Mar 9, 2026 · 32m · mixergy

#2299 Zapier is using AI to sell to AI

Wade Foster · 22m spoken Andrew Warner · 6m spoken
0:00 / 0:00

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Zapier co-founder and CEO Wade Foster joins Andrew Warner to break down how executive leadership is evolving through multi-agent decision frameworks, Model Context Protocol tooling, and AI-first workflows. Foster also shares critical insights on marketing to autonomous software agents, the realistic economics of vibe coding versus SaaS, and scaling AI literacy across an entire organization.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Andrew holds 21.9% of the talking time here. How this is scored →

Andrew as informed peer 4.5 Guest teaching 5.4 Guest disagreement 1.0 Andrew pushing back 1.4
05100:0010:0020:0030:000:47–4:42 · Andrew as informed peer 4/10 Optimizing Software and Documentation for AI Agent Marketing Andrew shares his own user experience being routed to Vercel by Claude Code before questioning how agent SEO testing works given indexing lag. Wade explains the emerging mechanics of serving plain text, fast docs tailored directly to AI agents.4:42–7:16 · Andrew as informed peer 5/10 The Reality of Building Versus Buying Enterprise SaaS Andrew questions why a company wouldn't vibe code its own tools to save money, then validates Wade's thesis using his own calorie tracking app edge-case maintenance issues. Wade breaks down the economics of low software spend versus the high engineering distraction of maintaining internal tools.7:17–11:23 · Andrew as informed peer 3/10 Developing the Multi-Agent War Council Decision Framework Andrew expresses his excitement for generative AI tools and asks how Wade structured his workflow. Wade introduces his War Council multi-agent framework and explains how he used Granola meeting notes and Cursor to automatically replicate it.11:24–15:35 · Andrew as informed peer 4/10 Evaluating the Commercial Viability of AI Agent Skills Andrew asks whether selling AI skills as subscriptions could become a viable SaaS-like business model. Wade reframes skills as low-barrier commodity documents similar to early WordPress themes and details how the War Council helps him bypass human evaluation bias.15:37–21:53 · Andrew as informed peer 5/10 Demonstrating Cursor as an Executive AI Chief of Staff Andrew prompts a live screen share to examine Wade's Cursor workspace and file setup. Wade demonstrates toggling between Claude Opus and OpenAI Codex to cross-critique bugs and explains letting the AI self-organize its executive knowledge base.21:53–23:58 · Andrew as informed peer 6/10 Connecting Tools and Actions Using Zapier MCP Protocol Andrew asks detailed questions about Zapier's Model Context Protocol (MCP) integrations and shares his hands-on observations comparing Claude and ChatGPT connector usability. Wade explains how authenticating the MCP server equips the LLM with tool-calling capabilities.23:59–29:46 · Andrew as informed peer 4/10 Scaling AI Adoption and Executing Complex Spreadsheet Analysis Andrew asks how Zapier drives company-wide AI adoption across non-technical staff. Wade walks through a practical case study where he analyzed four dense spreadsheets in 15 minutes, emphasizing the need for human domain intuition to spot model false alarms like maternity leaves.29:46–32:05 · Andrew as informed peer 5/10 Emerging Tools and Brian Halligan's Focus Hat Advice Andrew shares a new AI video editing tool he has been testing and inquires about a decorative hat in Wade's background. Wade shares advice from HubSpot founder Brian Halligan about using a 'NO' hat to enforce focus against distracting new ideas.0:47–4:42 · Guest teaching 5/10 Optimizing Software and Documentation for AI Agent Marketing Andrew shares his own user experience being routed to Vercel by Claude Code before questioning how agent SEO testing works given indexing lag. Wade explains the emerging mechanics of serving plain text, fast docs tailored directly to AI agents.4:42–7:16 · Guest teaching 6/10 The Reality of Building Versus Buying Enterprise SaaS Andrew questions why a company wouldn't vibe code its own tools to save money, then validates Wade's thesis using his own calorie tracking app edge-case maintenance issues. Wade breaks down the economics of low software spend versus the high engineering distraction of maintaining internal tools.7:17–11:23 · Guest teaching 6/10 Developing the Multi-Agent War Council Decision Framework Andrew expresses his excitement for generative AI tools and asks how Wade structured his workflow. Wade introduces his War Council multi-agent framework and explains how he used Granola meeting notes and Cursor to automatically replicate it.11:24–15:35 · Guest teaching 6/10 Evaluating the Commercial Viability of AI Agent Skills Andrew asks whether selling AI skills as subscriptions could become a viable SaaS-like business model. Wade reframes skills as low-barrier commodity documents similar to early WordPress themes and details how the War Council helps him bypass human evaluation bias.15:37–21:53 · Guest teaching 6/10 Demonstrating Cursor as an Executive AI Chief of Staff Andrew prompts a live screen share to examine Wade's Cursor workspace and file setup. Wade demonstrates toggling between Claude Opus and OpenAI Codex to cross-critique bugs and explains letting the AI self-organize its executive knowledge base.21:53–23:58 · Guest teaching 4/10 Connecting Tools and Actions Using Zapier MCP Protocol Andrew asks detailed questions about Zapier's Model Context Protocol (MCP) integrations and shares his hands-on observations comparing Claude and ChatGPT connector usability. Wade explains how authenticating the MCP server equips the LLM with tool-calling capabilities.23:59–29:46 · Guest teaching 6/10 Scaling AI Adoption and Executing Complex Spreadsheet Analysis Andrew asks how Zapier drives company-wide AI adoption across non-technical staff. Wade walks through a practical case study where he analyzed four dense spreadsheets in 15 minutes, emphasizing the need for human domain intuition to spot model false alarms like maternity leaves.29:46–32:05 · Guest teaching 4/10 Emerging Tools and Brian Halligan's Focus Hat Advice Andrew shares a new AI video editing tool he has been testing and inquires about a decorative hat in Wade's background. Wade shares advice from HubSpot founder Brian Halligan about using a 'NO' hat to enforce focus against distracting new ideas.0:47–4:42 · Guest disagreement 1/10 Optimizing Software and Documentation for AI Agent Marketing Andrew shares his own user experience being routed to Vercel by Claude Code before questioning how agent SEO testing works given indexing lag. Wade explains the emerging mechanics of serving plain text, fast docs tailored directly to AI agents.4:42–7:16 · Guest disagreement 1/10 The Reality of Building Versus Buying Enterprise SaaS Andrew questions why a company wouldn't vibe code its own tools to save money, then validates Wade's thesis using his own calorie tracking app edge-case maintenance issues. Wade breaks down the economics of low software spend versus the high engineering distraction of maintaining internal tools.7:17–11:23 · Guest disagreement 1/10 Developing the Multi-Agent War Council Decision Framework Andrew expresses his excitement for generative AI tools and asks how Wade structured his workflow. Wade introduces his War Council multi-agent framework and explains how he used Granola meeting notes and Cursor to automatically replicate it.11:24–15:35 · Guest disagreement 2/10 Evaluating the Commercial Viability of AI Agent Skills Andrew asks whether selling AI skills as subscriptions could become a viable SaaS-like business model. Wade reframes skills as low-barrier commodity documents similar to early WordPress themes and details how the War Council helps him bypass human evaluation bias.15:37–21:53 · Guest disagreement 1/10 Demonstrating Cursor as an Executive AI Chief of Staff Andrew prompts a live screen share to examine Wade's Cursor workspace and file setup. Wade demonstrates toggling between Claude Opus and OpenAI Codex to cross-critique bugs and explains letting the AI self-organize its executive knowledge base.21:53–23:58 · Guest disagreement 1/10 Connecting Tools and Actions Using Zapier MCP Protocol Andrew asks detailed questions about Zapier's Model Context Protocol (MCP) integrations and shares his hands-on observations comparing Claude and ChatGPT connector usability. Wade explains how authenticating the MCP server equips the LLM with tool-calling capabilities.23:59–29:46 · Guest disagreement 1/10 Scaling AI Adoption and Executing Complex Spreadsheet Analysis Andrew asks how Zapier drives company-wide AI adoption across non-technical staff. Wade walks through a practical case study where he analyzed four dense spreadsheets in 15 minutes, emphasizing the need for human domain intuition to spot model false alarms like maternity leaves.29:46–32:05 · Guest disagreement 0/10 Emerging Tools and Brian Halligan's Focus Hat Advice Andrew shares a new AI video editing tool he has been testing and inquires about a decorative hat in Wade's background. Wade shares advice from HubSpot founder Brian Halligan about using a 'NO' hat to enforce focus against distracting new ideas.0:47–4:42 · Andrew pushing back 3/10 Optimizing Software and Documentation for AI Agent Marketing Andrew shares his own user experience being routed to Vercel by Claude Code before questioning how agent SEO testing works given indexing lag. Wade explains the emerging mechanics of serving plain text, fast docs tailored directly to AI agents.4:42–7:16 · Andrew pushing back 2/10 The Reality of Building Versus Buying Enterprise SaaS Andrew questions why a company wouldn't vibe code its own tools to save money, then validates Wade's thesis using his own calorie tracking app edge-case maintenance issues. Wade breaks down the economics of low software spend versus the high engineering distraction of maintaining internal tools.7:17–11:23 · Andrew pushing back 1/10 Developing the Multi-Agent War Council Decision Framework Andrew expresses his excitement for generative AI tools and asks how Wade structured his workflow. Wade introduces his War Council multi-agent framework and explains how he used Granola meeting notes and Cursor to automatically replicate it.11:24–15:35 · Andrew pushing back 2/10 Evaluating the Commercial Viability of AI Agent Skills Andrew asks whether selling AI skills as subscriptions could become a viable SaaS-like business model. Wade reframes skills as low-barrier commodity documents similar to early WordPress themes and details how the War Council helps him bypass human evaluation bias.15:37–21:53 · Andrew pushing back 1/10 Demonstrating Cursor as an Executive AI Chief of Staff Andrew prompts a live screen share to examine Wade's Cursor workspace and file setup. Wade demonstrates toggling between Claude Opus and OpenAI Codex to cross-critique bugs and explains letting the AI self-organize its executive knowledge base.21:53–23:58 · Andrew pushing back 1/10 Connecting Tools and Actions Using Zapier MCP Protocol Andrew asks detailed questions about Zapier's Model Context Protocol (MCP) integrations and shares his hands-on observations comparing Claude and ChatGPT connector usability. Wade explains how authenticating the MCP server equips the LLM with tool-calling capabilities.23:59–29:46 · Andrew pushing back 1/10 Scaling AI Adoption and Executing Complex Spreadsheet Analysis Andrew asks how Zapier drives company-wide AI adoption across non-technical staff. Wade walks through a practical case study where he analyzed four dense spreadsheets in 15 minutes, emphasizing the need for human domain intuition to spot model false alarms like maternity leaves.29:46–32:05 · Andrew pushing back 0/10 Emerging Tools and Brian Halligan's Focus Hat Advice Andrew shares a new AI video editing tool he has been testing and inquires about a decorative hat in Wade's background. Wade shares advice from HubSpot founder Brian Halligan about using a 'NO' hat to enforce focus against distracting new ideas.

speaking balance: gold is Andrew, purple is the guest (3 minute bins)

0:00 · Andrew 24.8% · guest 75.2%0:00 · Andrew 24.8% · guest 75.2%3:00 · Andrew 38.1% · guest 61.9%3:00 · Andrew 38.1% · guest 61.9%6:00 · Andrew 28.1% · guest 71.9%6:00 · Andrew 28.1% · guest 71.9%9:00 · Andrew 16.6% · guest 83.4%9:00 · Andrew 16.6% · guest 83.4%12:00 · Andrew 4.5% · guest 95.5%12:00 · Andrew 4.5% · guest 95.5%15:00 · Andrew 24.4% · guest 75.6%15:00 · Andrew 24.4% · guest 75.6%18:00 · Andrew 14.5% · guest 85.5%18:00 · Andrew 14.5% · guest 85.5%21:00 · Andrew 35.4% · guest 64.6%21:00 · Andrew 35.4% · guest 64.6%24:00 · Andrew 13.2% · guest 86.8%24:00 · Andrew 13.2% · guest 86.8%27:00 · Andrew 12.5% · guest 87.5%27:00 · Andrew 12.5% · guest 87.5%30:00 · Andrew 33.5% · guest 66.5%30:00 · Andrew 33.5% · guest 66.5%
Sharpest disagreement ▶ 11:46 Reframing skills as commoditized theme files

Wade dismisses the premise that AI skills can become high-value SaaS subscription businesses, comparing them directly to commoditized $10 WordPress skins.

Hardest push from Andrew ▶ 3:43 Challenging the feasibility of rapid agent SEO testing

Andrew directly pushes back on Wade's description of agent marketing experimentation, questioning how anyone can quickly measure feedback when agents take time to discover changes.

Biggest teaching moment ▶ 5:16 Breaking down the true economic cost of internal software

Wade educates Andrew on enterprise software budgeting, explaining that software subscriptions represent negligible single-digit spend compared to the massive distraction of engineering maintenance.

Andrew holds their own ▶ 23:32 Comparing MCP connector reliability across LLMs

Andrew demonstrates his own technical fluency by explaining his direct testing of Zapier MCP with Claude versus ChatGPT, detailing exactly how each client handles tool access.

the scores for every segment, with the reasoning behind each
ChapterTopicAndrew as informed peerGuest teachingGuest disagreementAndrew pushing backWhy
Optimizing Software and Documentation for AI Agent Marketing 4513 Andrew shares his own user experience being routed to Vercel by Claude Code before questioning how agent SEO testing works given indexing lag. Wade explains the emerging mechanics of serving plain text, fast docs tailored directly to AI agents.
The Reality of Building Versus Buying Enterprise SaaS 5612 Andrew questions why a company wouldn't vibe code its own tools to save money, then validates Wade's thesis using his own calorie tracking app edge-case maintenance issues. Wade breaks down the economics of low software spend versus the high engineering distraction of maintaining internal tools.
Developing the Multi-Agent War Council Decision Framework 3611 Andrew expresses his excitement for generative AI tools and asks how Wade structured his workflow. Wade introduces his War Council multi-agent framework and explains how he used Granola meeting notes and Cursor to automatically replicate it.
Evaluating the Commercial Viability of AI Agent Skills 4622 Andrew asks whether selling AI skills as subscriptions could become a viable SaaS-like business model. Wade reframes skills as low-barrier commodity documents similar to early WordPress themes and details how the War Council helps him bypass human evaluation bias.
Demonstrating Cursor as an Executive AI Chief of Staff 5611 Andrew prompts a live screen share to examine Wade's Cursor workspace and file setup. Wade demonstrates toggling between Claude Opus and OpenAI Codex to cross-critique bugs and explains letting the AI self-organize its executive knowledge base.
Connecting Tools and Actions Using Zapier MCP Protocol 6411 Andrew asks detailed questions about Zapier's Model Context Protocol (MCP) integrations and shares his hands-on observations comparing Claude and ChatGPT connector usability. Wade explains how authenticating the MCP server equips the LLM with tool-calling capabilities.
Scaling AI Adoption and Executing Complex Spreadsheet Analysis 4611 Andrew asks how Zapier drives company-wide AI adoption across non-technical staff. Wade walks through a practical case study where he analyzed four dense spreadsheets in 15 minutes, emphasizing the need for human domain intuition to spot model false alarms like maternity leaves.
Emerging Tools and Brian Halligan's Focus Hat Advice 5400 Andrew shares a new AI video editing tool he has been testing and inquires about a decorative hat in Wade's background. Wade shares advice from HubSpot founder Brian Halligan about using a 'NO' hat to enforce focus against distracting new ideas.

Statements from this episode (10)

Insight
Wade Foster: Marketing to AI purchasing agents requires entirely new skills
“The agent is actually choosing what products to buy on behalf of a human. And so you're no longer advertising to the human. You are trying to get the agent to say, pick me. And that is often a very different skill set than marketing to a human.”
Wade Foster Mar 9, 2026 ▶ 1:00
Insight
Wade Foster: Companies are building plain-text webpages exclusively for AI agents
“Agents care a ton about, like, clean documentation, like, fast webpages, so, you know, it's like, the thing I'm seeing a lot of folks do is serve up a separate version of their webpage that is meant just for the agent, and so that content is, like, plain text.”
Wade Foster Mar 9, 2026 ▶ 2:18
Insight
Wade Foster: Vibe coding is for one-off utilities, not load-bearing SaaS
“I think vibe coding is fantastic for building tools that you can't buy, or these, like smaller utilities, where it's like, hey, this is sort of like a one-off piece of software that I need to use to solve a particular job. Fantastic. We're seeing so much more …”
Wade Foster Mar 9, 2026 ▶ 6:09
Disclosure
Wade Foster: Software accounts for a single-digit percentage of company spend
“It's minimal. You know, it's like a single digit percentage. It's so small compared to like headcount.”
Wade Foster Mar 9, 2026 ▶ 6:45
Disclosure
Wade Foster used Cursor to build a tool from Granola meeting notes
“I got off the call, and I have cursor hooked up to granola, And so the meeting was recorded with Granola, and so I said, hey, Granola, our, go, go fetch that meeting I just had with Ryan, and in it he describes the war council he built. Build that for me. And …”
Wade Foster Mar 9, 2026 ▶ 10:18
Prediction Not checkable as stated
Wade Foster: AI agent skills will become cheap commodities like WordPress themes
“I could see skills kind of having, like, a business model like that, where it's like, you could buy, you know, buy one for five bucks, 10 bucks, there would be a bunch of free skills out there. You know, it's, it feels kind of like a very commodity thing at a …”
Wade Foster Mar 9, 2026 ▶ 12:18
Insight
Wade Foster: Interviewers struggle to reject candidates approved by previous rounds
“It's hard for humans to say no to, like, candidates that are good. It gets harder to say no as the interview goes along. So, like, two people before you said yes, now are you gonna be the person that says no? So you get more of that. And also humans I find are…”
Wade Foster Mar 9, 2026 ▶ 13:07
Disclosure
Wade Foster: Zapier uses a multi-agent AI to check hiring decisions
“I find it to be a really helpful tool at keeping the panel honest, keeping me honest on a decision where I can find myself sort of like feeling one way or the other, and then you talk to it, and it like really grounds the decision, and you're like, then you ha…”
Wade Foster Mar 9, 2026 ▶ 14:05
Opinion
Wade Foster: Claude Opus struggles with hard coding tasks compared to OpenAI
“One thing I've found is that like Opus is like more enjoyable to just like talk to and go back to it. Like sort of is an easier conversation list, but like when I'm having it work on hard problems, it sometimes just like bugs out. It doesn't do as good of a jo…”
Wade Foster Mar 9, 2026 ▶ 16:52
Disclosure
Wade Foster: Zapier is rolling out AI coding tools to non-engineers
“This week now we are rolling out company-wide. We're getting everybody in every function, not just engineering. So like marketing, sales, accounting, HR, all set up on either cursor, cloud code, or codex. One of the three.”
Wade Foster Mar 9, 2026 ▶ 24:58
Made with StarZero

Turn any episode into a week of clips.

This entire site, about 40 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.