed or unfinished cases:
Setup time and tool costs:
Actual use and support needed:
Unresolved issues and comparison limits:
Proposal and permitted scope:
Next action / owner / due date / acceptance check:
Approver, if required / approval status:
Follow-up date:

Before sharing, check that figures have sources and units, forecasts are separate, conclusions fit the tested scope and each next action has an owner. For a company pilot, confirm quality, permissions, costs and incident handling before proposing expansion. Do not treat a proposal or saved report as deployment approval.

Evaluation reference

The Kirkpatrick Model distinguishes workplace application (Level 3) from organizational results (Level 4). Evidence labels, calculations and worksheets here are template conventions, not a Kirkpatrick ROI formula or certification.