Ad Copilot — scope, goals & phased plan

The zero-to-one scope for an AI co-pilot that generates ad creative, monitors live performance, and routes every AI action through confidence-based human review.

Owner Product · Status Draft v1.0 · Thesis the product's value is the trust boundary it makes explicit, not the AI itself.

Goals / Non-goals

Goals ≥40% suggestion acceptance by beta month 3 · incremental ROAS as north-star · visible audit trail on every budget action · LLM cost capped as % of managed spend

Non-goals (v1) No autonomous spend above threshold · no custom attribution engine · text-only creative · English-only · hourly refresh, not real-time

Personas

PrimaryPerformance marketer$50K–$500K/mo, team of 1–4. Distrusts black-box automation; won't adopt anything spending money with no paper trail.

SecondaryMarketing leadCares about incremental ROAS and LLM cost % of spend — not in-product engagement metrics.

SupportAdmin / complianceNeeds brand-safety filtering, configurable lane thresholds, immutable audit history.

Solution — three capabilities

1. Creative generation → 2. Performance monitoring → 3. Human review queue

Functional requirements 15 total, representative sample

FR-1	Generate 3–5 ad copy variants per ad set from product feed + historical top performers.
FR-8	Classify every AI action into exactly one lane: Auto-apply, Suggest, or Escalate.
FR-9	New / low-data accounts cannot land in Auto-apply regardless of model confidence.
FR-11	Immutable audit log of all AI actions; rollback for Auto-apply within 5 minutes.
FR-13	Enforce per-account monthly LLM generation caps before the model call is made.

Phased scope

Phase 1 · Meta, Suggest-only → Phase 2 · +Google, +budget/bid → Phase 3 · Opt-in Auto-apply