The advertising lifecycle runs from a business question to a creative brief, an approved asset, a submitted ad, observed delivery, and an evaluated result. AI can help with several of those steps. It does not eliminate the need to distinguish them.
The most useful lifecycle model makes each transition explicit. “Created,” “approved,” and “live” should not be interchangeable labels. A team needs to know what evidence moved an ad into its current state, who owns the next decision, and what should happen if a step fails.
Begin with a defined job for the advertisement
Before production, state the customer situation, offer, intended action, and business outcome. Check that the destination can deliver the promised experience and that the business can handle demand. An accurate advertisement can still waste money if the product is unavailable or the sales team cannot respond.
Create a brief with current evidence and a version identifier. Link the brief to the campaign or experiment it is intended to support. If the purpose is to test an explanation, record the hypothesis and the comparison before reviewing performance.
This stage should end with a production-ready assignment, not a vague instruction to generate more content. Our AI creative brief guide describes the information that should survive each handoff.
Use states that correspond to observable events
The following is a proposed operating model. Platform status labels differ, so map each provider's actual statuses into the internal model without losing important detail.
| Internal state | Evidence required | Next decision |
|---|---|---|
| Brief ready | Approved assignment, offer, and source references | Begin production |
| Draft ready | Identified asset version and completed checks | Request internal review |
| Internally approved | Approval for that version, placement, and use | Submit within authorized scope |
| Submitted | Provider response and returned identifiers | Check provider review and errors |
| Eligible | Current provider status permits serving | Verify schedule, constraints, and delivery |
| Delivering | Observed impressions or other delivery evidence | Monitor the defined outcome window |
| Evaluated | Mature results with context and limitations | Continue, revise, retest, or retire |
| Retired | Verified stop or removal from intended use | Preserve records and check replacements |
Add an explicit unknown or failed state when evidence is unavailable. A missing readback should not be treated as success. Keep timestamps so an old observation is not displayed as a current fact.
Approve the complete customer promise
Internal review should inspect the rendered asset, offer, and destination together. Confirm supported claims, required qualifications, placement fit, and asset rights under the team's established process. An approved script may still produce a misleading crop, illustration, or subtitle.
Tie approval to a version and scope. If the price or destination changes, determine whether the prior approval still covers the result. Do not let a convenient filename hide a material edit. The creative operations board guide explains how to retain these relationships across concepts and adaptations.
Keep a record of rejected claims and their reasons. That prevents the same unsupported promise from reappearing in the next generated batch. It also makes the review process easier to improve without relying on one person's memory.
Treat provider review as a separate stage
Google's ad review documentation describes platform review and notes that edits can restart it. Internal approval does not bypass that process. Build launch plans around observed provider status rather than promising a guaranteed review time.
Inspect returned errors at the level where they occur. A campaign may exist while one asset or ad fails. Keep the successful and failed items distinguishable, and decide whether the intended launch can proceed with the remaining configuration.
Provider acceptance also does not prove delivery. Check status, schedule, budget, targeting, and other relevant constraints. When an ad has not served, report the actual evidence and investigate the cause instead of declaring the lifecycle complete because an API returned an identifier.
Verify execution without accidentally repeating it
A request can time out after the provider has accepted it. Before blindly repeating a creation or budget action, reconcile the requested change with current provider state. The correct recovery depends on the provider and operation; a generic retry policy cannot assume every mutation is safe to repeat.
Google's API partial failure guidance describes situations where valid operations can succeed while others fail in a supported request. A useful internal receipt should therefore preserve item-level outcomes rather than reducing a batch to one optimistic success message.
Maintain an agent change log connecting intent, authorization, request, provider response, and readback. This is a recommended evidence pattern, not a claim that every advertising platform exposes identical logs or controls.
Observe before interpreting
Once delivery begins, evaluate the agreed business outcome and diagnostic measures over a suitable period. Check conversion maturity, tracking health, inventory, promotion changes, and other events that affect comparability. A recent decline may not justify a creative replacement if the data is incomplete.
Keep a control ad when the test design requires one. Record changes that make it less comparable. If automated delivery reallocates spend, distinguish that observed allocation from evidence produced by a controlled experiment.
The result record should answer a specific question: what did this execution teach us about this message in this context? It can also say that the evidence is insufficient. Labeling an inconclusive test honestly helps the next team avoid treating a weak observation as a permanent creative rule.
Retire deliberately and preserve the lesson
Ads may need retirement because an offer expired, a product changed, evidence became outdated, or the team chose a replacement. Confirm the relevant provider action and any scheduled processes that could reactivate the old configuration.
Preserve the approved version, campaign identifiers, observation window, and decision rationale. Link a replacement to the prior asset so later analysis can reconstruct the change. Do not delete the history simply because the ad is no longer useful.
For urgent problems, use an advertising automation stop runbook with named owners and verified controls. An orderly lifecycle makes normal production easier, but it should also make an interruption understandable: what stopped, what remains active, and what evidence is needed before work resumes.
Carry the evidence between stages
Start with the creative production brief, then use the campaign launch checklist to verify the approved asset, destination and outcome path. Keep the same version references through production and release so a later review can identify what actually ran.