A brand brief says to avoid discounts. A campaign request asks for a twenty-percent-off promotion. A product page still shows the full price. An AI advertising operator needs to do more than choose the newest sentence.
The conflict involves different kinds of information: a standing brand preference, a proposed campaign exception, and evidence about the offer currently available to customers. The right response is to identify the unresolved decision and its owner, then preserve useful work that does not depend on that decision.
Classify the statement before following it
Treat instructions, observations, proposals, and quoted research as different inputs.
An instruction comes from someone with authority over the relevant work. An observation describes a state, such as the current landing-page price. A proposal asks to change a state. Research describes external information that may help the team make a decision.
This classification prevents a competitor's example from becoming a direction to copy its claims. It also prevents an analyst's suggestion from being treated as an approved budget change.
The distinction has a security counterpart. The OWASP AI Agent Security Cheat Sheet identifies indirect prompt injection through external material and recommends boundaries around agent actions. In an advertising workflow, a webpage is evidence to inspect, not a source of permission over your account.
Use scope and ownership together
A practical precedence policy can describe four layers:
- Organization-level requirements that the campaign owner cannot waive casually.
- Account-level goals, spending authority, and exclusions.
- Campaign-specific instructions accepted by the appropriate owner.
- Task-level preferences that apply only to the current deliverable.
This is a proposed team policy, not a standard hierarchy imposed by ad platforms. The team must decide who can approve exceptions and which requirements need a different reviewer.
Recency helps only after authority and scope are clear. A newer comment from a reviewer who owns visual style does not necessarily override the budget owner's spending limit. A campaign-specific approval should not become a permanent rule for every campaign.
Work through a concrete promotion conflict
Consider an illustrative retailer with a standing preference for premium positioning and no routine discounts. A growth lead proposes a limited sale for a discontinued product. The product page has not yet been updated.
The operator should separate three questions. Is the sale an authorized exception to the brand preference? What are the exact eligible products, dates, and terms? Will the advertised destination present the same offer when the ad starts running?
While those questions are unresolved, the operator can prepare a creative concept that avoids a numerical discount, collect the approved product facts, and draft the required destination changes for review. It should not publish a discount claim merely because one message suggested it.
Once the owners approve the exception, record its scope and end condition. The exception expires with the promotion rather than silently rewriting the standing brand policy.
Keep a short conflict record
| Field | Useful detail |
|---|---|
| Conflicting statements | The exact relevant instructions or observations |
| Source and date | Who supplied each statement and when |
| Scope | Organization, account, campaign, or task |
| Decision owner | The person authorized to resolve this conflict |
| Dependent work | What cannot proceed until the decision is made |
| Independent work | What can continue now |
| Resolution | Accepted instruction, exception boundaries, and expiry |
The record should be short enough to review. It does not need an entire conversation transcript. Include the evidence necessary to explain the disagreement and the concrete decision required.
Ask a narrow question when authority is unclear
An operator that repeatedly asks “What should I do?” shifts too much work back to the team. It should explain the conflict and present the missing decision.
For example: “The standing brief excludes discounts, while the campaign request proposes twenty percent off Product A. The offer is not yet visible on the destination. I can prepare the creative, but publication depends on the campaign owner accepting the exception and confirming the live offer.”
This identifies the dependency without reopening decisions that are already settled. A budget owner who previously authorized the exact increase should not be asked to approve it again unless the relevant state or proposal changed.
Do not confuse account access with business authority
Software access is necessary for execution, but it is not the entire permission model. A user may technically have access to many accounts while the engagement covers only one.
The Google Ads access model describes access inherited through account hierarchies. Your business policy must still identify which of those accounts and actions the operator is authorized to use. A connection's technical reach should not determine the engagement's scope by accident.
Use the permissions matrix to establish that boundary. If a new instruction refers to a different account, resolve the identity and authority before preparing an executable change.
Update durable instructions after the decision
A resolved conflict should improve the next task. If the team decides that promotional exceptions require a named offer owner, record that policy where future campaign briefs can find it. If the decision was a one-time exception, preserve that limitation.
Avoid merging every comment into a growing list of equally authoritative instructions. Outdated promotions, rejected suggestions, and temporary workarounds should remain historical context rather than active rules.
During a handoff, include the current rules and the exceptions still in effect. For creative disputes, use a feedback workflow that distinguishes mandatory changes from preferences. The goal is a system that can explain why an instruction applies, what it permits, and when it stops applying.