Uber burned through its entire 2026 AI budget by April. I keep coming back to that story, because it says a lot about where marketing is heading. AI is moving out of personal chats and into shared team systems. It costs real money now. And someone has to decide what it’s for. I write predictions like these every year. This year, three threads run through almost everything I see coming, so let me start with them.
AI work needs a home. For the last two years, most of us used AI on our own. Our prompts, our chats, our files on our laptops. That stops working the moment a team depends on it. The positioning, voice and skills your AI uses need to live somewhere shared, with a history of every change, owned by your company and not by whichever AI vendor is winning this quarter. GitHub, the tool developers have used for years, is a strong candidate. The name sounds technical, but the idea is simple: one shared folder for your team’s AI work that remembers who changed what, and why.
Someone has to pay for it. In 2026, many companies told everyone to use as much AI as possible. Some even ranked employees by how much they used. Then the bill arrived. As AI costs move into marketing budgets, every use of AI will need a reason. “We used a lot of it” will stop counting as a result.
Some things shouldn’t be AI-written at all. This is the thread I care about most. Once AI use has to be justified, it gets easier to say out loud what should stay human. Saving money is only part of it. Some content works because a person wrote it: your point of view, your customer stories, the posts where your audience wants to hear you. In 2027, being firm about that line becomes part of the job.
Here are all 13 predictions, grouped around those three threads.
Your AI work needs a home
1. GitHub becomes marketing infrastructure
Teams that work seriously with AI build everything they do on a set of shared files: positioning, brand voice, audience research, and the skills that automate their reports, briefs and drafts. Every brief, draft and report starts from those files. Yet in most teams, they still sit on individual laptops, on Google Drive or inside one AI tool, where only one person can use them.
In 2027, those teams will keep it in a shared repository. A repository, or repo, is a shared folder that remembers every change: who made it, when and why. Developers solved this exact problem years ago, and GitHub is where most of them do it. It works just as well for marketing files. “Repo” stops being a developer word.
2. Model loyalty dies
A marketer I interviewed recently told me her team moved from ChatGPT to Gemini, and then to Claude. Both moves were messy, and some people still have projects stranded in the old tools.
Nobody wants to go through that again. In 2027, being able to switch AI vendors becomes a must-have. It’s one more reason your team’s AI knowledge should live somewhere you own.
3. AI asks for your strategy. Many teams don’t have one.
To get good work from AI, you have to write down your positioning, your audience and your voice. Clearly, and in detail.
Many teams will find they can’t. The strategy lived in one or two people’s heads and was applied case by case. Writing it down will be the hardest AI project of 2027, and the least technical one.
4. The first marketing fail caused by a markdown file
Once a team shares the same source files, one small edit travels everywhere. Someone changes the definition of a qualified lead in a plain text file. Three AI workflows quietly change with it, and nobody who owns positioning approved it.
Developers handle this with reviews and clear ownership. Marketing will learn the same lesson, probably the hard way. Suddenly, a markdown file needs the same sign-off as a press release.
The people who make it work
5. The Marketing Engineer gets a job title
Someone has to build and maintain the team’s AI workflows, keep the shared files up to date, and fix things when they break. Right now, that person is usually a marketer doing it on the side.
In 2027, it becomes a real role with its own budget, working next to the writers and strategists. Not every marketer needs to become technical. Every team needs someone who is.
6. The quiet promotion
The marketer I mentioned earlier didn’t wait for a mandate or a budget. She built something useful for herself, noticed it was stuck on her laptop, and made it available to her whole team. That made her the AI lead almost by default.
This will happen in many teams. The fastest promotions in 2027 go to people who built something their colleagues use every day. And when you’re job hunting, what you’ve built will say more about you than your CV.
The bill, and what’s worth it
7. Somebody has to pay for all those tokens
AI tools charge by the token, a small piece of text the model reads or writes. The price per token keeps falling. But AI agents use far more tokens than a simple chat, so total bills keep growing.
In 2026, many companies learned this the hard way. Uber reportedly used up its full-year AI budget in about four months, after encouraging heavy use and ranking employees by it. Then it capped spending per employee (Bloomberg, via LetsDataScience).
As these costs land in marketing budgets, AI spend has to show a return. Using a lot of it no longer counts as success. The teams that can say what each workflow costs, and what it brought back, will keep their budget.
8. Writers split in two
Writing on demand keeps shrinking. AI produces a passable first draft of almost anything, and cheaply.
At the same time, another kind of writing work grows. Someone has to write the examples the AI learns from and check that what it produces sounds like your company. Someone also has to decide which content a person should always write: the founder’s opinions, customer stories, and the posts where your audience expects to hear from you directly. When AI costs real money, that decision is easier to defend. But the bigger reason is that good writing has become a real advantage. When most content is AI-written and sounds alike, a clear, specific, human voice is what makes people stop, read and remember you. Writers who can do that will be in more demand in 2027.
9. Sales calls become your best content source
AI writes to the average. Ask it for positioning and you get the version that shows up most often online. Your competitors get the same one.
The way out is what they don’t have: your customers’ exact words from sales calls, the questions prospects keep asking, and the real reason you lost last month’s deal. In 2027, call recordings move from sales tools to the center of content strategy.
Getting found
10. A blog only humans can update becomes a problem
If AI can’t read and update your content without someone logging in, every refresh stays manual. Meanwhile, competitors whose content AI can work on directly will update theirs frequently.
Old statistics, outdated screenshots and broken links used to be a slow leak. In 2027, they become a visible gap.
11. Every company needs a sitemap for AI agents
AI agents are starting to research and choose products for people. They trust verified, structured facts more than clever copy.
A page that describes what you offer, written for machines, becomes as basic as a sitemap. If an agent can’t understand your product, you don’t make its shortlist.
12. Nobody pays a creator without an audit
B2B brands now pay creators the way they used to buy media. And where the money goes, fake engagement follows.
In 2027, nobody signs a creator deal without checking that the audience is real, in the right market, and made up of the people the brand actually wants to reach.
And the good news
13. The best time to be a marketer
Anyone can build a product in a weekend now, so you’ll have ten competitors by Friday. Most of them will have similar features and similar AI-written landing pages.
Building got easy. Getting people to care got harder. If you can earn attention and influence across channels, there has never been a better time to be a marketer.
Bet against me
Which one do you think I got wrong? Tell me the number in the comments. In October 2027, we’ll see who got it right.
