Opinion: marketers should produce fewer AI‑generated assets to avoid bottlenecks

After more than 25 years in marketing, I find it absolutely absurd that I’m about to argue for producing less. I’ve spent much of my career trying to get more out of a budget, a team, and an agency relationship. More has always been better. More ideas. More opportunities to get in front of customers. More time to do the work would’ve been nice, too, but I don’t want to sound whiny.

More is exactly what AI has promised us and delivered. Give a marketer the ability to turn a brief into a dozen executions before lunch, and they’re going to use it. I would (and have — don’t tell the boss!).

What I’m less convinced about is what happens after lunch.

Human work still needs to be done. Human work still needs to be done. Somebody still has to review those executions, verify the claims, check for brand consistency, etc. But the rest of the process hasn’t changed, so the speed gained from AI can disappear as the work moves through the department. Everyone else may still be operating exactly as they did last Tuesday.

This is where my enthusiasm for more starts to run into questions. We talk a lot about how much marketing AI can produce. The bigger question is how much marketing an organization can actually handle. AI does something else really well: exposing bottlenecks in the process.

The world’s most powerful SEO platform, purpose-built for Enterprise.

More output exposes the bottlenecks in your workflow

Think about the approval process in your own company. Where does work get stuck? Is it really the first draft? Sometimes, sure. But it might be an incomplete brief, three people with conflicting feedback, or a decision that nobody seems authorized to make. Adding 40 or more versions to the mix won’t resolve any of those process impediments. They’ll just pile on the work – and frustration.

Adobe’s 2025 research shows the scale of the problem. In a survey of more than 1,600 marketers, 89% said content went through at least three approval stages. About 58% said more than 40% of their time was spent managing reviews and approvals.

Stop and really ponder that — 40%. Wow. That’s very disproportionate and not a good use of our time, especially at a senior level. We should be focused on strategy, not monotony. But so many marketing executives will tell you that they spend countless hours reading and reviewing every week.

Imagine your team creates 20 assets a week for you to review. Now, imagine AI produces 100. Congratulations, you’ve just been demoted to full-time reviewer or (and this happens all the time) you hire someone to review it for you, negating the cost efficiencies of using AI in the first place.

An organization has a finite absorption capacity. There’s only so much work that creative, legal, brand, media, and analytics can get through.

Now, I don’t want to limit creative exploration. More ideas are a good thing. They expand our thinking and open us to novel approaches, but we don’t need to review all of them.

Separate creative exploration from creative approval

Having more creative may increase your CTR by 0.1%, and that may translate to substantial revenue for your org and pad your accomplishments trophy case, but at what cost? Time and again, the great creative idea will beat the one optimized to death. Maybe less is more, at least when we focus on quality.

Less doesn’t mean fewer ideas. It means selecting fewer ideas for approval. Before creative work enters the approval queue, someone should be able to explain who it’s for, what it will accomplish, and why it’s better. From there, decide which creative deserves senior approval. That’s the person who owns the bottleneck. There’s at least one in every org, and you know who you are.

The same discipline matters when we test. Say we change the headline, image, offer, and audience, then get a better result. Great. What should we do next? We know the combination performed better in that test. But is testing everything really the best approach?

There’s nothing wrong with comparing complete creative concepts, provided that’s the question we intend to answer, but if we want to isolate the effect of a headline, we should work from the premise that we don’t need to solve everything at once. We can build on what we learn across campaigns. AI can help turn our DAMs and share drives into living repositories of knowledge that grow over time.

AI can help develop the treatments and organize the findings. We still owe ourselves an honest interpretation, including the occasional “we don’t know yet.” I’d much rather hear that than watch an inconclusive result become a confident recommendation by the time it reaches the executive presentation.

Build the process around decisions, not organizational charts

Implementing any of this gets messy. Marketing departments come in every imaginable shape and size. At one company, a single person owns content, CRM, and reporting. At another, those responsibilities span several teams and a couple of agencies. Even companies with similar org charts can make decisions in completely different ways.

How do we standardize enough to introduce AI without pretending everyone works the same way?

Start by asking the same questions, rather than prescribing the same structure. What are we trying to achieve? Who owns the decision? Which information can we trust? Where does the work wait, and what happens when someone disagrees?

The answers should shape a marketing operating system where work gets requested, chosen, approved, distributed, and learned from. AI should orchestrate and help people follow that process. It should make approved claims easy to find, show teams what content already exists, and preserve what we learned from the last campaign. Sometimes, its most useful contribution will be saving us from commissioning something again.

In McKinsey’s March 2025 State of AI study, workflow redesign had the strongest relationship with self-reported generative AI earnings impact among the 25 attributes examined. That doesn’t establish causation. It does give us a reason to look at how work moves through the business before buying another tool.

Fix one marketing workflow before changing everything

Yeah, that sounds simple, but it isn’t easy to implement. I agree, but you don’t need to change everything overnight. Pick one recurring piece of work and follow it all the way through. A lifecycle email will do. Sit with the people involved and reconstruct what actually happened, including the approval buried in Slack and the brief that changed halfway through. Work out how much time went into creating, waiting, and doing things over.

Then fix the problem you find. If people can’t locate approved assets, help them retrieve those assets. If the brief is missing essential information, improve the intake. If nobody can make the final call, give someone that responsibility. AI may help with the first two. The third is yours.

Run the revised process for several real cycles. Check the total time and effort, including the work of checking AI’s output. Look at how much commissioned content reached customers and whether the tests informed a decision. Keep an eye on lead quality, complaints, and unsubscribes. A faster process that annoys more customers deserves some scrutiny.

Be honest about the economics. Twenty minutes saved on a draft can vanish in an extra hour of review. Integration, licenses, and implementation cost money. Recovered time is valuable, but we still have to decide how to use it before calling it a return.

If I’m taking an AI investment to the CFO, I want to show that we sold more, improved lead quality, reduced actual costs, or stopped funding work that went nowhere. Merely telling them, “We made another thousand assets,” will leave them questioning your motives or, worse, your competency.