How Abid Chaudhry, Sr. Director of PMM at ServiceNow, built a system that feeds AI the right context every time

There's a question that's been circulating internally at ServiceNow, and I bet it's circulating at your company too: 

What's the point of product marketing now that AI can do so much of what we've traditionally done?

My ADHD hyperfocus has pushed me deep down the AI rabbit hole. I’ve become obsessed with figuring it out, and that obsession has turned into a strategy for using AI to accelerate PMM and drive faster time to value for our customers.

I’m going to share that strategy with you here. Here’s a preview of what we’ll cover:

• Why AI is putting pressure on product marketing's output
• What it means to own your "source canon", and why that's now the job
• How to break your messaging down into structured, machine-readable blocks
• Practical steps for getting started, with a real example from ServiceNow

The visible layer is under pressure

For years, product marketing’s job has been fairly consistent. You go figure out what's happening in the market, understand your ICP, and make that intel useful for sales, product, and engineering. Then you produce the content that brings it all to life: beautiful decks, data sheets, FAQs, battlecards, et cetera, et cetera.

That second part, the content production layer, is where things have gotten complicated.

AI has made it possible for a salesperson or a product manager to put together a pitch deck that's probably more polished than what you could produce in two weeks, and they can do it in two minutes. 

So, there's a creeping anxiety in the function. If the visible part of our job, the thing we've always been able to point to and say "look what I made," is now something anyone can do with a well-placed prompt, then what exactly are we here for?

The answer to that question isn't "less than before." It's the opposite.

The market is moving faster than our processes

While we've been wrestling with what AI means for our output, the market itself has accelerated in ways that compound the problem. 

Your customers are constantly shifting their thinking. Your competitors are moving into your space. Analysts’ positions are pivoting. Your own engineering teams, thanks to AI, are shipping features at a velocity that would have been unimaginable a few years ago.

Where you used to plan one launch every half year, now you're being asked to support three launches in a quarter. The messaging refresh cycle that used to feel manageable has become something you can't keep up with through traditional means.

And then there's a subtler problem that deserves more attention than it's getting: content convergence.

Think about it. Most of us are using the same handful of LLMs: Claude, Gemini, ChatGPT, and maybe one or two others. If you and your biggest competitor use the same back-end models, your content will start sounding the same. Outputs converge. Differentiation erodes. That's not a hypothetical; it's already happening.

I ran a quick experiment a while back. I built a script to scrape the product and solution pages of our top 25 competitors, then asked Claude to do an AI slop check to identify who was most guilty of just pasting AI-generated content straight onto their production website. 

The results were eye-opening. A lot of companies used almost identical language – word-for-word overlap in places – to describe what they claimed made them unique. 

When we ran ServiceNow through the same rubric, we scored a 6.6 out of 10, with 10 being the most generic. Our HR product area was closer to seven or eight. That was a wake-up call.

Better prompts aren't the answer

When people start thinking about how to solve for AI quality and consistency, the first instinct is usually that we need better prompts, more precise prompt engineering, and a well-organized prompt library.

That helps, but it's not enough. Here's why.

Even with the most carefully constructed prompts, if the back-end source material that you, your colleague, and your competitor are all drawing from is essentially the same, the outputs are still going to converge in ways you can't fully control. The prompt is just the instruction. The real question is: what are you instructing the AI to draw from?

This is where the real shift in product marketing's role becomes clear.

PMM’s job is to own the source canon

The argument I've been making at ServiceNow is that the purpose of product marketing in a post-AI world is to manage the market zeitgeist.

Zeitgeist, in this context, means the defining mood, pressures, and shifts that shape how your market thinks, how your ICP is evolving, and what your buyers care about right now. Product marketers are the best people in any organization to gather that signal, make sense of it, and translate it into something the rest of the company can act on.

But there's a second part to this, and it's the part that changes everything operationally: you need to own the source canon.

The source canon is the organizational truth. It's the stuff that isn't in Claude or Gemini. It's in your Gong call recordings. It's in last Tuesday's conversation with your product team about why they built a feature a certain way because three enterprise customers asked for it. That's proprietary context. That's differentiation that no competitor can access, because it lives inside your organization.