Is AI actually saving time on your monthly marketing reports?

A draft arriving sooner is useful. A finished monthly report still needs reliable figures, supported explanations and whatever review the team requires before it reaches its readers.

To measure AI reporting time savings, compare the human effort and turnaround required to finish reports of similar scope and quality. Include preparation, checking, corrections and approval. Record drafting time separately so you can see where a change occurred.

The worksheet accompanying this guide summarises that comparison from your own activity records. Start with a recurring report you already own. You do not need a claimed saving or a financial return to begin.

Choose the report you are comparing

Write down the report's purpose, reporting period, data sources, channels or properties covered, required analysis and recipients. Name the person who can accept it, and agree what acceptance means.

For a monthly marketing report, that might mean checked figures, consistent comparison periods, explanations supported by campaign activity, and recommendations that follow from the evidence. A polished summary with unresolved source questions would still need work.

Use the same start and finish rules for both workflows. For this exercise, start when report preparation begins, including collecting or chasing inputs. Finish when the required reviewer has accepted the report and it is ready for its intended distribution. If actual distribution matters, record that separately rather than silently changing the endpoint.

Keep a reference description of the current workflow, including any automation it already uses. Your baseline is the process you actually use. It need not be entirely manual.

Record each report's scope and workflow version before work begins. Compare reports that belong to the same scope group. A client presentation with additional analysis and an internal summary are different jobs.

Separate human effort from turnaround

Human effort

The combined active time people spend producing the report. Include the analyst, reviewer and anyone contributing necessary work.

Turnaround

The elapsed time between your agreed start and finish. It includes unattended processing, queues and waiting within that interval. Use full dates, times and a consistent time zone; calendar elapsed time includes nights and weekends. A business-hours measure needs its own explicit working calendar.

Add people's active minutes to measure effort. Subtract the report's start timestamp from its finish timestamp to measure turnaround. Do not calculate turnaround by adding activity durations: work can overlap.

Consider this teaching example. An analyst checks a source export while a colleague reviews the previous section. Both people's work counts towards effort, even though it happens at the same time. The report's elapsed clock passes through that interval once.

The analyst then requests a missing campaign explanation and moves to another project. The report is waiting, but that other project's work does not belong in this report's effort. If the analyst returns to chase the response, record the chasing as active work. An unattended AI run also uses elapsed time without necessarily using human effort; prompting, monitoring and checking do use effort when someone is actively doing them.

Keep waiting reasons for diagnosis. Overlapping waits cannot simply be added into a total delay.

Record work as it happens

The worksheet takes stage totals for each report. Keep the underlying activity notes separately in a shared log or your existing time-recording system. Use a common report identifier, with an internal label rather than confidential client details.

The activity record

Keep a record you can compare.

Establish the context before work begins. Then record the time and what happened at review.

01 / Context

Identify the work

Start with the reporting period, scope group and workflow version. Give each contributor a stable label and record their role, so every entry can be traced to the right person and comparison.

Assign each activity to one stage: preparation, drafting, checking, corrections, stakeholder revisions or release preparation.

02 / Time

Separate work from waiting

Record active minutes spent on this report. Split interrupted work into separate entries where practical, and never count the same person’s full overlapping interval against two activities.

Keep activity start and end timestamps, plus the report’s overall start and finish. These help reveal overlap; they do not automatically measure active work. Note waiting dependencies, such as an export, campaign explanation or reviewer availability, and record active chasing separately.

03 / Review

Preserve the outcome and the gaps

Record material changes, including corrections and changes in scope, and whether review accepted the report, requested revisions or left an issue unresolved.

Explain missing or estimated records. Leave an unknown duration blank. Enter zero only when you know no active effort occurred.

Use one stage per activity. If checking discovers a wrong comparison period, log the checking once, then put the work to fix it under corrections. Do not also add that correction time to the original checking entry.

Before entering stage totals in the worksheet, reconcile missing entries with contributors. Check timestamps, plausible active durations and simultaneous entries from the same person. Different people working together legitimately contribute separate effort. These checks require your activity notes; aggregate totals cannot reveal duplicated intervals.

Transfer the checked active minutes into the corresponding stages. Enter setup effort separately, then enter elapsed time calculated from the agreed report timestamps. Record acceptance and comparability honestly. If a stage total is incomplete, leave it blank and retain the reason in your notes. A partial total would understate the work.

Record setup and training separately from recurring production. That keeps the learning cost visible without making every later report appear to require the same setup again. The National AI Centre includes training, preparation and ongoing oversight in its guidance on assessing AI costs[1].

Keep review and changes in scope visible

Agree the quality checks before the comparison. Check figures against the intended sources, verify periods and metric definitions, and distinguish supported explanations from possible causes. Record material corrections, unresolved issues and whether the report passed its first review.

Compare what the records support

For comparable completed reports, calculate the difference in total human effort and, separately, turnaround. Look at stage totals to locate the change. A positive baseline-minus-current difference means less time on that measure; a negative difference means more.

Keep the underlying observations visible. Missing reviewer time makes an effort total incomplete, even if start and finish timestamps establish turnaround. An unaccepted report has no completed turnaround yet. Neither should appear as a finished saving.

Repeat the comparison across reporting cycles. Note changes in campaign complexity, source readiness, staffing, templates and reviewer availability. Where feasible, keep a comparable group using the existing workflow. Microsoft recommends baselines, comparison groups and quality measures in its value-measurement guidance[2]. This is vendor guidance, not proof of an effect in your team.

A before-and-after observation cannot isolate AI from everything else that changed. METR's developer research describes how task selection and concurrent work complicated its own measurement. That study concerns software developers, but its methodological limitations[3] are a useful reminder to examine who did the work and which tasks entered the comparison.

Decide what to change next

Use the pattern in the records to choose a specific action.

If effort is released, record what it is used for. Reduced labour time can create capacity without reducing payroll or other expenditure. The National AI Centre also recommends tracking where released time goes when assessing AI's value[1].

Start the next report with the scope, reviewer and timing boundaries already agreed. At acceptance, you should be able to explain which work changed, whether the report met the same standard, and what the team should adjust next.