AI Marketing Leverage Rule
Use AI to scale execution while reserving judgment and curation for skilled marketers.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 7
- Confidence
- 99%
The rule positions AI as leverage applied to an existing marketing system. The team first defines the business outcome and identifies which parts of the workflow are routine enough to automate. Skilled marketers continue to supply taste: selecting the direction, providing context, curating alternatives, correcting hallucinations, and editing for nuance. Performance is judged by audience response and business quality rather than the number of assets generated. Campbell predicts asymmetric benefits: capable marketers can use AI to become substantially better, weaker marketers may improve only marginally, and lazy operators can create far more low-quality material. The practical implication is to automate production without outsourcing judgment. If an AI-assisted process cannot preserve factual accuracy, distinctiveness, and brand fit under human review, it should not be scaled merely because it is faster.
Origin
Patrick Campbell and Kieran Flanagan debated AI's effect on marketing quality, automation, hallucination, and human taste on Marketing Against The Grain.
Core principles
- 01AI is a leverage tool rather than a complete marketing strategy.
- 02Automation amplifies the quality of the judgment directing it.
- 03Strong output still requires curation, nuance, and editing.
- 04Low-taste channels are most vulnerable to commoditization.
- 05Human review is essential where hallucination or brand risk matters.
How to run it
- 1
Start With the Outcome
Define the customer or business result the workflow must improve. Do not begin with a mandate to use AI somewhere.
Pro tip Choose a workflow with a measurable baseline for speed and quality.
Watch out Tool-first adoption creates activity without strategic relevance.
- 2
Map Judgment and Production
Break the workflow into research, decisions, drafting, production, verification, and distribution. Mark which tasks are repetitive and which require taste or accountability.
Pro tip Automate transformations before decisions when uncertainty is high.
Watch out A task that looks repetitive may still contain hidden brand or factual judgment.
- 3
Supply Context and Standards
Give the system relevant facts, constraints, examples, voice guidance, and the intended audience. Define what unacceptable output looks like.
Pro tip Use approved examples to make quality concrete.
Watch out Generic prompts encourage generic output.
- 4
Generate With Bounded Scope
Use AI for drafts, variants, summaries, or personalization within the mapped boundaries. Keep consequential decisions outside the autonomous path.
Pro tip Ask for alternatives that expose different strategic choices rather than superficial rewrites.
Watch out Personalization at scale can also scale hallucinations.
- 5
Curate and Verify
Have a capable marketer check facts, nuance, differentiation, audience fit, and brand risk. Edit or reject output rather than accepting it by default.
Pro tip Review samples more heavily when a workflow or model changes.
Watch out Fluent language is not evidence of factual or strategic quality.
- 6
Measure Quality-Adjusted Leverage
Compare speed, cost, response, errors, and downstream outcomes with the prior process. Scale only when the total result improves.
Pro tip Track correction time so apparent production gains are not overstated.
Watch out Volume metrics can conceal lower trust, engagement, or conversion.
- 7
Reinvest the Gain
Use recovered capacity for deeper customer knowledge, more experiments, stronger concepts, or improved distribution. Let AI create room for higher-value marketing work.
Pro tip Set an explicit destination for saved time before automating.
Watch out If all saved capacity becomes more generic output, the team may increase noise without increasing value.
In the wild
A company uses LinkedIn information to draft personalized outbound emails at scale. A marketer reviews factual claims, removes hallucinated details, and adjusts the message to match the recipient and brand before sending.
→ The team gains production speed without delegating truthfulness or judgment to the model.
A content lead supplies approved examples and asks AI for ten headline directions. She rejects generic options, combines two promising concepts, verifies every factual claim, and tests the final variants against qualified engagement.
→ AI expands the option set while the marketer remains accountable for taste and performance.
Common mistakes
Calling AI the Strategy
AI describes a capability or tool layer; it does not define the audience, positioning, value, or business outcome.
Scaling Without Curation
Fast generation multiplies hallucinations, generic language, and poor strategic choices when skilled review is absent.
Measuring Output Volume
More emails, articles, or variants are not evidence of better marketing unless audience and business outcomes improve.
Is it for you?
Best for
It is best for marketing teams adopting generative AI for research, drafting, personalization, production, or operational acceleration.
Not ideal for
It is not ideal for autonomous high-stakes communications where factual, legal, reputational, or relational errors cannot be safely reviewed.
From the transcript
“I think it's just a tool. It's just a tool that is gonna make better marketers even better and worse marketers marginally better.”
“I think AI hallucinates a lot. That's one of the largest problems with it, right? And so you actually do need to have a human…”
“Is like AI is not a strategy, right?”
From the episode
$200M Founder Reveals Why Ai Can’t Replace Top 10% Of Marketers (#135)