For management and advisory consultancies, start recurring client reporting by measuring the complete production cycle and naming the person who accepts the output. Improve the method, template or current software first where it meets the need. A custom workflow is worth testing only when it can prepare or check defined work without taking over expert judgement.
Use the editable worksheet: download the report-production measurement worksheet (CSV).
Start with one report you already deliver
Choose one report type with a consistent client purpose. Do not combine every dashboard, board pack and ad-hoc request. Record the trigger, accepted output, accountable reviewer, permitted sources, calculation rules and the reasons work returns.
This makes the question concrete: which steps can be prepared or checked more reliably, while a qualified person still owns the conclusion? A system may assemble a draft workpaper, compare a current figure with an approved rule or flag a missing source. It should not silently approve a finding or issue a report without the named reviewer.
Measure complete effort, not draft speed
Use the same definitions for comparable reports:
Active production minutes = preparation + calculation + draft + review + correction + maintenance/coordination.
Record waiting time separately; it is elapsed delay, not active labour. Record one-off setup, testing and training separately too. They matter to a decision, but they are not routine report-production minutes.
| Activity | Record | Keep separate from |
|---|---|---|
| Preparation | locating permitted inputs, permissions checks, extraction and normalisation | drafting alone |
| Calculation | reconciliation, calculations and approved definitions | an automated decision |
| Draft | tables, charts, narrative and formatting | the final professional conclusion |
| Review and correction | reviewer minutes, returns and rework reason | proof that review can be removed |
| Maintenance and coordination | template, rule and integration updates; chasing inputs and approvals | one-off setup |
| Waiting | blocked time for missing inputs, approvals or client responses | active production minutes |
| Setup, testing and training | initial configuration and adoption work | the routine cycle |
The Behavioural Insights Team's AI-assisted evidence-review case study reported 23% less total time in its comparison, alongside more revisions to the initial draft. Its authors state that the case cannot be generalised. It is a reason to count review and correction; it is not an estimate for another firm.
Completed fictional example
This example is fictional and does not represent a client result. A fictional advisory team produces a monthly operating review from permitted sources. One fictional cycle records 30 minutes of preparation, 40 of calculation, 35 of drafting, 25 of partner review, 20 of correction, and 15 of maintenance and coordination.
| Field | Completed fictional entry |
|---|---|
| Report and comparable case group | monthly operating review; standard client portfolio |
| Accepted output and reviewer | partner-ready report; engagement partner |
| Source version | approved source register, period 2026-09 |
| Active production minutes | 165 minutes using the formula above |
| Waiting | two days for an owner to confirm a changed definition |
| One-off cost | not recorded in the recurring total |
| Return reason | one changed definition required a correction |
| First route to test | source checklist and visible stop when a definition changes |
One report does not establish a typical baseline. The next step is to record three comparable cycles, including returned work. Do not infer saved cost, capacity or revenue from fewer minutes alone.
Choose the proportionate route
| If the constraint is… | Test this first | Consider a custom workflow only when… |
|---|---|---|
| inconsistent structure | an agreed template, definitions and reviewer checklist | approved sources need to be joined or checked repeatedly |
| an existing reporting product | its scheduled reporting, access controls and audit trail | it cannot preserve necessary source, permission or review evidence |
| repeated copying between approved systems | a narrow integration or ordinary automation | exceptions and ownership have explicit handling |
| professional judgement | a workpaper and review process | assistance can prepare bounded evidence without deciding the conclusion |
| incomplete or disputed inputs | a clearer intake and stop/return rule | conflicting evidence reliably routes to a named person |
Keeping the current method, improving a template or using existing software better are valid outcomes. A build should only proceed with known permissions, a reviewer, exception routes, acceptance checks and an operating owner. The NIST AI Risk Management Framework is a useful voluntary reference for thinking about risk and accountability; it does not prescribe this worksheet or establish an outcome.
Use the next reporting cycle
Download the measurement worksheet, complete it for three comparable reports, then look for the same return reason, missing source or manual transfer. For a neutral route comparison and four test cases, read existing reporting tools, better templates or a custom workflow?.
For a concrete reporting example, see client reporting and the existing report-production check. A free Firm Capacity Fit Call can begin with a plain-language description; no confidential client files are needed.