Evidence-Grounded AI Marketing Campaign Planner for Teams and Agencies

Create an evidence-grounded, execution-ready marketing campaign plan from the inputs and source materials below. Use Claude to synthesize the supplied brief, documents, data, and constraints; do not imply that Claude accessed systems, analytics, files, websites, or current market information that were not supplied or explicitly made available in the active Claude session.

Campaign inputs

Campaign goal: [Campaign goal]
Target audience segments: [Target audience segments]
Offer or product: [Offer or product]
Market and applicable requirements: [Market and applicable requirements]
Brand voice and claims rules: [Brand voice and claims rules]
Unique value proposition: [Unique value proposition]
Marketing channels: [Marketing channels]
Available assets and source evidence: [Available assets and source evidence]
Team roles and approval authority: [Team roles and approval authority]
Campaign timeline: [Campaign timeline]
Budget and capacity constraints: [Budget and capacity constraints]
KPI definitions and baselines: [KPI definitions and baselines]
Tracking and reporting setup: [Tracking and reporting setup]
Known risks and exclusions: [Known risks and exclusions]
Definition of done: [Definition of done]

Input contract

Treat the campaign goal, offer, audience, market, timeline, available capacity, publishing authority, and definition of done as blocking prerequisites. If any is absent, contradictory, or too vague to support responsible planning, ask a short set of prioritized clarification questions before producing the full plan. If answers are unavailable, provide only a bounded planning scaffold, label the affected decisions Blocked or Unknown, and state what evidence or decision is required.

Treat brand documentation, approved claims, customer research, historical performance, channel benchmarks, budget allocation, tracking specifications, and existing asset inventories as useful supporting context. Their absence does not always block planning, but it must reduce confidence and must not be concealed with invented facts or benchmarks.

Evidence and uncertainty rules

1. Classify material inputs and conclusions as one of: Supplied fact, Evidence-backed observation, Assumption, Hypothesis, Unknown, or Conflict.
2. For evidence-backed statements, identify the supplied source by file name, document section, dataset, report period, URL, or other available reference. Do not create citations or pretend to have reviewed absent material.
3. Keep audience facts separate from inferred personas. Label inferred motivations, objections, channel preferences, and buying triggers as hypotheses until supported by research or performance evidence.
4. Do not invent market size, competitor behavior, conversion rates, customer quotations, legal requirements, channel benchmarks, baselines, attribution results, or performance targets.
5. A target without a supplied baseline or benchmark must be labeled Proposed and include the rationale and validation method.
6. Surface conflicting sources rather than silently choosing one. Explain the campaign impact and name the owner who must resolve the conflict.
7. If web search or a connected data source is explicitly available and authorized in the current Claude session, distinguish retrieved evidence from user-supplied evidence and cite the source and retrieval date. Otherwise, state that external validation was not performed.

Authority, privacy, and action boundaries

Claude may analyze supplied materials, organize evidence, calculate from supplied data, draft concepts, compare options, and propose a campaign workflow. Claude must not claim to have contacted customers, queried live analytics, changed budgets, configured tracking, obtained consent, approved claims, scheduled assets, published content, launched ads, or measured results unless the action was actually performed through an authorized capability and supported by execution evidence.

All public-facing assets require the designated human approval authority. Legal, privacy, regulatory, product, or subject-matter review is mandatory when claims, promotions, testimonials, regulated products, personal data, intellectual property, accessibility obligations, or material customer promises are involved. Do not use raw personal or sensitive data when aggregated, minimized, or de-identified information is sufficient. Do not recommend sensitive-trait targeting, deceptive personalization, fabricated testimonials, dark patterns, or unsupported scarcity.

Stop and mark the relevant work Blocked if the plan depends on an unsupported material claim, unresolved consent or privacy issue, missing publishing authority, contradictory offer terms, unavailable measurement capability, or a timeline that cannot accommodate required review. Propose a safe resolution; do not bypass the control.

Planning workflow

1. Normalize the brief and build an input-and-evidence ledger. Identify the campaign objective, funnel stage, conversion action, scope, exclusions, decision owners, supplied evidence, assumptions, unknowns, and conflicts.
2. Translate the goal into a measurable objective. Connect the primary conversion action to the offer, audience need, campaign period, baseline, target, measurement source, and attribution limitations. If measurement is not currently possible, specify the instrumentation prerequisite.
3. Develop audience segments and provisional personas. For each, distinguish known demographic or firmographic attributes from inferred goals, pain points, triggers, objections, trust requirements, preferred channels, and content needs. Include an evidence reference and confidence level.
4. Establish positioning and messaging. Produce a positioning statement, value proposition, messaging pillars, rational proof points, emotional angles, objection responses, and calls to action. Build a claims matrix linking every material factual or comparative claim to evidence, permitted wording, restrictions, reviewer, and approval state. Exclude unsupported claims from publication-ready recommendations.
5. Select channels using explicit trade-offs. Assess audience fit, funnel purpose, creative requirements, reach or intent, cost constraints, team capacity, measurement readiness, platform-policy exposure, and dependencies. Explain why each selected channel is included and why plausible alternatives were deferred.
6. Design the asset system. Define campaign anchor assets, channel adaptations, landing-page sections, email sequence roles, social concepts, video concepts, advertising variations, lead magnets, retargeting messages, and sales-enablement materials as appropriate. For each asset, specify its segment, funnel purpose, message, proof required, CTA, source asset, owner, review path, and repurposing limits.
7. Map AI and human responsibilities from brief to reporting. Cover research synthesis, ideation, drafting, creative variation, factual checking, design, brand review, legal or compliance review, approval, scheduling, publication, monitoring, analysis, and repurposing. State what Claude can propose, what a human must inspect, and who has final authority.
8. Build a dependency-aware calendar that fits the supplied timeline and capacity. Include briefing, production, review, revision, tracking validation, launch readiness, publication, monitoring, optimization review, reporting, and repurposing. Do not schedule publication before prerequisite approvals and tracking checks.
9. Define measurement and decision rules. For every KPI, include its formula, funnel stage, baseline, proposed target, data source, tracking owner, reporting cadence, attribution caveat, and the decision triggered by underperformance or overperformance. Do not treat proxy engagement metrics as conversions.
10. Assess operational and marketing risks. Include brand inconsistency, unsupported or misleading claims, legal or platform-policy exposure, privacy and consent, intellectual-property issues, accessibility, audience mismatch, model-generated inaccuracies, review bottlenecks, missed dependencies, budget overrun, tracking failure, channel underperformance, and reputational harm. Assign likelihood, impact, preventive control, detection signal, mitigation, contingency, owner, and residual risk.
11. Reconcile the complete plan against budget, capacity, dates, evidence, approvals, and measurement readiness. Reduce scope or mark work Blocked where constraints cannot be reconciled; do not hide infeasibility.

Required deliverable

Produce the following sections in order:

1. Campaign status and executive brief
State whether the result is Ready for human review, Partially planned, or Blocked. Summarize the objective, audience, offer, conversion action, strategic approach, main constraints, unresolved decisions, and immediate human actions. Make clear that this is a proposed plan, not proof of launch or approval.

2. Input and evidence ledger
Use columns: Item | Classification | Supplied value or observation | Source reference | Confidence | Conflict or limitation | Campaign impact | Required action or owner.

3. Strategy decision record
Use columns: Decision | Selected approach | Alternatives considered | Evidence and rationale | Trade-off | Assumption or dependency | Decision owner | Status.

4. Audience segment and persona matrix
Use columns: Segment or persona | Known attributes | Goals and jobs | Pain points | Trigger | Objection | Trust requirement | Channel preference | Message angle | Content need | Evidence reference | Confidence | Validation method.

5. Positioning, messaging, and claims controls
Provide the positioning statement, value proposition, messaging pillars, emotional angles, proof points, objection responses, and CTA options. Then use a claims matrix with columns: Proposed claim | Claim type | Evidence source | Permitted wording | Prohibited or risky wording | Market limitation | Required reviewer | Approval evidence | Status.

6. Channel portfolio and rationale
Use columns: Channel | Segment | Funnel purpose | Conversion path | Content format | Cadence | Capacity or budget assumption | AI contribution | Human owner | KPI | Tracking source | Dependency | Selection rationale | Status.

7. Campaign asset backlog
Use columns: Priority | Asset | Segment | Funnel purpose | Core message | Proof required | CTA | Source material | Owner | Reviewer | Due date | Approval gate | Repurposing rule | Status. Include concepts and briefs, not fabricated production or publication claims.

8. AI and human responsibility map
Use columns: Workflow stage | Required input | Claude may assist with | Claude must not claim or do | Human responsible | Required evidence | Approval gate | Failure or escalation path.

9. Approval gate checklist
Cover factual accuracy, evidence-backed claims, offer and pricing accuracy, brand voice, legal or regulatory review, privacy and consent, intellectual property, testimonial permission, accessibility, visual quality, CTA and destination integrity, channel and platform fit, tracking readiness, final publishing authority, and post-publication monitoring ownership. Use columns: Check | Acceptance criterion | Reviewer | Evidence required | Actual observation from supplied material | Status | Blocking issue.

10. Capacity-aware campaign calendar
Use columns: Date or period | Task | Owner | Channel | Asset | Dependency | Effort or cost assumption | Review deadline | Approval gate | Publication state | Monitoring action | Repurposing opportunity | Status. Use Proposed as the default state unless execution evidence supports another state.

11. KPI, instrumentation, and decision dashboard
Use columns: Objective | KPI | Formula | Baseline | Target | Target basis | Data source | Tracking requirement | Attribution limitation | Reporting cadence | Owner | Decision threshold | Action if threshold is crossed | Verification status.

12. Risk and control register
Use columns: Risk | Failure mode | Likelihood | Impact | Preventive control | Detection evidence | Mitigation | Contingency or recovery action | Owner | Residual risk | Status.

13. Verification and acceptance report
Test the plan using these concrete checks:
- The objective, audience, offer, primary message, CTA, and KPI form one traceable conversion path.
- Every material claim has a valid supplied evidence reference or is excluded, qualified, or blocked.
- Every channel has a distinct audience, funnel purpose, accountable owner, feasible asset requirement, KPI, and tracking source.
- Every public asset has a named human reviewer and publishing authority.
- Calendar effort, review time, budget assumptions, dependencies, and deadlines reconcile with available capacity.
- KPI formulas, baselines, targets, sources, and attribution caveats are explicit; unavailable instrumentation is identified.
- Privacy, consent, intellectual-property, accessibility, legal, brand, and platform-policy checks are routed to appropriate humans.
- Assumptions, unknowns, conflicts, and blocked items remain visible in the handoff.

Use columns: Acceptance check | Expected condition | Actual observation | Evidence reference | Status as Pass, Fail, Blocked, or Not assessed | Remediation | Owner. A Pass must cite observable support in the produced plan or supplied evidence.

14. Handoff and next decisions
List prioritized decisions, evidence requests, owner assignments, approval requests, tracking work, and launch prerequisites. Separate Ready for review, Blocked, and Optional optimization items.

Completion-claim rules

Use only these states unless execution evidence supports a more specific one: Proposed, Drafted from supplied inputs, Ready for human review, Blocked, Not assessed, or Verified from supplied evidence. Use Approved only when identifiable approval evidence is supplied. Use Published, Launched, Sent, Configured, Tested, or Measured only when corresponding execution logs, system output, dated screenshots, approval records, or exported results are available. Otherwise describe the work as recommended or pending.

Before finalizing, reconcile all section totals, owners, dates, statuses, dependencies, evidence references, and KPI definitions. Do not silently convert an assumption into a fact or a proposed action into completed work.