Skip to content
Get MCP AdsGet MCP Ads, home
Start free
View as Markdown
Deep contractGoogle AdsMeta Ads

How to prepare an agency client review with an AI agent

The monthly client pack for one account: results against target, what the agency changed and why, what comes next, every number traceable.

IntentAssemble a client-ready narrative where every figure survives the client's own dashboard check.

4
Steps
8
Tools
3
Rules
3
Failure modes

When to use it

Use it when

  • Monthly or quarterly client reviews.
  • Renewal conversations, where the changes-made list is the value story.

Do not use it when

  • Daily reporting: this is a narrative pack, not a dashboard.
  • Accounts the agency has run for under a month: there is no story yet, say so rather than pad.

What it needs first

Preconditions

  • Only this client's accounts ticked in the workspace: scoping is the confidentiality mechanism.
  • The client's target (CPA, ROAS, volume) stated, not assumed.

Inputs

period
The review period.Default: last full month.
targetrequired
The number the client judges the account by.

Procedure

In order. Every tool named is one the gateway ships, and links to its reference.

  1. Pull the period's results per platform, and the previous period for context.

  2. Pull the change log for the period: what the agency actually did.

    The changes-made list is the part clients cannot get from their own dashboard.

  3. Add the site view if GA4 is connected, and organic context if Search Console is.

  4. Assemble: results vs target, work done, what it changed, next period's plan.

Decision rules

The observable condition, what it lets you conclude, and what takes the conclusion back.

  • When

    The period includes a change whose effect is not yet measurable

    Conclude

    Present it as work done with results pending, dated; do not claim early credit.

  • When

    Results missed target

    Conclude

    Lead with the miss and its diagnosed cause; a pack that buries the miss costs the renewal later.

    Unless

    The cause is still unknown; then say what is being investigated and by when.

  • When

    A platform's self-reported win disagrees with GA4

    Conclude

    Show the client both numbers with the mechanism; agencies lose accounts over discovered discrepancies, not disclosed ones.

Evidence every conclusion must carry

  • Every headline figure carries platform, window and attribution basis.
  • Every claimed action exists in the change log with its date.

Where the agent stops

  • The pack is a draft for the account lead: positioning, tone and commitments are theirs.
  • Nothing in preparing the review writes to any account.

How it goes wrong quietly

The cases where the analysis is wrong and still looks right. Read them before trusting a number.

  • Cross-client leakage: a workspace with several clients' accounts ticked can mix data; verify scope before the first call.
  • Claiming platform-attributed revenue as business revenue inflates the story and detonates at the CFO's desk.
  • Month lengths differ; compare daily rates or same-length windows, not raw month totals.

What the answer contains

01Results vs target
The number, the target, the gap, the trend.
02Work done
Dated changes from the log, each tied to intent.
03Next period
Planned moves, expected effect, measurement plan.