In this article I show you how I actually use AI for product marketing day to day, where it earns its place, and where it quietly falls apart. The honest version, not the polished one.
The first time someone suggested I “leverage AI for competitive intelligence,” I cringed. Not because AI has no place there, it does, but because the line treated AI like a switch you flip instead of a rethink of how you spend your week. Most advice about AI for product marketing makes the same mistake, and it’s why so much of it goes nowhere.
I’m a Certified AI Product Manager and a PMM lead, which is less a flex than a way of saying I’ve spent enough time inside these systems to know when the hype outpaces the reality. And in product marketing, there’s a lot of hype.
The honest version is simpler. AI is a tool that lets me move faster on specific, bounded tasks, and it falls flat on the parts of the job that actually decide whether a launch works. Knowing which is which is the whole skill, so let me walk through where it earns its place and where it doesn’t.
What AI for product marketing actually changes
The pattern across every use below is the same. AI compresses the mechanical half of the work, the reading, summarizing, drafting, and formatting, so the time it gives back goes into the judgment half, the deciding. If you want the bigger argument for rebuilding the whole function around that split, I made it in AI native product marketing. This piece is the ground-level version: the tasks, the real wins, and the limits I keep running into.
Synthesis at scale
The biggest shift for me has been customer research synthesis. I used to spend hours working through call recordings, support tickets, and interview notes, and the task was never intellectual. It was mechanical: listen, take notes, find the pattern, repeat. Now I feed the transcripts and ticket threads into an AI tool and ask it to surface patterns first, not insights. Which problems come up again and again, what language customers use, where they get stuck.
That saves me close to ten hours a week, and it catches things I would have missed sampling five calls instead of fifty. When you can process the whole pile, the weak signals get loud enough to notice. The catch is that I still have to synthesize the synthesis. The tool can tell me customers mentioned integration complexity seven times, but it can’t tell me whether that’s a real blocker or background friction, or what it means for positioning. That judgment is the part I’m paid for.
Competitive intelligence that doesn’t make things up
Competitive intelligence is where AI gets oversold the hardest. “Monitor your entire competitive set continuously,” the pitch goes, and the reality is messier than that. What I actually do is have Claude scan competitors’ websites, blog posts, and job listings each week and flag what changed. New hires hint at product direction, job descriptions telegraph the features they’re building, and a rewritten pricing page is always worth a closer look.
But I don’t trust the interpretation, I read the source myself. The AI is a filter that tells me where to look, not a narrator that tells me what it means. It’s also useful for tracking how rivals talk about the problem: I collect their messaging, feed it in, and ask for the claims and positioning patterns underneath. None of that replaces talking to customers, but it does mean that when a prospect mentions a competitor, I already know what that competitor has been saying.
Content drafts, with the judgment kept back
I write a lot: product launches, sales enablement, internal alignment docs, the brief language that goes to design, and this blog. I used to draft linearly, opening first and building from there. Now I get a rough first draft out of AI in a few minutes, which beats staring at a blank page, and then I tear it apart. I write a detailed prompt with the context, the tone, and the exact point I want to land, feed it to Claude or another model, and rewrite the weak parts. Sometimes I throw the whole thing out and start again.
What I don’t do is hand the raw output to someone and call it finished. That’s the mistake I keep seeing juniors make, and it shows. The writing comes out competent and hollow, with no point of view and no risk in it, the prose equivalent of beige. I use AI here the way I’d use a research assistant: it does the work that would be dull for me to do, and I keep the judgment.
Message testing and positioning iteration
This is where I’m still learning. I’ll have AI generate positioning variations on the same core message, the feature framed as time savings, then as faster decisions, then as lower risk, and stress-test which one lands for which reader. The gap in the current tools is that the testing isn’t as good as a real conversation. I run surveys, but for a positioning call that actually matters I still get on the phone. AI helps me produce the variations faster; the customer still tells me which one is true. For anything important I build it on my message map first, so the variations all ladder up to the same positioning.
The parts AI can’t touch
A lot of the job sits outside what AI can help with, and accepting that has been freeing. Building cross-functional trust is human work: sales has to believe you understand their deals, and product has to know you’re carrying the customer’s voice rather than just relaying requirements. That trust comes from repeated interaction, and no tool compresses it.
Strategic judgment is the other no-go zone. “What should our positioning be” isn’t a question a model can answer. It can gather the inputs and lay out the options, but the call about who you want to be in the market, where you’re actually defensible, and which bets are worth making has to come from you. Customer empathy is the same. An AI can read a thousand tickets and report that customers are frustrated, but empathy means sitting with that frustration long enough to let it change how you think, and that’s a human problem, not a data one.
Where I’m still figuring it out
I’ve started using AI as a sparring partner for strategy, not to generate it but to pressure-test it. I’ll write out a positioning thesis, have the model argue against it, and then make myself answer. Some days that sharpens the thinking; other days it’s productive procrastination dressed up as work, and I’m still learning to tell the two apart. I’m also curious whether AI can help with internal storytelling, taking a complex product direction and making it land emotionally with the team, but so far the output stays too surface-level to trust with something that matters.
The real shift
The headline isn’t that AI is coming for product marketing. It’s that the parts of the job built on raw processing are getting cheaper and faster, which makes the parts built on judgment, taste, and relationships more valuable, not less. If you were hired to write competent copy and build clean decks, the floor is rising toward you. If you were hired to decide who the product is for and why they should care, AI just handed you back the hours you used to lose to the mechanical work. The tools in your stack will keep getting better at the grunt work, and your job is to spend the time that frees up on the work only you can do.
So yes, AI is automating the grunt work, and that’s good for the craft. Less time in spreadsheets, more time thinking. The product marketers who do well over the next few years won’t be the ones who resist it, or the ones who hand it the wheel. They’ll be the ones who use it to move faster on the mechanical work and spend the difference on the judgment. Not less work. Better work.
Frequently asked questions
What’s the best way to start using AI for product marketing?
Start with one bounded, mechanical task you already do. Customer research synthesis is usually the easiest win. Feed it real inputs like call transcripts and tickets, ask for patterns rather than conclusions, and check its work against the source. Once that saves you real time and you trust the output, expand to the next task. Don’t try to AI-enable your whole function at once.
Which AI tools do you actually use?
I use Claude and Gemini Pro for writing and synthesis, plus a couple of competitive intelligence tools that run AI under the hood. The specific tool matters far less than knowing how to prompt well, and knowing when not to use one at all.
Will AI replace product marketers?
It replaces tasks, not the role. The mechanical work, drafting, summarizing, formatting, is what’s getting automated, and that was never the reason you were hired. The judgment, the positioning calls, the cross-functional trust, and the taste only get more valuable as the rest gets cheaper.
I help B2B SaaS companies fix their go-to-market when positioning is unclear, launches don’t land, and sales can’t explain what makes them different. Ten years doing exactly this, from scale-up to enterprise. Contact me at zackalami.com/#contact.
