Brandon McGraw Is Teaching AI to Make Sense to People

The first version was a little vain and rather strange. Brandon McGraw fed old writing into GPT-3, cloned his own voice, and built a machine that would talk back as a rough simulation of Brandon McGraw. It was 2022. The resulting conversation was uncanny, occasionally profound, and often wrong. The errors were the useful bit. Each one made him stop and decide what he actually thought.

That experiment left him with a question he would carry through six prototypes, a startup, a turn at Anthropic, and eventually into OpenAI: “What would it mean if a computer could understand you?” The question sounds philosophical. McGraw treated it like a product brief.

In June 2026, he announced that he had become interim CMO of OpenAI’s Consumer and Hardware business. The appointment placed him around technology with planetary reach, but the central problem was still intimate. A person meets a new machine. The machine can do unfamiliar things. Someone must make the encounter legible without sanding away what is unfamiliar about it.

The career is a sequence of difficult introductions

McGraw has repeatedly arrived when a product, category, or audience needed introducing. Early in his career, he worked in advertising for Apple and Lenovo, then moved inside Google Creative Lab. At Facebook, he led product marketing for Groups and helped create the company’s first Communities Summit. He also served as product lead on a brand-engineering project that delivered personalized videos to 690 million people.

At Twitch, the audience was not merely consuming a product. Creators were making the place worth visiting. McGraw says he rebuilt product marketing and established a clearer foundation for communicating with them. At DoorDash, beginning in 2020, he built consumer and business product marketing through the dislocations of COVID-19, the company’s IPO, and the period beyond. The assignment sprawled across customers, merchants, and couriers, each with a different idea of what the same service was for.

The résumé can be read as a tour of familiar technology names. The more revealing reading is a tour of translation problems. How do you show that Groups are more than a tab? How does a platform speak with the people who create its culture? How do you explain one marketplace to three constituencies at once? McGraw’s own distinction is crisp: brand is the promise; product marketing is the proof. The phrase is neat enough for a slide, but it also imposes a useful discipline. A promise that outruns the experience is simply a debt.

“Know the difference between an anecdote and an insight.”Brandon McGraw on product marketing

His advice to product marketers tends to return to trust. Get close enough to product managers to be invited upstream. Listen beyond the loudest customers. Judge promotions through both ownership and autonomy. He has spoken candidly about his own nontraditional path from brand into product marketing, including the imposter syndrome that once made him reluctant to ask basic questions. His conclusion is generous: atypical backgrounds are useful precisely because product marketing sits at the center of a wheel and rarely commands it by decree.

Then the marketer opened the code editor

McGraw had been writing software since someone handed him a Microsoft BASIC tutorial book. His first triumph was a number-guessing game. The scale was modest; the sensation was not. Words had made the computer do something enjoyable. Decades later, OpenAI’s Whisper gave him a similar jolt. Speech could suddenly be transcribed with striking accuracy, and GPT-3 could respond to the text. The keyboard was no longer the narrow gate between a thought and a machine.

The early self-conversation became a voice journal. McGraw adapted a set of prompts from the “36 Questions That Lead to Love,” which he had enjoyed discussing with his wife, Whitney. A prompt arrived each day. Speaking rather than typing reduced the editorial polish people apply to their own thoughts. With enough entries, McGraw reasoned, the system might help someone notice a blind spot, connect an old idea to a new one, or decide what to do next.

His friends did not fall in love with it. They found the prompts deep and the experience “heavy.” Most did not wish to deliver a daily soliloquy to a machine. This was the sort of result a marketer might be tempted to reframe and a builder has to accept. McGraw did the latter.

He and his co-founders kept the question and changed the product. They formed Reflective Works in 2023 and developed Monograph, an AI-powered journal designed to remember what its user had shared. The work took six prototypes in all. By late 2024, McGraw estimated that he had spent about 1,500 hours using it. This is either commendable product diligence or a spectacular refusal to leave one’s own diary alone. Probably both.

The product’s uses were often pleasingly unglamorous: making a packing list, working through a book, planning a trip, deciding whether to call someone, or preserving an account of a child’s early years. McGraw wanted an AI that accumulated context without accumulating authority over the person using it. His critique of algorithmic feeds was that computers had learned to direct attention too well. Personal AI, in his formulation, offered a chance to return some agency.

“We can build AI that doesn't sacrifice our humanity or agency, it deepens it.”From McGraw’s account of building Monograph

Back to scale, carrying the prototype scars

In 2025, McGraw moved from his startup into consumer marketing at Anthropic. He publicly championed Claude’s Max subscription as a flexible answer for people who kept hitting usage limits. Later that year he joined OpenAI, initially leading product marketing for apps as the company prepared its first major ChatGPT brand campaign. By 2026, his public role had expanded to VP of Marketing for ChatGPT and then to the interim CMO remit across Consumer and Hardware.

That path contains a curious loop. McGraw left executive product marketing to build on the OpenAI API. A few years later, he returned to executive marketing at OpenAI with a founder’s education in everything a polished launch omits: awkward onboarding, vague feedback, technical constraints, the loneliness of an empty product, and the insolence of users who decline to behave as planned.

His recent public enthusiasm has centered on uses that feel immediate. He highlighted a class of young ChatGPT builders working on scientific discovery, accessibility, disaster response, and education. When ChatGPT Voice reached the desktop, he described leaving it open in the background while he spoke an idea or assigned a task. For a GPT-Live campaign, OpenAI put the product in conversation with grandmothers to demonstrate pauses, interruptions, translation, and the rhythm of ordinary speech. The technology was the news. The proof was three people talking.

OpenAI’s consumer and hardware brief is broader than any journal prototype. Hardware adds all the difficulties of atoms: cost, supply, ergonomics, and the small scandal of an object taking up space in a home. Consumer AI adds a category problem. People have heard plenty about what the technology may become. Marketing must explain what they can sensibly do with it on Tuesday afternoon.

A builder, even when the medium is a sentence

McGraw lives in Brooklyn and presents a life with interests that resist the algorithmic torrent: morning runs, dinners at home, books, history, architecture. He made Landmarks, a walking guide to New York City’s notable buildings. Monograph and Landmarks look unrelated until one notices that both are instruments of attention. One asks you to look at your inner life. The other asks you to look up from the pavement.

There is an appealing tension in his career. Marketing compresses. Reflection expands. A launch needs the short sentence; a life does not fit inside one. McGraw has worked on both sides of that divide, trying to make technology clear enough to approach without pretending that clarity makes it simple.

When he announced the interim CMO role, he ended with “Builders wanted. Onward!” It is recruitment language, certainly. It also sounds like the boy with the BASIC book, delighted that a sentence can cause a machine to act. The tools have changed. The pleasure has not. The consequential part of McGraw’s work now is deciding which sentence should be said, what action ought to follow, and whether the person on the other side feels more capable when it does.