Changes to How You'll Spend a Marketing Budget

Introduction

Katrina Wong, Chief Marketing Officer at New Relic, has spent more than 20 years building go-to-market engines at Twilio Segment, Salesforce, Zuora, SAP, and Hired — which makes her read on where marketing dollars are heading worth paying attention to. New Relic is an enterprise observability and APM platform, so her buyers are developers and technical teams, exactly the audience that has adopted LLMs fastest. Her position on what happens to marketing budgets when platforms like OpenAI turn on advertising is direct: the shift is coming, and the time to buy your seat is before the ROI shows up.

Fund the Hypothesis Before You Can Prove the ROI

Wong's operating instruction to her team is to deliberately suspend the ROI requirement on a small slice of spend. "Forget the ROI just quite yet. Let's take a small piece of our budget, experiment," is how she frames the mandate internally. That's not a permission slip for undisciplined spending — it's a carve-out, sized small enough that it doesn't threaten the performance budget that funds the pipeline number.

What That Looks Like in Practice

New Relic is currently participating in OpenAI's beta ChatGPT app program. Wong is candid that it isn't paying off yet: it's "not quite yielding the results in the same way yet as some of the other cha els." Her justification for staying in it anyway is the more useful takeaway for anyone building a 2026 plan — "you're always going to have to invest in your hypothesis of where the market's going to go." Early access to a cha el buys you learning curve and relationships that money can't retroactively purchase once the auction is competitive.

Is This the Death of SEM?

Wong's view is that LLMs may become "perhaps the default for how we find answers," with adoption already broad across business users, everyday consumers, and technical teams. Her phrasing on the consequence is blunt: "maybe it is the death of the SEM as we know it." For marketing leaders, the operational question isn't whether search spend disappears tomorrow — it's whether the line item that reliably converts today is structurally exposed, and whether you have any alternative in market when the erosion starts.

Segment by Surface, Not by Cha el

Paid advertising isn't going anywhere — the more interesting variable is which surface your buyer is standing on. A marketer building agents in Claude or Codex never touches a website and never sees an impression. That same marketer's parents are still on Google, still fully addressable through traditional cha els. Both behaviors coexist right now, inside the same total addressable market, and the value exchange around who sees what, when, and where is what's shifting dramatically.

The Streaming Precedent

Wong draws the parallel to television: audiences migrated to streaming, ads largely disappeared from the experience, and then reappeared around tentpole events like the Super Bowl and the World Cup — where they're actually welcome. The pattern worth noting is that ad inventory didn't die when the surface changed. It relocated, re-priced, and re-earned attention in a different context. Budget pla ing should assume relocation rather than elimination.

Conclusion

Three things to take into your next pla ing cycle. First, carve out a small, explicitly ROI-exempt test budget for LLM surfaces — small enough that underperformance is survivable, real enough to generate learning. Second, get into beta programs like OpenAI's ChatGPT app program now, while access is cheap and the competition is thin. Third, stop pla ing around cha els and start pla ing around surfaces, because your agent-building technical buyer and your search-native buyer require different bets from the same budget. The cha els you can currently prove are the ones most likely to erode — fund the hypothesis anyway.

--- One note on sourcing: the transcript provided is a short closing segment (roughly three minutes), so this post is built from the four exchanges it contains — the ROI-exempt test budget, the OpenAI beta participation, the SEM comment, and the streaming analogy. I kept the framing and takeaways tight to what was actually said rather than extrapolating additional strategies. If you have the fuller episode transcript, I can expand the middle sections with more of Katrina's specifics.