OpenAI scales product operations with AI-powered Airtable
OpenAI runs its product operations in Airtable—turning scattered launches and Slack threads into a system the whole company can see.
50–60
Active projects AI-summarized weekly
8-week
Rolling view of upcoming milestones
1
Source of truth for all of product development
Shortly after ChatGPT launched, Blake Samic joined OpenAI. He'd led product operations at both Uber and Stripe, and what he heard in his interviews felt familiar: "We need to become this great product company now, in addition to this amazing research company."
Different companies, same core problem.
"The constituents were different," Blake says, "but it was still about how to build this connective tissue inside the organization."
Today, Airtable is where OpenAI's product organization tracks what's shipping, spots what's stuck, and stays aligned:
50–60 projects summarized by AI fields every week
Blockers flagged before anyone has to escalate
Months of Slack conversations turned into ranked customer priorities
Quarterly planning docs processed into audience-specific briefings
One launch calendar connecting engineering, marketing, and leadership
None of that existed when Blake arrived—here’s how they built it in Airtable.
When the product moves faster than the ops
OpenAI wasn't just shipping fast. It was trying to become a fundamentally different kind of company—and the gap between the people building products and the rest of the organization was growing by the week. More features in parallel. More teams. More launches. But no shared view of any of it.
New features blindsided teams who didn't know they were coming. People who'd been at the company longer knew who to ask and where to look—but newer employees couldn't find basic information, and didn't feel like their voices were heard. Someone had set up a form to capture feature requests from the sales team. Well meaning, but barely anyone used it. It didn't fit the way people actually worked.
Meanwhile, a Slack channel between sales and product was buzzing. "They were talking all day every day about this stuff," Blake recalls. Real customer intelligence, completely unstructured and impossible to act on at scale. Quarterly planning docs piled up unread. The faster the company grew, the more information it produced and the less of it anyone could actually find.
"You're not just shipping a lot of products. You're shipping a lot of information in your company."
Structure first, then intelligence
With launch information scattered and teams getting blindsided, Blake's first move was a launch calendar—one place to see what's going out the door. He'd pushed Google Sheets to its limit at Stripe and eventually had to build custom tooling with code. His instinct was to start scrappy with Sheets again when Peter Dang, coworker and former Airtable employee, told him to take a fresh look at the platform.
What he found impressed him. The building blocks could scale with the company—and because executives, PMs, and go-to-market teams all needed to see the same data differently, the ability to build tailored views on top of one shared foundation mattered. He could also pull data out for analysis and bring it back in. "I really liked that I could get it out, do some analysis somewhere else, get it back in," he recalls. "I still love that today."
"Using [Airtable], you can feel the craft in the product."
That launch calendar was his first quick win. And once it was in a good place, he started expanding from there.
Today, every significant initiative at OpenAI lives in Airtable: a project for each feature, milestones tracking key dates, and short weekly updates from each project lead. Blake pushed for this cadence early. "We were not a ton of people then, but we were gonna be a ton of people later. So I really wanted to get some of that hygiene going."
Once those updates started flowing in every week, Airtable's AI fields started working on them automatically—condensing long updates into short summaries, cross-referencing what people wrote against the milestones they'd logged, flagging mismatches. As Blake puts it: "An AI column that is just running, and it automatically does things based on whatever comes into some other column."
The eight-week view
Every week, those AI fields condense project updates into summaries short enough to scan. The result is a rolling eight-week view across 50 or 60 simultaneous projects, with a tighter two-week lens for what's about to ship. Blake's term for it: ambient awareness. Executives, PMs, and go-to-market teams all see the same data, shaped for their role. Updates also get repurposed into newsletters and internal podcasts so people can consume them however they want.
"So many more people in the organization just have some idea of what's happening, and they can dig in when they need it."
Finding who's stuck before they say so
Product ops doesn't wait for someone to raise a hand. A separate AI-powered view flags projects where a team appears stuck. "We can very quickly isolate where someone needs help," says Blake, "and go have an unblocking conversation."
No blame, just accuracy
When a project lead's update doesn't match the milestones they've logged, AI catches the mismatch. "No blame. It's just, how do we get it to the truth as fast as possible?"
There's appetite for more. "You can think about what are the twenty other columns that I would want to have."
The team doesn't stop at what Airtable can do natively. They also point AI at messier sources:
From Slack threads to ranked priorities
That Slack channel full of customer intelligence nobody could act on? Blake's team pulled months of those conversations, ran them through OpenAI models, and produced a ranked view of top feature requests with click-through to the original quotes.
"You can go from a very unstructured thing, just where people are conversing, to something that gives you the structure that you need."
Planning docs that actually get read
Every quarter, product teams write two- to three-page reflections and plans. Advanced models, including o3 Pro, now process all of them at once—producing one-slide summaries for leadership, tailored briefings for specific teams, and analysis that flags where product strategy and sales strategy aren't aligned.
"Tell us the four things we need to go talk about that we're not totally aligned on," Blake says, "versus everyone just restating the stuff that is really obvious."
What two years of commitment looks like
The system now shows Blake things he couldn't see before. Two launches landing on the same day—and the chance to shift one by a day so they land with more impact.
Blake says they generally try not to slow anything down for timing, "but sometimes it makes sense if it's really a few days to kind of bundle some things together." Upstream models being trained that will power new features. Marketing moments tied to product releases.
The full picture of how OpenAI builds and ships products, visible in one place.
"You have to really commit to it at least weekly," Blake says of keeping the system current. The payoff:
50–60 projects visible to the entire company at a glance, with an eight-week horizon and a two-week lens for what's about to ship.
Ad hoc information requests eliminated. When accounting needs to know what shipped last quarter, the answer's already there. No one gets interrupted.
Two-sentence hallway intros on every project, creating a lightweight knowledge layer across 50-plus simultaneous initiatives. "You'd be surprised at how many connections just get made," Blake says.
"I never could have asked any size team to go read all these updates and summarize them to a tweet. That's a tall ask for a professional. But now you can do that pretty much endlessly—and the use cases it opens up at scale are really interesting."
What Blake's building next
Custom interfaces are first. The team plans to vibe-code a denser eight-week roadmap and a news-feed experience for project updates on top of Airtable. After that, Blake wants agents handling specific ops functions: organizing dogfooding, managing betas, orchestrating launch readiness.
"The tinkerer, the builder," he says of what he looks for in hires now. "You're going to be building things. And that can lead to really interesting places."
But not replacing the people who hold it together.
"The folks that we have are amazing cross-functional bridges. Even if you had fifteen agents, it wouldn't work yet fully without those people there."
Based on an interview with Blake Samic, Head of Product Operations at OpenAI, conducted at Airtable LA in 2025.