Description: Every woodworker builds their own workbench first, because it's the thing that makes every other project possible. Chris Geoghegan, VP of Product Strategy and AI Transformation at Zapier, sees the power users of AI doing basically that, and he's in a unique position to know. Zapier connects over 8,500 apps and has spent more than a decade watching how people actually get work done.
In this episode, get a look behind the curtain at how one of the most widely used workflow platforms is thinking about AI at this point in time. From building personal and shared context for your tools, to where AI collaboration is most advanced, to a not-so-far-off future of not having to configure automation.
A workbench metaphor
In woodworking, a workbench is often one of a craftsman’s first projects because it’s the thing that makes every other project possible. He sees a similar pattern happening among some of the power users of AI. They’re building a personalized infrastructure of tools that works for them.
His own version of this lives in Cursor, a workspace he calls “Brain” that connects to his calendar, his documents, and his tools. When he opens it in the morning, it already knows what he’s working on for the day.
“It creates my daily brief. Who am I going to meet with today? Does a bit of research on them, prep work, connect it to all the tools I use.”
That’s the noteworthy difference between using AI reactively (opening a tab, pasting some context, and getting an output) vs. using it as infrastructure. The workbench doesn’t wait to be asked, it’s already running.
The lack of shared context remains mostly unresolved
Building a personal workbench is a fantastic start. But getting a whole team working the same way is a tougher nut to crack, and one Chris thinks the industry hasn’t (quite yet).
“We’re transforming how individuals work. We’re not yet at the point where we’re transforming how teams work together.”
Part of what makes this difficult is structural. AI collaboration is likely most advanced in coding, and that makes sense when you think about it. Software development already has a human-to-human collaboration model that translates naturally to human-to-AI. The context lives in the code, it’s version-controlled, and it’s shared. Everyone’s working from the same source of truth.
Most other knowledge work isn’t built that way. Context lives in Slack threads, in email chains, in someone’s head. Every time you sit down with AI to do something meaningful, you’re rebuilding that context from scratch.
Zapier’s current approach is to change the underlying structure of how context is stored and shared, so that whatever tool anyone on the team reaches for, the shared knowledge is already there waiting. They have a shared Google Drive folder of Markdown files synced to everyone’s machine. Simple, portable, and tool-agnostic. Whether someone is working in Cursor, Claude Code, or something else entirely, they’re all pulling from the same well.
Having an open posture for disruption
Chris has been at Zapier for nearly a decade, and one of the things he’s most proud of is how the company responds to disruption:
“Every moment where there’s been something disruptive, whether that’s coming from customers, coming from the market, we don’t think, how do we make the existing thing survive? We think somebody’s going to disrupt us. Let it be us.”
So when ChatGPT dropped, the response was consistent with that. Their CEO called a code red to stop normal operations, set aside the roadmap, and rethink the fundamentals. It opened up a period of exploration, with parallel experiments running, some competing with each other. Chris says that a lot of what Zapier has shipped recently traces back to that time.
Now the work looks a little different! Chris describes the current phase as consolidation - taking everything that was learned and pulling it into something more coherent. The risk of moving fast and experimenting broadly is that users end up having to make one-way-door choices between products solving similar problems in slightly different ways. Closing that gap is where the focus is now with their team.
The continued blurring of roles
The idea of T-shaped skills have been around for some time: deep expertise in one area with broad capability across others. What’s changing is the width of that horizontal bar.
Becoming deeply expert at something still takes time. But becoming capable enough to contribute meaningfully across product, design, and engineering is more accessible. Some of Zapier’s designers are reaching for Claude Code instead of Figma when they need to prototype. Not because they’ve become engineers, but because the distance between having an idea and having a working version of it has collapsed. The strict boundaries that used to define who did what are getting a little harder to justify.
Even still, Chris and George are both quick to acknowledge the distinction: capable isn’t the same as expert. The polish a seasoned designer brings in Figma or the architectural decisions an experienced engineer makes under the hood is critical judgement that takes years to develop. What’s shifted is the on-ramp.
“It is way easier to become more capable. But becoming an expert still takes 10,000 hours.”
Chris connects all of this to a bigger question for the industry. Referencing the “N of one software” - the idea that when building is cheap enough, anyone can create something hyper-personalized for themselves or their specific team rather than buying off the shelf. That has potentially scary implications for the SaaS market broadly. And for product managers specifically, it changes the nature of the work. Overall, Chris thinks that the blurring of the lines between user, builder, and product manager are mostly a good thing.
What’s the next great frontier at Zapier?
Zapier’s brand reputation was built on making automation accessible to people who couldn’t code. The next version of that is more ambitious: what if you never had to think about automation at all?
Right now, even with an AI-assisted setup, there’s still cognitive load on the user. The editor is visible, the field mapping is your problem, and when something breaks, you’re the one who has to fix it. At Zapier they are exploring what it looks like when the system understands what you’re trying to accomplish and handles the rest. Sans the configuration and troubleshooting.
And it’s not as far off as it sounds! The gap between hitting an error and resolving it is already closing. When it closes entirely, there can be a whole new class of users who will use automation without ever knowing that’s what they’re doing. For a platform that has spent a decade watching how people actually get work done, it’s one of their most exciting chapters to date. At Crema, our staff can’t wait to see it unfold.
People of Product is brought to you by Crema - a design & technology consultancy
