AI can do the marketing work. It can’t make the marketing decisions.
LLMs like Claude, Perplexity, ChatGPT, and TEAM LEWIS’ own Marketing SideKick are some of the most methodical media planners and buyers that I’ve ever worked with.
These LLMs are taking prompts for campaign structure, naming conventions, budget allocations, targeting, forecasts and automating the delivery of these critical building blocks for paid media programs. The turnaround time is lighting quick, the iteration is in real time, and the identification of typical pain points and pitfalls are identified pre-launch. Given enough data and the right context, these tools will work at all hours of the day without complaining about the brief, the timeline, or the rounds of feedback.
Yes, AI can do the job. Or at least, an increasingly large portion of it. But that’s not the point.
Why AI automation is so attractive to marketing leaders (the seductive math).
If you’re sitting in the C-suite or on a procurement team, you’re seeing dollar signs.
Finally, the agency commission model that marks-up the buying is getting exposed. If AI can systematize the execution work, improve go-to-market timelines, reduce headcount, and automate production then many companies are wondering “why can’t I just do this myself?” It’s a good debate; what could a company do if it shaves off agency costs? Consolidate agencies, put the savings into sales and product development, aid margin, stuff the coffers – it’s all pretty sexy math.
But the math only works if the doing was the advantage. If hands to keys, campaign builds, optimizations, reporting decks and trafficking were where the value lived.
This may come as a shock to performance-driven teams, but they weren’t. Those execution tasks were expensive because they required people (it was a slog of a process). That didn’t make them the most valuable part of the job.
AI has collapsed the cost of doing. And it will continue to collapse it. But in doing so, it has raised the price of deciding.
When everyone can execute faster, produce more and optimize continuously, execution stops being the differentiator. The advantage shifts upstream: deciding what deserves to exist in the first place.
What AI can actually automate in marketing
There’s an important distinction between automating marketing work and automating marketing judgment. AI is exceptionally good at the first.
A huge portion of modern marketing is built on repeatable tasks: resizing creative, generating copy variations, building audience segments, analyzing campaign performance, summarizing research, creating taxonomies, QAing links, forecasting budgets and identifying patterns across enormous datasets.
Those tasks used to consume hours. Some consumed days. AI can now do many of them in minutes.
That’s a meaningful change, and marketing leaders should absolutely take advantage of it. I don’t believe the answer is to protect work simply because people used to get paid to do it. If a machine can build 50 ad variations faster, let it. If it can catch a pacing issue before a human analyst, great. If it can turn a six-hour reporting exercise into 20 minutes, nobody should mourn the spreadsheet.
But automation doesn’t eliminate the need for expertise. If anything, it exposes where expertise actually matters.
Someone still has to decide:
- Which audience matters?
- Which signals are meaningful?
- Which KPI is telling the truth?
- Which platform recommendation is serving your business versus the platform’s business?
- Which trend deserves investment and which is just LinkedIn having a collective moment?
AI made the mechanics easier. It didn’t make the decisions easier.
Three marketing problems AI makes harder, not easier.
The irony is that marketing teams are losing expertise precisely at the moment the decisions are getting harder.
Companies are cutting agencies, flattening teams and moving capabilities in-house while simultaneously introducing AI into every layer of the marketing organization. On paper, that can look incredibly efficient.
But efficiency without architecture just helps you move faster in whatever direction you happened to be heading.
And there are three questions becoming much harder to answer.
1. AI is making clicks disappear. What do you measure when the journey ends inside an answer?
- For years, digital marketing built an entire measurement economy around the click: Search > Click > Visit > Convert > Attribute
- Now discovery increasingly happens inside AI-generated answers, zero-click search results, social feeds, Reddit threads, creator content and other environments where influence happens without producing a neat trail back to your dashboard.
- Your customer may know your company, trust your point of view and put you on a shortlist without ever giving your attribution model the satisfaction of taking credit for it.
- AI can analyze the data you have. The harder question is whether you’re measuring the things that actually matter.
2. Performance is crowding out brand. Who defends the spend that’s hardest to attribute?
- Optimization engines will always favor what they can see. That makes them extraordinarily good at harvesting existing demand and dangerously convincing when they tell you to keep feeding the bottom of the funnel.
- Someone still has to ask where tomorrow’s demand comes from. Someone has to defend investment in attention before intent, build memory before the buying window opens and recognize when an efficient campaign is simply getting better at reaching people who already know you.
- That is a business decision disguised as a media decision.
3. You’ve decided to in-house. Who is architecting the operating model the AI runs inside?
- Bringing capabilities in-house can make enormous sense. But headcount is not an operating model, and neither is a stack of AI subscriptions.
- Who owns strategy?
- Who challenges channel recommendations?
- Who connects brand and demand?
- Who decides what gets centralized versus localized?
- Who owns the measurement framework?
- Who sees across functions when every internal team is understandably focused on its own objectives?
- These are charged questions your AI solution isn’t prepared to answer. Instead it is learning from what already exists. And some of the most valuable knowledge in this industry doesn’t exist in a public playbook because agencies, operators and experienced marketers have spent years learning it the expensive way.
- Sure, there are plenty of playbooks online. But are you really going to build the next marketing operating model from every business influencer’s MLM scheme? Go ahead and try, comment “playbook.”
Why marketing judgment becomes more valuable as AI improves
The job was never the point. The system was. The system is built on judgment: what to build, what to own, what to measure, what to stop doing and where to place the next bet.
It’s knowing where attention can be earned, where audiences can be owned and when you need to buy access to them. It’s knowing when the numbers are telling you something useful and when they’re simply telling you what’s easiest to measure. It’s understanding that a channel performing well doesn’t automatically mean the marketing is working.
AI is helping teams accelerate execution. That is good for marketing. But the faster execution becomes, the more important the architecture around it becomes.
The remaining moat isn’t knowing how to launch the campaign. Everyone will know how to do that. It’s knowing why you’re launching it, how it connects to everything around it and whether you should be launching it at all.
Four questions leaders should ask about AI and their marketing operating model
If you’re a marketing, finance or procurement leader assessing the marketing function this year, I’d start with four questions.
- Do you actually know what’s working, or just what’s attributable? Those are not the same thing, and the gap between them is getting wider.
- Does your in-house model have an architecture, or just headcount? Moving roles onto your payroll doesn’t automatically create integration, governance, or strategic clarity.
- Are you reaching buyers before they’re in market, or only nurturing the people who already are? If every optimization is pointed toward existing intent, eventually you run out of demand to capture.
- When an AI answers a question about your category, are you in the answer? Because your next buyer may never make it to page one of Google, let alone click your ad.
If you answer “no” or “I don’t know” to any of these questions, ask your team. And if they’re struggling to answer them too, that’s the problem worth solving.
Where agencies fit in an AI-powered marketing organization
The benefit of an independent, integrated and global agency isn’t that we have more hands to put on keyboards. That advantage is disappearing for everyone.
It’s perspective.
We get to see what happens across categories, markets, platforms and operating models. We can challenge the brief because we don’t live inside the organization that wrote it. We can tell you no. We can tell you something isn’t working. We can say, “I don’t know. Let’s test it and see what happens.”
AI will make the doing faster and cheaper. Let it.
Then make sure you still have people around who know what’s worth doing.
