“PhD-Level AI Doesn't Mean Better Enterprise Results": Kelly Xu on What Actually Matters in B2B AI

Welcome to the latest episode of Silicon & Spice, where tech leaders share life's unique flavors! Today I'm joined by Kelly Xu, Head of Industry Marketing at Glean, an enterprise AI company, with previous marketing leadership roles at Snowflake and DocuSign.

What was your path into product marketing?

Like many product marketers, I started right out of business school. In 2016-2017, drones were the hottest technology, and I got introduced to a drone company. At first, I didn't know what product marketing was or what I wanted to do, but they pitched the idea: "How about you start as our first product marketer?" I said sure, that sounds interesting!

That's how I got into product marketing—pretty lucky. And then I found out this is where my passion lies and what I want to do long-term.

You're currently Head of Industry Marketing at Glean, and previously worked at Snowflake and DocuSign. How different are startup marketing versus more mature company marketing roles?

I've worked at companies ranging from 100-150 people to public companies, and it's actually pretty similar. The nature of work for a product marketer remains the same: understand your product, understand your customers deeply—their use cases, pain points, and what value you can bring. You also need to understand the competitive landscape and your ecosystem.

I don’t think it matters that much whether you're at a startup or established company. Obviously at startups you have a wider scope, while at bigger companies you go deeper into particular areas. Different companies have different ideas of what product marketers own, so your scope and team dynamics vary more based on organizational culture than company size.

The work pace is similar too. Both DocuSign and Snowflake were pretty fast-paced. If you're at a company that's generally fast-paced or in a rapidly changing market, the pace is fast regardless of size.

What does a typical day look like for you as a marketing leader?

I think of it more as a typical week with different tasks. On Monday, maybe I'm doing enablement for our sales team—sharing new messaging or use cases. Then I catch up on industry trends and what's happening in the market around our customers and partners.

The next day might involve events I host virtually or in-person—either our organized field marketing events or industry events we use as go-to-market channels. Another day I'm working on deliverables like decks, reviewing assets like white papers, outreach sequences, or blogs.

A big part of my job is research. Especially with AI, the technology exists, but mapping new technology to existing workflows to add value requires industry know-how and customer expertise. There's a lot of deep research—very interactive. I ask questions, dive deeper into specific topics, test hypotheses, validate with potential customers, then roll out crisp messaging.

What do you think is the biggest challenge in the AI B2B space?

There are a few major challenges. First, there's so much noise in the market. Customers tell us, "What do you do? Someone else is saying something very similar" because vendors pitch similar things. As a vendor, how do you stand out? As a customer wanting to roll out AI, how do you assess what's real versus smoke, or differences between solutions?

You never know how accurate an AI model is for your organization until you test it. What works for another similar company may not work for you because everyone's technology toolkit is different. Your data setup is different. Your culture and people's habits are different.

There's also the adoption challenge. Experienced employees can often do things faster than AI because they've done it hundreds of times and optimized their process. When you put an AI tool in front of them and it doesn't give the results they want immediately, they think, "I don't have time to train a robot." This creates internal hesitation around adopting AI at scale.

Enterprise AI tools have zero tolerance for inaccuracy. In consumer space, if ChatGPT hallucinates, you give a thumbs down—no big deal. But in workplace analytics that impact company strategy, there's much bigger risk. It's less about PhD-level reasoning and more about helping with existing workflow and business results.

What advice do you have for people wanting to transition into product marketing from other domains?

The good thing about product marketing is almost everything you do is public anyway. If you're interested in product marketing, look at companies with good storytelling, messaging, and positioning. Watch their keynotes, look at their websites, blogs, ebooks—see what you like and understand why it resonates, why the messaging is clear, why it differentiates.

Learn from the best because great product marketers' work is public for everyone to see. There are also organizations like Sharebird and Product Marketing Alliance that host conferences where you can network and meet fellow product marketers.

Do you think the product marketing role will be "dead" or significantly impacted by AI evolution?

Actually, I think product marketers' roles have been elevated and we can be more impactful with AI.

First, the deep research capabilities that AI provides are incredible. Historically, I could only be deep in one or two industries or personas, but now deep research saves so much time. I can get much deeper in a variety of industries, personas, use cases, or markets much faster.

Second, content generation capabilities make broadcast work much faster.

Third, at Glean specifically, we help companies find enterprise context. As a product marketer, I'm the internal SME for marketing materials and use cases. People can find my work in our system and reach out when they need deal support or have questions. AI elevated my profile and made more people able to use my work at scale.

AI is making product marketers more powerful and capable.

What's your favorite clothing brand?

I actually just like getting stuff from Shein! They have so many varieties—any color, different thickness, materials. I just like the wider selection.