A Three-Variable Equation Determines Whether AI Helps or Hurts Your Marketing

But here's the stat nobody puts in their LinkedIn post: the share of marketers who can actually prove ROI from AI dropped from 49% to 41% over the same period.

More adoption. Less proof it works. That's not a technology problem. It's a delegation problem. Most businesses start in the same place with AI: writing. Ad copy, social posts, blog content, email sequences. It feels natural because writing is slow and AI is fast. But speed and quality are different things. And the quality question follows a precise mathematical pattern that Wharton professor Ethan Mollick calls the Equation of Agentic Work.

This equation doesn't just apply to coding or consulting. It explains exactly why some marketing tasks generate massive returns when handed to AI, and why others degrade your results the moment you stop doing them yourself. We've argued before that most businesses automate the wrong half of marketing. The equation that follows explains precisely where the line falls.

The Equation That Governs AI Delegation

• Human Baseline Time: how long the task takes you to do manually
• Probability of Success: how likely the AI is to produce an acceptable result on each attempt
• AI Process Time: how long it takes to write the prompt, wait for output, and evaluate the result

With GPT-5.2, expert judges rated AI output as equal to or better than human professionals 72% of the time. At that hit rate, on a 7-hour task with 1-hour evaluation cycles, you save roughly 3 hours on average. Tasks the AI fails still cost you (wasted prompting and review time), but the wins more than compensate.

Here's where it gets interesting for marketing. Not all marketing tasks have the same Probability of Success when delegated to AI. And the tasks most businesses hand over first are the ones where that probability is lowest.

The Delegation Spectrum for Marketing Tasks

Here's how common marketing tasks map against the equation, categorised by what determines their Probability of Success: whether the task has clear rules, measurable outputs, and well-defined "done" criteria.

Why Activation Delegates Beautifully (and Brand Building Doesn't)

Les Binet and Peter Field's analysis of nearly 1,000 IPA effectiveness case studies produced one of marketing's most validated frameworks: the split between brand building and sales activation.

Activation work is short-term, targeted, measurable, and designed to convert existing demand. Brand work is long-term, broad-reach, emotionally driven, and designed to create future demand.

Here's the insight that connects this to AI: activation tasks are exactly the kind of work where AI has a high Probability of Success. They have clear inputs, measurable outputs, and feedback loops that let AI self-correct. When you ask AI to optimise bids on a Google Ads campaign, there's a right answer, and the AI can find it faster than you ever will.

Brand tasks are the opposite. They require you to decide what your business means, which memories you want to build in buyers' minds, and how to feel different from every competitor in the category. There is no "right answer" to optimise toward. There's only judgment.

Binet and Field's recommended budget split (roughly 50/50 to 60/40 between brand and activation for established businesses) maps cleanly onto the delegation boundary. The activation side of your marketing budget is where AI earns its keep. The brand side is where AI produces its worst results.

The 2026 Supermetrics Marketing Data Report confirms this at scale: while 91% of marketers actively use AI, only 6% extract meaningful bottom-line value. The teams that do? They deploy AI for operational execution, data analysis, and performance reporting. Not for strategy. Not for positioning. Not for the creative direction that determines whether everything else works.

AI Will Never Find Your Best Marketing Idea

Rory Sutherland, Vice Chairman of Ogilvy UK, has spent three decades proving that the most valuable marketing ideas are the ones that seem irrational. He calls them psychological moonshots: small, cheap interventions that produce outsized effects because they work with human psychology rather than against it.

An e-commerce site changed one button label from "Register" to "Continue." Revenue increased by $300 million per year. The London Underground's greatest improvement in passenger satisfaction wasn't faster trains. It was dot-matrix countdown displays. The wait didn't change. The uncertainty did.

These ideas share a trait that makes them structurally impossible for AI to generate: they don't make logical sense until after you've seen them work.

Sutherland puts it bluntly: "It is much easier to be fired for being illogical than it is for being unimaginative. The fatal issue is that logic always gets you to exactly the same place as your competitors."

AI optimises for the logical answer. It draws from the largest corpus of existing ideas and converges on the consensus. That makes it brilliant at tasks where the consensus IS the right answer (bid management, data analysis, reporting). And it makes it structurally incapable of generating the counterintuitive insights that create the most value in creative and brand work.

This isn't a temporary limitation that the next model update will fix. It's how the technology works. AI can execute against marketing principles, but it cannot create meaningful distinctiveness on its own. When you delegate creative strategy to AI, you get back the average of everyone else's creative strategy. That's not a moonshot. That's a race to the middle.

What Byron Sharp's Research Says You Can't Delegate

Sharp's research at the Ehrenberg-Bass Institute, across 130+ brands in 13+ product categories, shows that brands grow by being easy to think of (mental availability) and easy to buy (physical availability). Not by being differentiated or deeply loved.

Building mental availability requires three things that AI cannot do well:

The Management Multiplier

Mollick's equation has three variables, but only one is within your direct control: Probability of Success. You raise it with better instructions, faster evaluation, and more precise feedback. In a word: management.

His experimental class at Wharton proved this. Executive MBA students used AI to build working startup prototypes in four days. They outperformed expectations by an order of magnitude. Not because they were technical. Because they had years of experience scoping problems, defining deliverables, and recognising when output was off.

"The skills that are so often dismissed as 'soft' turned out to be the hard ones," Mollick writes.

For marketing, this means the businesses that get the most from AI aren't the ones with the best tools or the biggest budgets. They're the ones that know what good marketing looks like BEFORE the AI starts generating output. They can evaluate whether an ad concept will build mental availability. They can judge whether a landing page will convert. They can spot when AI copy sounds like everyone else's AI copy.

Without that judgment, no amount of AI tooling will compensate. You'll produce more output, faster, with less distinctiveness. Which is exactly what the 91%-adoption, 41%-ROI gap tells us is happening across the industry.

What This Means for Your Business

The practical framework is simpler than you'd expect.

• Performance reporting and data analysis (highest ROI per hour saved)
• Bid management and budget pacing
• Keyword research and expansion
• Ad copy variations from an existing, human-written creative brief
• Audience segmentation from conversion data
• Email sequence drafting from a defined strategy

• Brand positioning and messaging strategy
• Creative direction and concept development
• Offer design and pricing architecture
• Category Entry Point identification and prioritisation
• Distinctive brand asset creation
• The "what should we actually do?" question

Research shows AI can reclaim roughly 27 hours per week from manual marketing execution. Those hours are real. But where you redirect them is what separates the 6% who extract value from the 85% who don't. The businesses winning with AI marketing aren't using those hours to produce more content. They're spending them on strategy, creative development, and the kind of counterintuitive thinking that AI will never replicate.

The equation doesn't lie. Know the Probability of Success for each task before you hand it over. And remember Sutherland's warning: the most valuable ideas in your marketing are the ones that don't make sense in a spreadsheet.

Further Reading

• Management as AI Superpower by Ethan Mollick - The Wharton framework for AI delegation decisions
• GDPval: AI vs Human Expert Performance - OpenAI's benchmark comparing AI output to human professionals across 44 occupations
• The Long and the Short of It by Les Binet & Peter Field - The IPA effectiveness research on brand building versus activation
• Why a 2026 Marketing Shelf Needs Byron Sharp - Mental availability in the AI era
• 2026 Marketing Data Report by Supermetrics - The AI adoption versus value extraction gap