Workflow example / 01

A client report draft you can actually review.

Start with approved, aggregate performance data. Draft a concise narrative and a list of questions. Keep data checks, client interpretation, approval, and sending in human hands.

Download the starter kit PDF Take the five-minute assessment

The sample below is fictional. This public guide is not a checkout or order-delivery verification.

The use case

Small input. Accountable output.

A repeatable client update often starts with the same metrics and handoffs. This workflow organizes those inputs and drafts plain-language context without claiming to determine why a metric moved.

Required tools

A spreadsheet or CSV export you are authorized to use
An AI assistant your agency and client approve for this data and use case
Your existing report format and a human reviewer who knows the metrics

Steps are manual and tool-agnostic; compatibility with a specific vendor, plan, connector, or privacy setting is not implied. Confirm terms, controls, and client permission before use.

Sample input / fictional

Period: April (fictional)
Compare: March (fictional)
Channel: Paid search
Leads: 160 vs 142; target 150
Spend: $4,000 vs $3,700
Context: Ad copy changed mid-period.
Cause: not verified.
Caveat: lead quality not reviewed.

No client name, account ID, or real client data. The numbers are invented for demonstration.

Sample output / draft

Paid search recorded 160 leads, up 18 from March (+12.7%) and above the supplied target. Spend rose 8.1%. Cost per lead was $25.00 vs. $26.06, down about 4.1%.

The input does not establish that the ad-copy change caused the movement. Lead quality has not been reviewed. Confirm source totals and calculations; add verified context before sending.

Human review required

Setup

A six-step pilot.

01Choose one recurring report and note the audience, cadence, assembly time, required metrics, and approver.
02Write down each metric's definition, unit, period, comparison period, source, and target.
03Remove client identifiers and unnecessary detail. Use approved aggregate data only; if permission is unclear, test with fictional data.
04Ask for a clearly labeled first draft. Keep the existing report handoff; do not auto-send.
05Independently check every calculation and claim, resolve open questions, and get a named person's approval.
06Track assembly time, review time, and corrections for three cycles before deciding whether to keep the workflow.

Human review

Before a report goes out.

Generated text can be wrong or incomplete. A person remains accountable for source checks, interpretation, approval, and delivery.

Source totals, periods, units, filters, and metric definitions match.
Changes, percentage denominators, rounding, and cost-per-result math are correct.
No unsupported causes, benchmarks, attribution, strategy, or promises remain.
Tone, client context, and any contract-specific language are appropriate.
No confidential or unapproved data was entered; every flagged question is resolved or stated.
A named person approved the final report before it is shared.

Limitations + safe use

Useful does not mean automatic.

No live integrations, metric validation, attribution, strategy, approval, or auto-send are included. Never enter passwords, API keys, personal data, row-level records, or confidential client details into an unapproved tool. If a figure cannot be traced or a claim cannot be supported, remove it or ask the account lead.

Delivery + use

Keep the guide with your process.

Download this PDF, retain a team copy, and use the template with data you are authorized to process. If you receive a kit through a storefront purchase, follow the download or library instructions displayed by that storefront. This guide's availability does not verify checkout or order delivery.

Get the printable guide