AMA: UiPath Sr. Director of Product Marketing, Kuber Sharma on AI and Product Marketing

The clearest wins for our team at UiPath have been in synthesis and first-draft speed, and the gains are actually measurable. On the research side, we used to spend 3 to 4 days collating competitive intel before a major product launch. We were pulling analyst reports, reviewing call recordings, synthesizing win/loss data. That is now a 4-hour job. The quality of the synthesis is often better too, because AI can hold 40 documents in context simultaneously while a human can realistically hold abou ...Read More

Neel and Paul both hit on something important: the narrative has to absorb AI, not be replaced by it. The principle I have found most reliable is this: before you introduce any AI claim, do a customer language audit of your existing user base. Pull 20 to 30 support tickets, renewal call transcripts, and QBR notes from customers who have been with you for 2 or more years. Look for how they describe the job your product already does for them. That language is your anchor. At UiPath, when we integr ...Read More

Gray, Vikas, and Dee are all pointing at the same thing: the claim "AI-powered" has become meaningless as a differentiator. The follow-up question is what you do instead. The approach I use is specificity as the differentiator. When every vendor in your category is running the same AI claim, differentiation lives in the mechanism and the proof, not the label. Not "AI-powered workflows." Instead: "AI that compresses a 3-day competitive analysis to 4 hours, measured across enterprise PMM teams." O ...Read More

The biggest shift will not be in what PMMs produce but in what they spend their time on. Over the next 12 to 24 months, I expect the execution layer of PMM work to become largely AI-assisted: first-draft messaging, competitive monitoring, win/loss synthesis, battlecard updates, sales enablement content, persona documents. The time cost of these tasks drops by 50 to 70%. For teams with enough insight to know what good looks like, that is a significant unlock. What does not change: the judgment la ...Read More

Neel, Charlotte, and Ahmed have covered the "AI as a workflow tool" angle well. The frame I would add is about sequencing. For PMMs at companies where AI is not central to the product, there are actually two separate conversations happening, and conflating them creates confusion. The first is internal: AI as a PMM productivity tool. This works regardless of what your product does. Competitive intel synthesis, customer language extraction from call transcripts, messaging QA, battlecard generation ...Read More

The most underused AI application in product positioning is customer language synthesis, not content generation. Here is what I mean. When we rebuilt Tableau's positioning framework, the hardest part was bridging the gap between how we talked about the product and how customers described the problem they were solving. That gap is where positioning goes wrong. AI closes it significantly faster than before. With UiPath, before major positioning work, I now run 30 to 50 customer call transcripts th ...Read More

The prompts worth automating are the ones with a predictable input structure, a consistent output format, and high repetition. If you have run the same prompt manually more than 5 or 6 times this month, it should be a workflow, not a one-off. In practice, this breaks into a few categories for PMM teams. Synthesis prompts are the highest-value ones to automate: summarizing customer call transcripts, extracting competitor claims from announcement posts, pulling top themes from win/loss interviews. ...Read More

The framing I find most useful here is distinguishing between loops that AI makes faster versus loops that AI makes possible for the first time. The loops AI makes faster: win/loss synthesis feeding back into messaging, competitive monitoring feeding into battlecards, customer call analysis feeding into ICP refinement. These existed before. AI compresses the cycle from monthly or quarterly to weekly. At UiPath, our competitive battlecard refresh went from quarterly to monthly because the synthes ...Read More

Yes, and the ones worth building are different from the ones most people start with. The first GPT I built was for competitive battlecard generation. I fed it a structured prompt with our product positioning, key differentiators, win/loss themes, and a template format. Then I gave it a competitor's website, their G2 reviews, and their most recent announcement. The output was a first-pass battlecard that took about 20 minutes to review and refine rather than 2 days to write from scratch. It now r ...Read More

The short answer: you don't need to write code to be an effective AI-augmented PMM, but you need to understand the architecture well enough to ask the right questions of people who can. The skills that have actually mattered in my work: prompt engineering at a practical level. Understanding what temperature does (higher gives more creative and variable output, lower gives more precise and consistent output), knowing when to use structured outputs versus free-form generation, and understanding wh ...Read More

The honest answer is that my stack has iterated a lot over the past 18 months, and a few tools have survived every round of reevaluation. For synthesis work, Claude is my primary tool. I use it for compressing large volumes of customer call transcripts, analyst reports, and win/loss interviews into structured positioning inputs. The context window matters a lot for this use case. At UiPath I am often working with 30 or 40 documents simultaneously, and Claude handles that without degrading. For r ...Read More

The tools that go org-wide tend to solve a workflow problem that is already causing friction before the tool shows up. When someone on the team is doing a workaround repeatedly and then finds a tool that eliminates that workaround, adoption spreads without a rollout plan. That bottom-up pull is a reliable signal that a tool will stick. At UiPath, the tools that have gotten traction fastest are the ones where we could point to a time-before versus time-after that was obvious to anyone watching. W ...Read More