If you’re in marketing, you don’t need another hot take about how AI is going to change everything.
You’ve already seen how every company’s current “pivot” is to AI. You’ve already seen how every board under the sun wants their companies to use AI for everything.
Likely, you’re already using ChatGPT or one of its imitators every day.
The BT crew has been at the frontline of AI adoption for over two years. We’ve seen inside a hundred companies, and seen how they’re using (or not using) AI, and getting (or not getting) results. We’ve played with all the tools.
We thought it was time, in this Q4 of 2025, to give a few sharp takes. Here we go.
Execs Adore Control and Will Not Hand Their Companies Over to AI
We keep hearing about armies of “fully autonomous agents” that will soon run sales, marketing, or even entire companies. Maybe one day.
But here’s the current reality: most executives still struggle to relinquish control over something as basic as website copy. They will likely never hand over ops wholesale to bots.
Anyone who has ever tried to push new messaging or a homepage redesign through an executive team knows the truth: people guard brand, language, and style like their life depends on it. As they should.
But it means the leap from obsessing over a tagline to handing over entire departments to AI agents is… a big one.
Expect tools to augment. To draft. To speed up. But don’t expect execs to willingly turn over the keys to the kingdom anytime soon. Small tweaks to ad creative? Sure. Deciding what wording to put behind six and seven figure ad spend? Not so much.
AEO Is Still 90% SEO
Otterly. Profound. SEMRush’s new toolkit. “What is ChatGPT saying about us?” “Will we vanish in the zero click era?”
No doubt, search is changing. LLMs and AI summaries are changing how information is packaged. But let’s be clear: the fundamentals are still the same.
The content that works in AI Overviews is the same content that works in Google search. Clean structure. Clear H1s and H2s. Schema markup. Concise, scannable answers. Authoritative backlinks. Recognizable brand entities.
Even the extras, like bigger FAQs, trying to appear in higher-karma Reddit threads if you can, it’s all garland.
In other words: AEO is SEO with a new coat of paint. If you’re already doing the work of high-quality SEO, you’re already 90% of the way there. The last 10% is testing, tweaking, and packaging for AI readability.
The LLM Voice Is Obvious, Cheesy, and Generic (and Everyone Can Spot it)
It’s now a common experience: You receive an email, and immediately, you can smell it: the sender didn’t write this. They barely even edited it.
They pasted it from ChatGPT. The tone, the style, the syntax, the airy cheeriness. It’s all there.
Does it make you feel great? Seen? Paid attention to?
No. It’s the new form of direct spam. Grammatically flawless emails that are too polite, overexplain, overuse transitional phrases, and lean on words like “delve,” “realm,” or “journey.”
Brands that rely too heavily on raw AI output are already sounding the same. And sameness is death in marketing.
Perhaps as these LLMs improve, they will learn to cloud their native tone of voice, and sound human. But that day isn’t today, nor likely anytime soon.
AI Promises Speed – but the Problem was Never Speed, It was Bottlenecks
Most AI tools brag about making teams faster. And to be fair, they deliver: more content, more iterations, more ideas, all at lightning pace.
But here’s the catch: at Buttered Toast, we’ve seen that, pre-AI, slow production was rarely the real problem. The real problem is getting things finalized and out the door.
Too many stakeholders. Vague processes. Endless review cycles. Last-minute edits from the exec who skimmed it on a Sunday night.
AI cannot fix organizational bottlenecks. A team will never move faster than its slowest signoff. Until companies tighten up decision-making and trust the process, all the AI horsepower in the world won’t stop campaigns from stalling. It will just create more crowded bottlenecks.
The More Complex the Sale, the Less AI Can Help
What you sell shapes how you use AI. If your product is complex, if the sales cycle is long, if average deal size is high, then the stakes are too big to hand real autonomy to a machine. The harder the decision, the more human oversight it requires.
That doesn’t mean there’s no role for AI. Far from it. Custom GPTs can churn through content drafts or scheduling. Campaigns can be optimized within clear constraints. Internal ops can run smoother with process automation.
AI has a place, and we’ve seen it deliver value. But the bigger and more complex the deal, the more it works as an assistant, not a decision-maker. We don’t see this changing anytime soon.
The Next 12 Months: Guardrails, Not Autonomy
So where does this all go in the year ahead? Well, we don’t think 2026 will be the year of fully autonomous agents running enterprise companies. But it will be the year where AI moves faster than humans can comfortably keep up with, and that will force companies to build real guardrails.
Expect to see decision trees, gates, and milestones built into workflows. AI might be given freedom to optimize campaigns, but only within a set spend. It might update a website, but only in safe zones like FAQs. It might propose new ideas, but execution will still pass through human review.
The pattern will be partial autonomy with clear boundaries. AI moving quickly in the background, humans deciding where the lines are drawn.
Need help drawing the lines? Sign up for our AI Workflow Diagnostic, and we’ll show you how.