Brad Smith — Head of Startups and Investor Partnerships at Zapier — will present How Zapier AIs at Mond(ai)y Coffee on Monday, August 3.
What are you building?
How Zapier enables AI adoption across the whole company — from individual productivity to shared, governed workflows. Brad will share how Zapier is building its product for the new AI era, where people and agents use the same connected automation layer, and will show the new Zapier AI assistant experience. He’ll also cover how Zapier adopts AI internally at scale, and ideas for founders looking to increase AI adoption.
The core idea
Zapier is the connective and action layer of the AI stack: it connects models to the apps, data, and business processes where work actually happens, turning intelligence into reliable action.
What’s working / not working
Working: Integrations and automations remain the dependable foundation — AI becomes useful when it can securely connect to business context and take action across the tools a company already uses. Teams are finding value by embedding AI into real workflows rather than treating it as a standalone destination.
Not working yet: Governance and model sprawl remain major market challenges as companies adopt more models, agents, and AI-enabled tools. The value and ROI of AI investments are cloudier than ever — pun intended. It’s still difficult to separate durable productivity gains from experimentation, novelty, and infrastructure spend.
Open questions
- Where should a startup draw the line between an agent making a judgment and an automation executing a known process?
- What trust and review models let teams safely give AI more autonomy over time?
- What keeps promising AI experiments from becoming durable habits and systems?
- Which founder workflows are producing the fastest, clearest ROI today?
- As AI capabilities improve, how are founders deciding between hiring another person and buying more tokens for AI to do the job?
Spicy takes
- Six to twelve months from now, what does the new tech stack look like when agents become first-class users of software?
- What happens to SaaS and labor pricing when software spend, headcount, and token spend increasingly compete for the same budget?
- Over the next few quarters, do IPO markets, AI valuations, and concerns about an AI bubble reset — or accelerate?
- Which parts of today’s AI stack become durable categories, and which collapse into features or infrastructure?
If we only discuss one thing
“You can delegate the work, but you can’t delegate the accountability.” — Bryan Helmig, Co-Founder & CTO, Zapier
Where should your organization draw the line for AI accountability? When AI recommends, decides, or acts, who remains responsible — and what oversight should change as the system earns more trust?
Meeting Notes
Every MLK weekend we go with a group of families to Great Wolf Lodge. An unexpected perk of that tradition is watching all the kids grow year over year.
The first year, this slide is too scary. The next year, the same slide is too boring.
I say all that to say — it was interesting to hear Brad Smith share what Zapier is doing to evolve the company in a world of AI. Hearing a new company share their ideas is one thing. Hearing it from a company you’ve known and used for 15 years is another. You need the same slide two years running to see how much the kid changed.
For me, the most useful part of the robots is building the giant markdown brain of connections. But when I show that to most people, they aren’t interested in a life of CLIs and .md files. What Zapier is building — and has built — seems like a bridge to that kind of connection.
Very cool.
PS: Thanks to Ankush Tewari for the coffee.
Coming up:
- 8/10 — John Avery is going to walk us through how ATDC AIs — building an approved AI path inside an institution with real compliance constraints
Come on and join the fun and the conversation. Everyone welcome, Mondays 8:30a.