A marketing director we worked with told us the truth about her first year using AI tools: "I have a thousand drafts and not one approved campaign brief." The drafts were everywhere. The bottleneck was somewhere else — the meeting where the team aligned on the angle, the approvals, the brand-voice review.
The campaign-brief copilot is the AI employee whose entire job is to compress that bottleneck. Not by skipping the human approvals. By feeding them better drafts, faster, with the inputs already organised.
The shape of the role
Title. Marketing Operations AI — Campaign Brief Specialist.
Mission. Convert raw inputs (meeting transcripts, Slack threads, executive priorities, performance data) into approved campaign briefs at 5× the team's current pace.
Outcomes. Brief turnaround time, brief approval rate on first review, campaign performance correlation with brief specificity.
Reports to. Head of Marketing or Director of Brand.
Tools. Meeting-transcript ingest, brand-voice eval, brief template library, calendar/scheduling, CRM integration, performance-data read access.
Boundaries. Drafts only. Doesn't approve, doesn't publish, doesn't make creative-direction calls.
The four-stage workflow
1. Intake. The kickoff meeting transcript, the Slack thread, the priority docs, the prior-campaign performance data — everything goes in. The agent assembles a structured digest: stated objectives, target audience, channels under consideration, key deadlines, constraints (budget, brand, regulatory), unresolved questions.
2. Draft. Using the digest plus the brief template the team has standardised on, the agent drafts the brief: positioning, audience, key messages, channel mix, asset list, success metrics, timeline. Variants where multiple paths are reasonable.
3. Review loop. The marketing director reviews. Edits. Each edit pair (original vs. final) feeds the eval set. The agent gets sharper at the team's specific style, the team's specific brand, the team's specific constraints. Within a quarter, first-review approval rate climbs from 40% to 85%.
4. Approval routing. Final brief routes through whatever approval flow the org needs (legal, brand, exec, channel). The agent tracks status and prompts on slipped approvals.
The reviewer loop is the system
Most AI marketing tools fail at the second item — the review loop that sharpens the agent. Without it, the agent generates the same kind of output forever, and the marketing director spends quarter two doing what she was doing in quarter one.
With it, the agent is a teammate that learns the brand by getting the brand wrong and being corrected. After 6 months, the agent's first drafts are better than the team's average mid-level marketer's first drafts.
The mechanic that makes this work: every edit to the brief gets logged with context (what was original, what was final, why if the editor noted). Weekly, the team reviews the edit log for patterns: "we keep softening the urgency language" or "we keep replacing our default CTA." Those patterns become updates to the agent's prompt and eval set.
What this saves
A typical mid-market team:
- Spends 6-10 hours per brief, end to end.
- Lands 60-70% first-review approval.
- Manages 4-8 active campaigns at any time.
With a working campaign-brief copilot:
- Time per brief drops to 1.5-3 hours of human time (the agent's draft is the heavy lift).
- First-review approval climbs to 80%+ within two quarters.
- Active campaign capacity expands to 10-15 without adding headcount.
The marketing director's day shifts from drafting to deciding. That's the trade. Decision-making is what the team should be doing.
What we won't ship
Auto-approval. Briefs are approved by humans. Always.
Creative direction. The agent drafts. The director picks the angle.
Channel-specific copy without channel-specific eval. A campaign brief is not channel copy. The agent drafts the brief; channel copy is a downstream task with its own discipline.
The KPIs the manager watches
- Brief turnaround time (target: 50%+ reduction within 2 quarters).
- First-review approval rate (target: 80%+ by quarter 2).
- Edits per brief (should decline as the eval sharpens).
- Brand-voice eval pass rate (should be 100% before reaching the director).
If 2 and 3 don't move over a quarter, the eval feedback loop is broken. Investigate before scaling.
How to start
Pick one campaign type — email nurture, product launch, holiday promo. Run the copilot on that one type for 60 days. Build the eval set. Once the metrics move, expand to a second type.
Most teams try to deploy across all campaign types at once. The ones that ship pick one and master it.
Close
The campaign-brief copilot is a marketing employee whose job is the unglamorous middle of the work — the part nobody enjoys but everyone notices when it's slow. Build it as a teammate. Manage it as a teammate. Review its outputs as a teammate. The lift on team capacity is what makes the headcount math work.
Related reading
- An AI employee isn't a bot — the framing for this whole series.
- Agents in marketing: brand voice — the eval discipline that powers the copilot.
- LLM evals are restaurant health inspections — the discipline that catches drift.
We build AI-enabled software and help businesses put AI to work. If you're hiring an AI marketer, we'd love to hear about it. Get in touch.