Over the past year, we’ve seen the same process play out again and again. Someone opens an AI tool, they describe the product from memory, generate a few value props, polish the language, and send it to their boss. Then it ships.
One thing is missing: the customer.
We often ask marketing teams, “When did a customer’s words last appear in your copy?” Not a polished paraphrase. Not an objection imagined for a persona. Something a real customer actually said. And the pause that follows usually tells us everything.
More and more companies are creating marketing from inside a bubble. And AI is making it easier for that bubble to grow.
The bubble isn’t new. What’s new is how sealed it’s gotten.
This pattern started long before generative AI. In its 2019 B2B research, Content Marketing Institute found that only 42% of content marketers talked with customers as part of their audience research. Roughly three-quarters relied on sales feedback or website analytics, and 65% used keyword research.
In other words, marketers were already more comfortable studying signals about customers than speaking with them directly.
Even so, the work still required marketers to talk to people. You needed a quote for a case study, so you called the customer. You needed to understand an objection, so you asked the sales rep. It was slow and sometimes annoying. It was also quietly essential.
Generative AI removes much of that friction. A draft can now appear before anyone talks to anyone. The assumption comes back instantly, fluent, polished, and neatly formatted. It looks like insight. But often, it is simply the original assumption just rewritten with more confidence.
Researchers are beginning to measure the downstream effect. In a Science Advances study, writers with access to up to five AI-generated ideas produced stories that scored 8.1% higher for novelty and 9% higher for usefulness. But the AI-assisted stories were also 8.9% to 10.7% more similar to one another.
Better one by one. Less distinctive as a group.
That tradeoff feels familiar to anyone scrolling through a B2B feed right now: polished posts, competent writing, and the same handful of ideas repeated in slightly different language.
Buyers seem to feel it too. In the 2024 Edelman–LinkedIn B2B Thought Leadership Impact Study, only 15% of decision-makers rated the thought leadership they consumed as very good or excellent. The same research found that the strongest quality signals included supporting data, concrete guidance, and case studies, all things that require contact with the world outside the marketing team.
An inside-only process can produce more content. But it cannot produce that evidence.
Where the best stories are already happening
One observation changed how we work: at most companies, the best marketing stories have already been told. Marketing just never heard them.
They surfaced on a sales call when a prospect explained what was making them hesitate. In a support ticket where a customer described the product’s value better than the homepage did. Or in an exit note that finally made clear why an account left. These moments happen every day. But unless someone captures them, they stay buried in a CRM, support platform, or call recording. Then they disappear.
That matters because direct customer conversations are already rare. Gartner’s research on the B2B buying journey found that buyers spent only about 17% of the purchase process meeting with potential suppliers. When several vendors are competing, each one gets only a small piece of that time.
That makes every real conversation valuable. Yet companies often treat what comes out of those conversations as sales exhaust instead of marketing source material. It is one of the most expensive missed opportunities we see.
We call the alternative story extraction: deliberately looking inside sales, support, onboarding, and churn for the language, objections, proof, and experiences that marketing should be built around.
We saw this firsthand with Olde South Bulldogges. The company was relying on paid ads while valuable material from sales and post-sale conversations went unused. We rebuilt the marketing around what customers were already saying. Common sales questions became the structure of the website. Sales conversations, thank-you notes, and care instructions became organic content. Real customer language replaced keyword guesses.
Over time, the company reached the top national search position in its category and stopped paid advertising entirely.
We did not invent a new story. We found the one customers were already telling.
What we see working
The teams that break out of the bubble don’t do it with a research budget. What they share is a listening routine and a low tolerance for paraphrase.
They sit in on sales calls. Five is enough to start, live or recorded, writing down exact phrases: how prospects describe the problem, what they’re afraid of, what they call the product. Verbatim only. The moment it gets summarized, it’s back in the bubble.
They read support tickets like source material. The last 90 days, tagged for two things: moments where a customer described value in their own words, and questions that keep recurring. The first turns out to be messaging. The second turns out to be a content calendar.
They talk to wins AND losses. Three of each, asking what they read or heard that moved them, and what was missing that would have. Losses talk more freely than most teams expect, and they know exactly why the deal was lost. The companies that formalize this see it show up in results: in Klue’s win-loss research, 70% of teams running win-loss programs feed the insights into messaging, and 56% report higher win rates.
Then they ship one asset built only from extracted language. One page, one email, one article where every claim traces to something a real customer said or asked. Compared honestly against what would have been generated instead, it isn’t close.
This is also, we’ve noticed, where AI finally starts paying off. AI amplifies its inputs. Fed the market’s average, it returns the average, faster. Fed fifty verbatim customer phrases and a real loss reason, it finally has something worth amplifying.
The pattern is simple: teams that extract before they generate don’t sound like everyone else. They sound like their customers. The best lines are already being said. The job is to catch them.