How Marketing Managers Build an AI Workflow That Compounds
For Marketing managers · Based on Jeff Su AI From Scratch 3-Level System
// TL;DR
Marketing managers can use the Jeff Su 3-Level AI System to convert scattered ChatGPT use into a compounding workflow. Commit to one chatbot (prioritising paid access from your employer), master the OC Framework, then build Projects for your recurring streams — weekly status updates, campaign reporting, and competitive research. Store approved past reports as markdown knowledge files, run the Reconcile Move after every edit so outputs match your voice, and eventually migrate to an AI System that cross-references campaign spend, calendar deadlines, and email threads simultaneously. The result: fewer repeated corrections and reports that get sharper over time.
Why do marketing managers need a structured AI system?
Most marketing managers use AI ad-hoc — a prompt here for a subject line, a prompt there for a report intro — and reset context every session. That means re-explaining brand guidelines, tone, and format constantly, and outputs never improve. The Jeff Su 3-Level System fixes this by treating AI as a compounding asset rather than a one-off tool. You go deep on one chatbot, supply the right context, and build Projects that remember your recurring work so quality climbs week over week.
Which chatbot should a marketing manager commit to?
Apply the three selection principles in order. First, Paid Tier Priority: if your employer provides paid Gemini but you only have free ChatGPT at home, use Gemini — the free-to-paid gap is night and day. Second, match to work type: Gemini is strong for heavy Google Workspace use and mixed media, ChatGPT for research and web search, Claude for writing and design. Third, vibes — the one you enjoy is the one you'll actually use. Then commit; don't split attention across tools, because skills only compound when you go deep on one.
How do you set up Projects for marketing work streams?
Identify your 2-3 recurring streams and build a Project (or Gemini Gem) for each:
- Weekly Status Updates — Project Instructions with constraints like 'always under 200 words, bullet format'; Knowledge Files containing your last 3 approved updates as markdown.
- Campaign Reporting — brand guidelines and past reports as knowledge files, plus instructions on required sections and metrics.
- Competitive Research — named frameworks (e.g. 'analyse using Jobs-to-be-Done') and source documents like competitor press releases and earnings calls pasted as real examples.
Always select the most powerful model in settings — companies default you to the weakest one, and the capable models catch nuances that matter for positioning and analysis.
How do you make reports sound like you wrote them?
Use the Reconcile Move. When the AI drafts a campaign report, edit it to match your voice and standards, then paste your final version back with: 'Reconcile my final version with your initial draft and propose rules to remember for next time.' The AI dissects every change, extracts the rules behind them, and applies them going forward. Do this for a few cycles and you'll stop giving the same corrections — the system learns your tightening of hype, your metric framing, your executive-summary style.
When should you scale to a full AI System?
Once all three Projects are running, migrate to an AI System — for most marketing managers, Gemini Spark is the right non-technical tier. It's pre-connected to Gmail, Calendar, and Drive, so it can cross-reference campaign spend data, calendar deadlines, and email threads at once. This unlocks two things Projects can't: synthesising insight across silos (spotting that a campaign deadline collides with a reporting cycle) and self-updating rules from your feedback.
What's the fastest way to start today?
Pick your one chatbot using the paid-access rule, switch to the most powerful model, and build a single Project for the report you write most often — loading two approved past reports as markdown. Run one draft, edit it, and do the Reconcile Move. That single loop will show you the compounding payoff before you scale to the full system.
// FREQUENTLY ASKED QUESTIONS
Which AI tool is best for marketing managers?
Whichever you have paid access to first — often Gemini through Google Workspace. For work type, Gemini suits Workspace-heavy and mixed-media marketing, ChatGPT suits research and web search, and Claude suits long-form writing and design. Commit to one rather than splitting focus, since fluency on one transfers to the others.
How do I get AI to match my brand voice in reports?
Load 2-3 approved past reports as markdown knowledge files inside a Campaign Reporting Project, then run the Reconcile Move after each draft: paste your edited final version back and ask the AI to extract rules from your changes. Over a few cycles it learns your tone and stops repeating the same mistakes.
Can AI handle competitive research reliably?
Yes, when you supply real context. Paste competitor press releases and earnings calls as examples and name a framework like Jobs-to-be-Done in your prompt. Named frameworks carry more context than long descriptions, and real source documents beat asking the AI to recall from memory.