Last week, some of the brightest minds in marketing joined us at Index ‘26 in San Francisco. It was a day of illuminating and invigorating conversations, with an all-star speaker line-up sharing what's actually working in their AI strategies right now.
One of those voices was Dane Vahey, who heads up B2B marketing at OpenAI. We were dying to know how the marketing team of one of the world’s most innovative AI platforms uses AI to supercharge their work.
Vahey didn’t disappoint, and now, you can steal his playbook.
1. Stop using AI like a chatbot app.
This is Vahey’s central thesis: While most marketing teams are still in prompt-and-response mode, the ones seeing the most gains from AI have woven the tech into every workflow and decision point.
“Most people have largely used AI as an app, essentially asking ChatGPT a question and getting an answer back... It’s a change of mentality to think about AI as a system: Using intelligence everywhere you go, making it accessible to all the work that you’re doing,” he said, noting that the real power of AI is that it enables every member of your team to adopt a “builder’s mindset.”
Vahey gave a live demo of what this looks like in action, spinning up a fictional chocolate company from a blank screen. He ran a deep research query that searched 300+ sources in under 20 minutes, used the output to build a growth model dashboard, then dropped the file into Codex and had a fully functional website running before anyone finished their coffee. The whole exercise took less than half an hour.
Takeaway for marketers: Your team needs to stop thinking of AI as a chatbot and start thinking of it as a companion engineering team that can build creative solutions to drive real impact.
2. Every individual contributor is about to become a manager.
This was one of the most provocative things Vahey said all day—essentially, every team member is now a manager. “We’re coming to the end of the individual contributor. Every single IC in your organization is going to be managing agents,” he said.
His demo made the point concrete. He showed three types of agents in action: research agents (Deep Research), building agents (Codex), and analysis agents (ChatGPT for Excel). None of them required a technical background to direct; any marketer can now conduct research or build agents without touching code. “The idea of ‘not being technical’ is no longer relevant,” he emphasized.
Takeaway for marketers: “We bought ChatGPT licenses” is not an AI strategy. Think about what managing agents actually means for team structure, roles, and how you hire—and bring that conversation to leadership ASAP.
3. Engineering and marketing should be in the same room.
Even though your marketers don’t need a CS degree to direct agents, Vahey still believes marketing and engineering now belong under the same roof.
OpenAI has put that philosophy into practice. Rather than treating engineers as a shared resource or an on-call dependency, OpenAI embeds them directly into the marketing team. “We’ve brought engineering in so we have more throughput to help us innovate and move faster,” he said.
Takeaway for marketers: If your marketing team doesn’t have dedicated technical resources, that’s the next conversation to have with your CEO.
4. Hire for curiosity and agency, not just experience.
Vahey has a specific term for the trait he prizes most in his hires: “training to failure,” a phrase borrowed from weightlifting.
“A lot of people are using this technology, but aren’t really pushing it to the limits. We look for people who are going to push the models and the technology to the highest-level capability,” he said. “We want people to honestly kind of break things.”
He added that this mentality is essential for a technology as fast-paced as AI, where new functionalities roll out weekly. The team members who probe for the ceiling today are most likely to adapt quickly as the limits shift.
Vahey also shared a litmus test he uses in interviews: When’s the last time you learned a new piece of software or built something with a new product? It’s a simple question that quickly separates the curious from the comfortable.
Takeaway for marketers: Reconsider what “AI experience” means when you’re hiring and sizing up your existing team. Years in the industry matter less than whether someone actually pushes the technology to its limits.
5. Make “what did you build this week?” a standing agenda item.
Every week, the first 15 minutes of Vahey’s staff meeting are reserved for show-and-tell. “We go around the room, and everyone has to share something new they built with AI in the last week. It could be something in their professional life or something in their personal life. But it forces us to actually be using the technology in new ways,” he said.
This creates peer-to-peer accountability—which, as anyone who’s tried to drive org-wide adoption knows, is a far stickier approach than top-down mandates.
Takeaway for marketers: This is the simplest tactic you can steal from this session. Add it to your next staff meeting. It forces real usage and surfaces what’s working across the team without a single new tool or budget line.
6. Stop measuring time saved. Instead, measure output and impact.
Vahey is skeptical of “hours saved” as an AI success metric. In fact, when he asked the room to raise their hands if AI had actually given them meaningful time back, not many hands went up. The better frame: AI expands what’s possible.
“Seventy-five percent of workers say that they can accomplish things that they could never do before,” he said. “When I think about the right metrics, it’s not savings of time. It’s how are you increasing your output and how are you generating more impact.”
Takeaway for marketers: If you’re making the case for AI investment internally, ditch the time-savings argument. It rarely lands. Show what your team can do now that was impossible before.
Vahey left us with a lot to think about (and a lingering craving for chocolate). He's been building an AI-native marketing team longer than almost anyone, and it showed. Hard to sit through that session and walk away without at least one thing you're doing differently on Monday.
This was just session one—stay tuned to AI Native for more exclusive insights from top AI marketing leaders in the weeks to come.
AI news is as dense as a Claude Mythos security briefing. Here’s what you need to know this week.
Anthropic's most powerful model won’t be released to the public. Dozens of organizations, including AWS, Apple, Google, Microsoft, and NVIDIA, received early access to Claude Mythos Preview, an unreleased model that has already found thousands of previously unknown security vulnerabilities. Anthropic is keeping it restricted because it says the model is too easy to weaponize. Credible safety call or the most alarming launch announcement in years? Maybe both.
Anthropic ends the OpenClaw free ride. Claude Pro and Max subscribers can no longer use their flat-rate plans to power OpenClaw, the popular open-source AI agent framework. Users now have to pay separately, with costs reportedly jumping up to 50x for the heaviest users. The open-source community is not pleased.
Perplexity’s search-to-agent pivot is paying off. The company’s ARR hit $450 million in March (a 50% jump in a single month) after launching Computer, its AI agent product, and moving to usage-based pricing.
Ads in AI Search may cost more than they’re worth. A new Ipsos survey found that 63% of US adults say ads in AI Search results would reduce their trust in results. The timing is awkward: OpenAI is actively expanding its ChatGPT ad pilot, and Google has been running ads in AI Mode since late 2025.
Who does AI actually cite? MuckRack analyzed 15 million AI response citations and shared the data with Press Gazette. Reuters is the most-cited publication globally, followed by Forbes, and a quarter of all links cited by AI are journalistic. Of particular note for PR teams chasing GEO wins, specialist and B2B titles dominate rankings over general news outlets.
GoDaddy sites can now control AI crawler access. Cloudflare and GoDaddy announced a partnership to give website owners visibility and control over which AI crawlers can access their content. The companies are also supporting a new open standard called Agent Name Service (ANS), essentially a verified identity layer for AI agents.
Anthropic is getting serious about compute. The company signed an expanded deal with Google and Broadcom for 3.5 gigawatts of next-gen TPU capacity, its “most significant compute commitment to date.”
Google launched a free dictation app. Google AI Edge Eloquent is an offline-first dictation tool that transcribes in real time, automatically strips filler phrases, and runs entirely on-device using Gemma models. A true gift for those of us who can’t seem to kick that high-school habit of saying “like” or “um” every other word.
Director, Marketing Strategy & Operations @ Audible (Newark, NJ; $137,900–$186,500 USD annual salary)
Go-To-Market Strategy Director @ Fetch (Madison, WI; $173,801–$204,472 annual salary)
Head of Marketing @ Valon (San Francisco, CA/US Remote; no salary listed)