Most of what I read about AI in marketing is somebody’s pitch deck dressed up as a hot take. Either the tools have changed everything and you’re already behind, or they’ve changed nothing and we’re all hallucinating together.
I wanted a quieter version of the story. So I sat down with three working marketers — Jordan, a director of product marketing at a B2B SaaS company in the mortgage space; Maria, a senior product marketing manager at a software company; and Lauren, a marketing manager at a SaaS company in property management — and asked them what AI has actually done to their day-to-day.
Three different roles. Three different companies. Three different stories. And exactly one thing they all agreed on.
What they agreed on
All three of them use AI every day. Not occasionally, not for special projects — AI is woven into the work. Jordan put it bluntly: “everything starts with an AI prompt.” Maria’s first move on most assignments is to open Claude. Lauren has it touching her email copy, ad copy, social, blogs, dashboards, and even her website audits.
All three of them also landed on Claude as their primary tool. Two of them — Jordan and Maria — actively abandoned ChatGPT after using it for a while. Jordan’s reason was that ChatGPT “feels more like a ‘yes man’ in certain areas where it’ll tell you what you want to hear, whereas Claude seems to be more straightforward.” Maria’s read was that ChatGPT is better at messaging and Claude is better at executing tasks. Lauren still uses ChatGPT for “quick things,” but the heavy lifting happens in Claude Cowork, where she has connectors and a browser extension wired in.
That’s a notable pattern in a three-person sample. I wouldn’t extrapolate to the whole market from it, but if you’re a marketing leader still standardizing on a single LLM, it’s at least a signal worth checking.
What was much more interesting than the tool agreement was what happened when I asked each of them what their relationship with AI actually feels like now. That’s where the three stories split.
Jordan’s story: the accelerator
For Jordan, AI is mostly a multiplier. He uses it for competitive research — he’ll feed in a competitor URL or piece of collateral and ask for a side-by-side against his own product, and what comes back is a “fully fleshed-out comparison chart.” He uses it for copywriting, positioning, messaging statements, email, launch comms, slide decks, internal training. He puts the work in on the prompt itself — target market, persona, product vs. value framing, writing style — and he says AI gets him “about 95% of the way there.” The last 5% is his.
His summary line was the cleanest I heard in any of the three conversations:
“AI has made taking products and messages to market exponentially faster.”
He doesn’t think AI is going to replace marketers. He thinks marketers who don’t get good at it “will become obsolete.” That’s a real distinction, and I think he’s right about it.
Maria’s story: the slow erosion
Maria’s story was quieter, and more honest about what’s actually hard.
She uses AI the same way Jordan does — to jumpstart slides, to dig into customer data, to draft. But when I asked her what part of her job looks most different now, she didn’t say “faster.” She said it has gotten harder to sit with her own thoughts.
“AI has made it harder to have personal critical thinking, because it’s so easy to just have AI think for you and then decide whether you like what it’s saying or not.”
She walked me through a recent example. She was building a security one-pager. Instead of starting with the task — what’s this for, who’s it for, what story should it tell — she opened Claude and asked it to write a draft so she could “see what Claude would give me.” Then she started riffing off what came back. A long way into the process, she realized she didn’t actually agree with the story the draft was telling. She had to scrap most of it and go back to step one: sitting with the problem on her own.
The honest version of her current state, in her own words: “Rather than spending time ideating and creating, now I spend more time fixing what AI has given me.” On the same one-pager, she asked AI for a tweak and it destroyed the formatting — and fixing it took longer than just writing the page herself.
Two things stand out in Maria’s story. First, the failure mode isn’t AI being wrong. It’s AI being plausible enough to make you stop thinking. Second, the most useful thing AI was supposed to do for her — pull together signal from customer interviews, web pages, and CRM data — has not really shown up, because the underlying data in Salesforce and the other systems isn’t clean enough for AI to find anything in it. The model doesn’t fix the upstream mess.
Lauren’s story: the enhancement
Lauren’s story sits between the other two. She’s pragmatic about what AI does and doesn’t do.
The thing she lit up about was reduced tedium. She walked me through a website audit she runs: tools like Claude Cowork can use the Google Analytics connector and the browser extension to read pages, pull data, and fill out the audit columns naturally — work that used to mean opening every page by hand, exporting reports, and matching numbers with formulas. Content creation is the other big shift. Drafts that used to take her several days to write and route through review now come together in minutes, with what she described as “minor review.”
But she’s blunt about the limits.
AI hasn’t fully taken anything off her plate. Reporting still needs a human eye — she gave the example of dashboards that count direct sessions full of spam the model doesn’t know to filter out.
Ads still need a human in the loop. Brand voice still does, too, because some phrases are “very obviously AI.”
When I asked her where she thinks most marketers are getting AI wrong, she didn’t hedge: “Marketers are trying to fully replace certain roles by fully automating when AI just isn’t advanced enough to get there yet.” Her one-line takeaway, which I keep coming back to: “AI is a great enhancement to a marketer’s role, but not a replacement.”
The pattern underneath the three stories
You can read those three conversations and walk away with “well, it depends on the person.” That’s fair. But there’s a sharper pattern, and it shows up in two questions.
The first question: who is the editor and who is the starter? Jordan walks in with a tight prompt, an opinion about the audience, and a clear point of view. AI fills in the middle; he edits the last 5%. Lauren does the same thing with tedious work — she stays the editor, AI does the legwork. Maria’s hard moment came when she let AI be the starter instead of the editor. The moment AI is the one introducing the idea, your role shifts from author to reviewer. Some work survives that shift fine. Some work — anything where the story matters more than the words — falls apart.
The second question: is the input clean enough for the model to help? Maria’s Salesforce point is the one most leadership teams I talk to are quietly dodging. AI does not fix the upstream mess. If your customer data is sitting in three systems with inconsistent fields and half-loved CRM hygiene, AI is going to confidently make that mess look more organized than it actually is. You’ll get a polished synthesis of noise. The teams getting real lift from AI right now are the ones who already invested in clean inputs — accurate analytics, real customer conversations, structured product information. Without that, you’re just rendering the same questions faster.
What I’d tell a marketing leader reading this
A few things, in plain terms.
First, speed is not the only metric, and it’s not even the most important one. Jordan’s “exponentially faster” and Maria’s “I now spend more time fixing than creating” can both be true on the same team in the same week. If you’re measuring AI ROI by output volume, you’re measuring the easy thing. Measure whether the work is getting better — clearer narratives, tighter positioning, fewer “what do you actually do?” cycles with prospects, shorter time from customer insight to in-market message. Output volume without those is just a faster way to make the same noise.
Second, headcount fantasies are dangerous. Lauren said it directly, and it tracks with what I see in the field: the marketers trying to replace whole roles with automation are running into the same wall every time. AI is not advanced enough to do the judgment part of marketing well, and the judgment part is most of the job. Plan for AI to make a smaller team more capable, not for it to remove the team. The teams I’ve watched fail with AI weren’t the ones who under-adopted it. They were the ones who eliminated the people who would have caught the model’s mistakes.
Third, the new skill isn’t “use AI.” It’s “stay in the chair.” The marketers I’d hire today are the ones who can still think when the tool is turned off — who can sit with a brief before they open Claude, who know what story they’re trying to tell before they ask for a draft, who notice when the model is being plausible instead of true. AI rewards taste. If anything, it rewards taste more than it used to, because the floor on competent-looking output has gone up. The ceiling is still set by the human who edits it.
Fourth, fix your inputs before you scale your outputs. If your customer data is messy, your analytics is full of spam sessions, and your product information lives in five places, slow down. The cheapest thing you can do for your AI strategy this quarter is clean your data. Everything you do with AI on top of clean inputs compounds. Everything you do on top of dirty inputs just multiplies the confusion — and you’ll pay for it twice, once in tokens and once in the hours someone spends untangling what the model produced.
The honest read
Jordan, Maria, and Lauren are three reasonable people doing real work, and they don’t fully agree on what AI is. That’s the honest state of the industry right now. Anyone telling you they’ve got it figured out is either selling you something or hasn’t tried to use the tools on a hard problem yet.
The marketers who are going to win the next two years aren’t the ones with the longest stack or the fanciest prompts. They’re the ones who treat AI like a sharp junior teammate — useful, fast, occasionally wrong, never the one with the final say. They edit. They stay curious about the inputs, not just the outputs. They keep their judgment turned on.
Or as Lauren put it, in the single best sentence anyone said to me across the three interviews:
AI is a great enhancement to a marketer’s role, but not a replacement.
I think she’s right. And I think the marketers who internalize that — and run their teams accordingly — are going to look back on this stretch and realize it was the cheapest, fastest leverage they ever had, precisely because they refused to outsource the part that actually mattered.