Baseline
Measure the current time, cost, quality and risk before introducing a tool.
PILOT & EVIDENCE LIBRARY
Prove the workflow before you publish the success story.
These are measurement blueprints—not invented client results. Pick a realistic workflow, record the baseline, run a controlled pilot and keep the evidence needed for a defensible decision.
Control
Use the same inputs, acceptance rules and human reviewer for a fair test.
Decision
Count accepted work and correction effort—not generated volume or demo theatre.
Evidence
Publish a result only when the source, method, limits and review date are clear.
Start with work your team already understands.
Each guide includes a baseline, four-stage workflow, success measures and stop/go controls.
01 · Estate agencies
From property brief to reviewable listing pack
A seven-day pilot that measures production time, correction effort and enquiry quality—not vanity output volume.
02 · Marketing teamsCampaign brief to channel-ready creative
A controlled comparison using one brief, one audience and one acceptance rubric across copy, image and video.
03 · NewsroomsSource-led research to publishable briefing
A research workflow that rewards traceable claims and correction speed rather than fluent unsupported copy.
04 · Accountancy practicesMeeting notes to controlled client summary
A low-volume pilot with redacted inputs, an approved template and partner sign-off before any client delivery.
05 · Software teamsIssue description to reviewed code change
A representative task set that measures accepted changes and reviewer effort—not lines of generated code.
06 · Recruitment teamsRole intake to auditable shortlist administration
A human-led workflow for drafting, scheduling and evidence organisation; not an automated hiring decision system.
REPORTING STANDARD
A case study must disclose what could change the result.
- InputTask mix, sample size and data sensitivity.
- MethodTools, plans, prompts, people and test dates.
- OutcomeAccepted output, correction time and total cost.
- LimitsFailures, exclusions and where human review remained.