AI agent team guide
AI Ad Manager: Run a Weekly Performance Review
A weekly paid-performance decision register that separates tracking gaps from creative and landing-page decisions

An AI ad manager is most useful when it explains what changed, which numbers can be compared and what deserves the next test. Give the team one complete reporting window, a clear conversion definition and the campaign context. The result should be a short performance brief with one proposed action, not a confident instruction to increase every budget.
Use this desk when paid results, website behavior and commercial outcomes live in different reports. Ad Manager leads the campaign read; the other coworkers check comparability, inspect the destination and preserve the decision. This is a review-and-planning workflow. It does not turn importing templates into permission to launch ads or spend.
Build your four-agent team
| Agent template | Its job on the team | Bring to its setup | Reviewable return | Team |
|---|---|---|---|---|
Ad ManagerLead agent | Own the campaign read and proposed test. | Campaign exports, the agreed conversion and complete comparison periods. | A campaign table, material exceptions and one proposed test with its budget assumption. | Add to team |
Metric Change InvestigatorSupporting agent | Check whether the apparent movement is comparable. | Metric definitions, attribution windows, event changes and coverage notes. | Recalculated changes, definition breaks and competing explanations. | Add to team |
Landing Page Conversion ReviewerSupporting agent | Inspect what happens after the ad click. | The actual creative, landing page and verified offer. | A promise-to-page review and one bounded conversion hypothesis. | Add to team |
Decision Journal KeeperSupporting agent | Preserve what the team decides to test. | The accepted evidence, alternatives and budget owner’s decision. | A dated decision draft with success criteria, stop conditions and a review trigger. | Add to team |
Set up your team in Finta
- Add each template individually. Use the Add to team action for each named coworker, review the import setup and complete its Getting Started skill.
- Give each role the right context. State the shared objective, then supply the authorized documents, notes or exports that role needs. Do not assume an unconnected system is available.
- Review the handoff. Accept or correct one return before giving its evidence and open questions to the next coworker. Importing four templates does not itself run a bulk team.
- Make recurrence deliberate. Complete a useful manual pass first. Review any supported routine and enable it only when the scope, inputs and permissions are right.
Prepare the evidence
- A campaign-level export with spend, currency, dates and the exact event counted as a result.
- A matching website report with eligible visitors and completed outcomes kept separate.
- Current creative, the destination page and any offer or tracking changes.
- The previous test decision, including what actually ran and any delivery limits.
Run the workflow
1. Agree the question before sorting winners
Choose a decision such as whether to keep observing one audience or prepare a landing-page test. Ask the metric investigator to confirm the same currency, conversion event, attribution window and complete dates across reports. A tracking change belongs in the brief before a growth percentage.
2. Build the paid-to-page bridge
Have Ad Manager report campaign spend and the defined result. Give the landing reviewer the creative and its real destination. Preserve platform-reported conversions separately from completed attributed signups in your own analytics; do not silently choose whichever is larger.
3. Choose one explanation to investigate
A weak result may reflect audience fit, creative, destination, measurement or ordinary variation. Ask for the evidence that would distinguish the explanations. The next test should change a bounded part of the journey, not audience, offer, page and budget simultaneously.
4. Hand the proposal to its actual owner
Review the test cost and the authority required to change the campaign. Decision Journal Keeper records the accepted version and review trigger. Approval of a proposal is not evidence that an ad platform applied it; check the actual campaign state after any authorized execution.
A fictional review that resists the cheapest-click trap
These are invented teaching inputs, not Finta customer results or advertising benchmarks. Both campaigns use the same complete period and currency.
| Evidence | What it establishes | Next review |
|---|---|---|
| Campaign A: $600 spend and 20 platform-reported signup events | $30 per platform-reported event, not necessarily $30 per verified new account. | Reconcile the event definition and attribution window. |
| Campaign B: $400 spend and 10 platform-reported signup events | $40 per reported event. The lower spend does not make this the better campaign. | Check destination and audience differences before comparing. |
| Website report: 24 completed attributed signups in total | This is a different measurement source from the 30 reported ad events. | Keep the six-event difference unresolved until definitions are reconciled. |
| Landing page omits the worksheet promised in A’s creative | A verified promise mismatch, not proof of the conversion effect. | Propose one destination correction for the owner to review. |
The team can prepare a useful test without declaring a winner. Preserve the reporting gap and inspect the corrected journey before drawing a performance conclusion.
Copy the team brief
Prepare one paid-performance review for [decision and budget owner].
Use [campaign exports], [metric dictionary], [complete comparison periods], [website outcome report] and [creative/destination].
Ad Manager: return spend and the precisely defined result per campaign.
Metric Change Investigator: reconcile definitions, currency, windows, coverage and tracking changes.
Landing Page Conversion Reviewer: compare the promise with the actual page and next action.
Decision Journal Keeper: draft one proposed test, alternatives, approval owner, stop condition and review trigger.
Show calculations. Keep platform conversions, completed attributed signups and unknown attribution separate.
Do not infer causes, launch ads, change budgets or enable a routine.Review the return
- Each percentage and cost uses a visible denominator and matching period.
- The proposed test addresses one evidenced question and has an actual decision owner.
- The saved brief distinguishes proposed, approved, applied and reviewed states.
Questions before you start
Will these agents manage my advertising account automatically?
Not simply because you import them. This guide uses authorized reports or supplied exports to prepare a review. Any connected advertising tool, permitted action and budget change must be configured and approved through the controls actually available to you.
What if conversions disagree across tools?
Keep both definitions and counts. Check attribution windows, time zones, event deduplication and eligibility before comparing. An unexplained gap is a measurement task, not permission to report a more flattering total.
Sources and next steps
This playbook uses the current published Finta marketplace packages linked in the team table. Open each profile to review its instructions, context, Skills and any routines before importing.