The discourse around AI in marketing has become difficult to navigate. In the last eighteen months I've heard AI positioned as a solution to everything from attribution to creative development to customer insight. Some of these applications are genuinely useful. Many are not. Most fall somewhere in the messy middle. On top of that, AI has also been positioned as the harbinger of doom, likely to wipe out thousands of creative roles.
I've used AI usefully in my own work, but I've also watched teams use it as a substitute for thinking and waste enormous amounts of money on the results. The challenge is separating the signal from the noise.
So here's what actually works, based on what I've seen succeed and what I've seen fail.
What AI really changes
Narrative development speed
Software product teams can spend 6-8 weeks on positioning. Three days of facilitation, weeks of back-and-forth refinement, another week of wordsmithing. The goal is to pressure-test assumptions about how the market is thinking and land on a narrative that feels both true and differentiated. I saw it firsthand when I worked at Peak.
Today, you can do that pressure-testing in days. You can generate five positioning frameworks in an afternoon and test them against market research in another afternoon. By the end of the week you've refined the winner. This isn't because the AI output is perfect. It's because having multiple options instantly reveals which ideas are actually working and which are just sounding good in a meeting room. Bad assumptions surface faster. The strategic judgment still sits with the person in the room, but they have way more evidence to base it on.
This is real and valuable. A positioning that would have taken a month to feel confident about can now be pressure-tested in a week.
Content scale without the coordinator role
At StickerYou, we run a full content operation: SEO content, product guides, email sequences, social media, case studies and more. Before AI, that meant either a content manager and multiple writers, or a constant stream of outsourced work that required management overhead.
Now a single strong content strategist can direct the work at scale. The AI generates the first draft. The strategist shapes the narrative. The output quality is substantially better than it would be if the strategist were doing all the writing themselves.
This is also real. You can do more with fewer people. If the strategist is good, the output quality improves.
Rapid A/B testing of messaging
If your headline isn't working, you don't wait two weeks for creative to come back with alternatives. You generate ten alternatives in twenty minutes, test them, and move on. Again, this is real and valuable.
Cognitive acceleration & automation
I love using AI as a thought partner more than anything. Nearly every project, task, or conversation is improved thanks to deep discursive back-and-forth with a machine. Use AI to challenge your thinking, research complex questions, and think through new angles on problems you might be stuck on.
What AI doesn't change
Attribution
I mention this because it comes up constantly in every conversation about AI and marketing. "AI will finally solve attribution." It won't. Attribution is hard because the underlying data is messy. Customers take non-linear paths. They touch multiple channels. Some interactions are offline. Some are dark social. No amount of AI fixes that.
What AI can do is help you model the uncertainty better. You still need clean data inputs and sensible assumptions. Garbage in, garbage out still applies.
Creative judgment
This one is subtle but here's the key: AI is good at generating options. It's terrible at knowing which option is actually right. I've watched teams use AI to generate fifty creative concepts and pick the one that "feels" best. Then watch it completely miss the mark in market.
The AI didn't fail. The human judgment failed. Humans are still the only thing that can make that call. This matters because a lot of the current conversation is "AI will now do creative." It will do the first draft and the iteration. A real creative strategist still has to make the final call. Someone who understands your audience, your category, your brand voice. And ultimately the real test is whether that creative actually generates results.
Customer insight
This is the one I see most overstated. "AI will now understand your customers better." Here's what's actually happening: you feed the AI your existing customer data and it finds patterns in that data. But if your data is biased (most is), if you're not asking the right questions (most teams aren't), or if you're using the insights to confirm what you already believed (all of us do), the AI output is simply sophisticated nonsense.
Real customer insight comes from actual conversation with customers. Qualitative research. Win/loss interviews. Customer advisory boards. AI can help you scale the analysis of that research. It can't replace the research itself.
What actually matters
The through-line is simple: AI is powerful at increasing velocity on work you already know how to do. It's weak at replacing judgment. It's useless at telling you what questions to ask in the first place.
The marketing teams I see winning with AI are not the ones who think AI is a replacement for strategy. They're using AI to move faster on execution so they have more time for strategy. They test positioning frameworks faster. They scale content. They test messaging. They focus on what's actually resonating rather than what feels right.
The teams losing with AI are the ones who think it replaces the hard work. They use it to generate content without a strategy. They use it to test fifty variations of the same bad idea. They use it to confirm biases rather than challenge them.
If you're in marketing in 2026, the question isn't "how do I use AI?" It's "how do I use AI to buy myself time for better judgment?" That's where AI is actually powerful.