> ## Content Index > Fetch the complete content index at: https://notes.jasonjhonson.com/llms.txt > Use this file to discover other available public pages before exploring further. # Just when we figured out what product marketers do … - URL: https://notes.jasonjhonson.com/product-marketing-ai/ - Published: 2026-10-17T02:07:23.000Z - Updated: 2026-10-17T02:07:23.000Z - Description: AI changed the job. - Author: Jason Jhonson - Tags: Product Marketing, AI When I first became a product marketer at Yahoo, it honestly wasn’t 100% clear to me what a product marketer was. It was described more as a problem than a function - we have a complex competitive product, and we have thousands of people in dozens of countries around the world selling it… help us make that work. The job made sense to me quickly enough. Technology companies had become very good at building things, but the connection between what they built and what the market understood about those products was surprisingly loose. Part of that was cultural. Technology companies tend to believe that if something isn’t selling, the answer is to improve the product: add the feature, fix the workflow, make the technology better. Consumer businesses don’t always have that luxury. If the cereal is already sitting on the supermarket shelf, you have to get better at packaging, positioning, and selling the thing you already made. Technology has always had the luxury, and sometimes the curse, of believing the next product change might solve the commercial problem. At Yahoo, that problem became almost physical. We had people selling and operating products all over the world. Every quarter, teams would make what amounted to a pilgrimage to Sunnyvale and try to get time with product managers and engineers over a cup of coffee. They had a country that needed something, a partner asking for something, or a customer behavior they thought mattered. They wanted to make their case directly to the people building the product, but over time, the product people were legitimately hiding from them (people can really only drink so much coffee). It wasn't because the market input was unimportant but more because it was a terrible information architecture. There were too many people selling things and too many people building things for every useful signal to travel through a series of person-to-person meetings. The way I've come to think about the split is like this: In commercial strategy and execution, product management represents, in a broad sense, everybody in the company who builds things, and product marketing represents everybody who sells things. The two functions have to stay very close so the company can keep answering three basic questions: What are we building? Who is it for? Why should anyone choose our product over the alternatives? ![Even “support functions” like HR and finance are building and selling a “product”](https://storage.ghost.io/c/ec/41/ec4178b6-0a2a-4b63-a5d1-a95a2fdc0a72/content/images/2026/10/A-Simple-Theory-of-Companies--2-.png) Even functions like HR and Finance are building and selling a “product” My time at GitHub made me rethink that model. Yahoo and PayPal had hundreds of millions of users. Nobody personally knew "the Yahoo customer" in any meaningful way. We relied on research, segmentation, usage data, local teams, and a lot of synthesis. GitHub was different. The people building the product often knew the people using it. They knew open source maintainers by name. They understood how developers worked because many of them came from the same community. In that environment, inserting a product marketer between the builder and the customer could actually make the company less informed. A product marketer arriving with a generic process and announcing that they had synthesized the customer was not really adding anything. That experience left me wondering where customer and market understanding should live inside a technology company. AI makes that question much more interesting and has changed how I answer that question. ## Product marketing has always been a sampling problem A good product marketer has traditionally spent a lot of their time trying to reconstruct reality. You sit on sales calls. You interview customers. You read win/loss reports. You talk to product managers and engineers. You commission research. You look at competitors. You audit the website and sales decks. If you're diligent, you do enough of all of those things to form a coherent view of what the product can do, what customers care about, and whether the company is explaining itself well. But even a very good team is sampling. AI changes the constraint. It is becoming practical to maintain something much closer to a continuous record of both product truth and market truth. On the product side, the system can consume documentation, roadmaps, support material, release notes, pull requests, and changes in the code base. That doesn’t mean that a model should reverse-engineer the product from source code and declare itself an expert. The more useful application is closer to a change log. What changed? What new capability does that create? Does it materially change how we should describe the whole product? That last question matters because one of the easiest failure states in product marketing is becoming a launch factory. New feature. New feature. New feature. Customers generally don't buy the newest feature of a product they didn't understand in the first place. The job is to decide what each change means in context. Does it reinforce the existing story? Make a new use case credible? Solve a problem we previously had to work around? Change who the product is for? On the market side, the change may be even larger. A product-marketing system can now potentially consume every sales-call transcript rather than the five calls a product marketer had time to attend that week. It can look across the language buyers use to describe their problems, competitors they mention, objections, support interactions, form fills, website behavior, and what happens after particular messages show up in a conversation. That is not just doing the old job faster. It is a much fuller pull of the signals the market is producing, much closer to real time. ## The hard question is still: what do we do about it? If AI can gather more of the evidence, summarize it, classify it, and keep the basic product-marketing artifacts current, what is the human product marketer for? The answer sits in a question technology companies have struggled with forever: **Is this a product problem or a communication problem?** A prospect tells you the product doesn't solve an important need. Maybe the product genuinely doesn't solve the problem. Or maybe it already does, but the market doesn't understand that it does. Getting those two confused is extraordinarily expensive. Treat a communication problem as a product problem and engineering can spend months building features customers didn't need. Treat a product problem as messaging, and the company can spend months finding increasingly clever ways to describe something customers still won't buy. AI can give the product marketer and product manager a much better body of evidence for that decision. The machine shouldn’t be making the decision on the company's behalf. That requires judgment about which signals matter, which customer to optimize for, what the company is willing not to be, and whether following market feedback too literally would improve the product or simply make it look like everybody else's. A system trained to continuously adapt messaging to whatever it hears from the market seems to me like a fairly efficient machine for producing homogenization. It might become very accurate. I'm less convinced it becomes interesting. ## Should we collapse product marketing back into product management? There is an obvious organizational response to all of this. If AI can do so much of the listening, synthesis, and artifact production, why not simply collapse product marketing back into product management? I understand the appeal. I would still argue against it in most reasonably complex technology companies. Managing everybody who builds things is already a hard job. Product managers live with roadmaps, engineering constraints, backlogs, internal stakeholders, releases, tradeoffs, and the thousand decisions required to move a product somewhere intentional. It is very easy for that role to become overwhelmingly inward-facing. Having product managers spend time with sales or join customer calls is useful. It is not the same as having someone whose job is to look outward continuously. I still like the idea of two tightly connected people looking in opposite directions. The product manager is responsible for understanding what we've built, what it can actually do, what we're building next, and the constraints involved in getting there. The product marketer is responsible for understanding customers, sellers, competitors, buyer behavior, and the environment into which the company is trying to sell. The two reconcile those views and keep the rest of the organization aligned. That alignment matters because bad communication creates bad feedback. Sales decks drift. Reps promise things the product does not do. New capabilities never make it into the story. Marketing emphasizes a benefit that nobody mentions in successful deals. Then the company looks at the resulting market response as though it were a clean signal about the product. It isn't. If the market wasn't shown the product accurately, the feedback is contaminated from the start. ## How the product marketing team changes My guess is that many companies will eventually need fewer people doing traditional product-marketing tasks (which makes me, as a former product marketer, sad), while the people who remain become more senior and more central to commercial strategy. AI can already do or materially accelerate a lot of the mechanical work: first-pass competitive research, battle-card production, meeting synthesis, messaging audits, launch artifacts, enablement updates, and the work of keeping information current. The human job shifts toward designing and maintaining the system, interpreting contradictions, deciding what the evidence means, and setting a point of view for the company. The analogy I keep coming back to is an editor-in-chief. The system can gather a lot of information and produce a lot of material. Someone still has to decide what is important, what is distinctive, what is true enough to stand behind, and what should be left out. That may also change who companies hire. Unfortunately, this brings me to *Armageddon*. The movie's basic premise is that it is easier to teach expert oil drillers enough astronaut skills to complete a space mission than to teach astronauts to become expert oil drillers. I feel like they should have thought that through better, but, like so many things, it made me think about product marketing. ![I don’t remember the movie that well … I think something like this happened.](https://storage.ghost.io/c/ec/41/ec4178b6-0a2a-4b63-a5d1-a95a2fdc0a72/content/images/2026/10/Armageddon.png "Armageddon") I don’t remember the movie that well … I think something like this happened. If AI makes much of the mechanical craft easier to execute, the harder thing to teach may be the market. The function still needs to be led by someone who deeply understands product marketing: positioning, segmentation, packaging, launches, evidence, feedback loops, and how those pieces are supposed to work together. But the people around that leader can increasingly start from deep customer and domain knowledge rather than perfect mastery of the traditional PMM toolkit. They need enough technical fluency to understand what is being built, and they need to understand what it is actually like to be the person who buys the product. What worries them? What do they depend on? What language do they use when nobody is feeding them a marketing message? What makes an idea credible or ridiculous inside that industry? AI can teach someone how to make a battle card considerably faster than most companies can teach a generalist the accumulated culture and judgment of a complicated market. ## Product marketing may finally become easier to measure I've been asked more than once for a quantitative measure of a good product marketer. I've never found revenue to be a particularly satisfying answer. Product marketing influences revenue, obviously, but it doesn't independently control the product, sales call, website, or most of the other things between positioning and a purchase. AI makes the operating objective easier to observe: **fidelity and speed**. At every important customer touchpoint, are we communicating what we intended to communicate? And when the product, strategy, or market changes, how quickly and intentionally does that change make its way through the organization and back out to the market? Historically, companies answered those questions through periodic audits and a lot of qualitative assessment. Now it is increasingly possible to observe them continuously. Are salespeople using the intended value proposition? Where are they improvising? Are old claims still circulating? Did the new capability make it into customer conversations? Is the website describing the product differently from the sales team? When the market responds, does that evidence get back to product and product marketing quickly enough to matter? If that system works, I would expect the downstream business measures to improve too. Better conversion, sales effectiveness, adoption, and ultimately more wins. But those are shared outcomes. The product-marketing operating system itself should be measurable before the quarter closes. AI may remove a lot of the work that made product marketing busy without removing the work that made it valuable. That seems like progress.