2026: Marketers, Think Through These 10 Things First

A while back, I went to a small dinner with a group of marketing leads. Seven people, four of them CMOs.

A few drinks in, the conversation turned, uncannily, to the same question: what on earth will my department look like in 2026?

One of them set down his glass and said something I still remember: "I'm not afraid of AI taking my job. What scares me is that I can't say clearly what I'll actually be doing next year."

2026 is more than half over. I've taken everything from these past months — conversations with dozens of marketing leaders, the research reports I read, the case studies I saw — kneaded them into 10 judgments, and packed them into 6 topics. Some are exciting. Some will send a chill down your spine.

That's fine. Let's take them one at a time.

Let Me Start With a Boat Race

Every year, Oxford and Cambridge race each other on the Thames. The race has existed since the 1830s, stopped only by World War II and the pandemic. It's 4.3 miles of water, with hundreds of thousands of spectators lined along both banks.

One year, the day before the race, a rower on one of the boats got injured and was replaced at the last minute. That hadn't happened in decades.

The next day, the winning boat won by just 0.5 seconds.

4.3 miles. 0.5 seconds. Measured along the hull, that margin is about one foot.

Think about it: on a racing shell, the coxswain calls the rhythm, the four in the middle are the engine, and someone handles balance while someone handles direction. Who each person is, and what each person does — crystal clear.

Now swap that boat for your 2026 marketing team.

On the roster of rowers, how many are human, and how many are AI agents? Who's the coxswain, and who pulls an oar?

Can't answer right away? That's exactly the point.

The org chart hasn't had time to change, but your certainty about who you are and what you do has already shattered.

The layoff list hasn't been drafted yet, but the identity crisis has already arrived. Are you the engine doing the execution, the coxswain setting the direction, or the one holding the balance? This question will chase you all through 2026.

And there's a study that cuts even deeper.

Researchers compared two kinds of leaders: one kind manages only people; the other manages mixed teams of humans and agents. The conclusion: leaders who can manage agents well do just fine when they go back to leading all-human teams too.

The reverse does not hold.

What's it like to lead a team that's half human, half software? Your people start fighting agents for tasks. Your agents start "bullying" your people. Sounds like a sci-fi movie?

In 2026, it's becoming a documentary at many companies.

AI Isn't Abstract Anymore. We Are.

The new generation of reasoning models can already post frighteningly good scores on the CFA (Chartered Financial Analyst) exam.

The CFA is widely recognized in finance as one of the hardest exams there is. The key point: these models never ground through years of past papers. They relied on their own reasoning — and have even developed something eerily close to "intuition."

Let that sink in.

The models are getting stronger at a speed you can see with the naked eye, sprinting toward artificial general intelligence. And on the other side?

The way we use AI is still stuck at opening accounts for the team and running a few isolated pilots.

In December 2025, the executive search firm Spencer Stuart surveyed several hundred CMOs at mid-to-large brands. More than half said: we're piloting AI projects.

And then? And then nothing. The pilots ended. Nobody scaled.

Put bluntly: the engine has already moved to a new generation, but the way we drive is still the same old routine.

I think of AI adoption as a four-story building.

The first floor is experimentation: open accounts, open sandboxes, run pilots. Curiosity through the roof; value in scraps. The second floor is enablement: tools like Midjourney, ElevenLabs, and Adobe Firefly enter the daily workflow, and a few enthusiasts pop up in the team, lowering the barrier for everyone else. The third floor is AI-first workflows: creative, media, and operations connect into one system, designed for speed and consistency. The fourth floor is end-to-end redesign: from strategy, insights, briefs, creative, and media to measurement, the entire marketing function is rebuilt the AI-native way — which segments go to people and which go to agents is decided by deliberate design.

Most companies are stuck on the second floor. 2026 is more than half gone, and they're still stuck.

Why? Because getting to the third and fourth floors means smashing the existing processes and rebuilding them. Too painful, too risky — and the people haven't learned how yet.

But hidden inside this is an opportunity: the more big companies seize up, the bigger the opening for small teams that were born AI-native. Old ships are hard to turn; new ships are built straight from the new blueprints.

First, Figure Out Where Your 20% Is

At dinners like these, a few drinks in, there's always a CMO who leans in, lowers his voice, and tells me: the boss has already hinted that the 2026 and 2027 budgets will be redone — cut costs, and cut them with AI.

Where does the saved money go? Into enterprise licenses for ChatGPT, into data, into research.

There's another set of numbers in that Spencer Stuart survey: more than 40% of marketers say their CEOs and CFOs expect the marketing department to save more than 20% in costs over the next 12 to 24 months.

20%. Think about it: on a budget of 10 million, 20% is 2 million. Roughly the payroll of a mid-sized team.

If your boss hasn't asked you this question yet, don't celebrate too early. It's coming, sooner or later.

So don't wait. Start taking stock now: which 20% is cut outright, and which 20% is won back through efficiency. If you only start thinking on the day you're asked, it's already too late.

In 2026, the luxury of "let's just try things and see" is gone. The bets are fewer and harsher — and not yours to choose.

An Ad No Human Touched

Here's a piece of copy:

"For too long, greatness has been defined by suffering. Pain is power, struggle is a badge of honor, and play has no place. But we were never like that. Greatness is the courage to do it your own way. Because greatness was in your nature to begin with."

Good copy, right? Emotion, attitude, a twist — it's all there.

Now the reveal: this Puma film — from consumer insight, to brief, to storyboard, to rough cut, to final cut — was made by a relay of AI agents. The team at the agency Monks was mainly there to watch from the side.

You'll say: doesn't this just scream AI? AI slop, all over the streets.

Right — and there will be more and more slop. But here's an uncomfortable truth: many things you judge as "not good enough" are, when deployed, exactly good enough.

Coca-Cola, two years in a row — end of 2024 and end of 2025 — made its Christmas ads with AI. Not only did the films air; they also passed the System1 tests that measure ad effectiveness.

Fast, cheap, and able to produce a thousand versions at once, each served to its own audience.

"Good enough," plus cheap and fast, is, in many scenarios, the new "good."

Follow this production line onward, and the 2026 marketer's toolbox is changing too: from scattered single-point tools to orchestrating whole workflows.

Who provides the orchestration? The large model vendors, enterprise platforms, specialist startups, no-code agent builders — or simply raising your own in-house agents. Most teams use a mix.

For what? Brand-compliance guardrails, content and personalization, chatbots, ad orchestration, real-time research, landing-page generation.

Where does the money flow first? Campaign orchestration. A single campaign — from insight to deployment to optimization — runs its whole chain through agents, with humans holding the key checkpoints. This is where 2026 is most likely to see its first real, hard-cash returns.

The Foundation of Search Is Being Hollowed Out

Let me tell you a sad story.

Chegg, an American company doing online high-school education, was built on SEO. Good content, high rankings, students flocking in — for a while, a perfectly respectable business.

Once ChatGPT arrived, students with questions stopped searching and just asked. Chegg's business today is down to a shell.

If a business is built on the arbitrage of "users search first, then discover you," its foundation is being hollowed out by large language models, one shovelful at a time.

So will search die? No. Google's data shows that the highest-intent users are still searching. But the decision-making in between has been churned up by large language models — and how it plays out depends entirely on the industry: products with high decision costs and products with low decision costs are seeing two entirely different traffic stories.

How fast is the change? On Black Friday, November 2025, traffic to US retail sites from AI surged 805%.

805%. Read that number three times.

What is answer engine optimization (AEO)? It used to be that you optimized your ranking on the search results page. Going forward, what you optimize is the moment AI opens its mouth to answer: whether you're in it — and who it says you are.

In 2026, you'll hear that term until your ears grow calluses.

You Didn't Order This Dish, but the Bill Still Arrives

In the fall and winter of 2025, Google released Gemini 3.0. Across a series of tests, it was faster and stronger than the latest version of ChatGPT.

OpenAI's CEO Sam Altman sounded the highest-level internal alarm, and the entire staff sprinted to rework the models.

The alarm wasn't wrong. But it was aimed in the wrong direction.

What should truly keep every brand awake at night is something else that happened in July 2025.

In the United States, a high school student was still chatting with ChatGPT at 4 a.m. As the conversation went on, the kid expressed thoughts of ending his life.

The machine replied with a passage that amounted, more or less, to a farewell and a blessing.

A few months later, the child was gone. The family took OpenAI to court.

This story makes me uncomfortable every time I think of it. OpenAI patched things up round after round afterward — but they too will get blindsided by the very things they built.

What I want to talk about isn't this one incident, but the pattern behind it: after AI scales, the side effects are endless. Misinformation, disinformation, ethical black holes, product misuse. Pick any industry, and there's a pit waiting.

The trouble is, none of these risks knocks first. And no matter whether it starts in product, technology, or customer service, the hand that reaches out at the end always finds the brand. And when it touches the brand, it becomes marketing's problem. Whether you catch it or not isn't up to you.

So in 2026, governance is no longer a PDF in the compliance department's filing cabinet. It's a required course for the marketing leader.

So What Do You Do?

The researcher Ethan Mollick has a saying that I consider the sharpest knife for understanding AI: AI's capabilities are jagged.

The same model: at its best, stunning enough to raise goosebumps; at its worst, stupid enough to make you question your life choices.

Fans only look at the tips of the sawtooth; haters only look at the valleys. Both sides are wrong.

One slip doesn't mean the whole thing is useless. One stunning moment doesn't mean it's reliable everywhere.

So how do smart people get along with it?

At first, the "centaur" model: this step goes to AI, that step goes to the human, the division of labor written on paper — who does what, crystal clear.

Later, the "cyborg" model: a stretch of work is finished, and you can't say which part was human and which part was machine. The boundary has dissolved.

In 2026, more and more marketers will live as cyborgs.

Back to the boat race at the beginning.

The sound of the oars is still there, the riverbanks are still there, the cheering is still there. The only thing that's changed is the roster of rowers — it's different every single week.

And back to that man at the dinner. He said he was afraid of not being able to say what he'd be doing next year. In fact, he'd already worked out half of it: knowing you can't say — that's where thinking it through begins.

Salesforce's CEO Marc Benioff once said something weighty: AI and generative AI may be the most important technology of any era.

Note: "any era." Not our era, and not our children's era.

A few years ago, a line like that would have sounded arrogant. Today, hearing it, you just go quiet for a few seconds.

The changes are indeed big. But standing on top of every one of those changes are the people who thought it through first.

For the rest of 2026, I hope you keep your seat steady. And when the day comes that agents come for your work, I hope you're still holding something they can't take.