Zapier’s CEO shares his personal AI stack | Wade Foster

Wade explains why CEO-level involvement in AI adoption is critical, contrasting effective leadership with the common 'delegation trap' where executives write AI memos but don't engage directly. He emphasizes the importance of hackathons, show-and-tells, and creating playspace for teams to experiment with AI tools hands-on.

• •Avoid the delegation trap: Don't just write an AI memo and ask your exec team to cascade it down—that poor IC won't succeed
• •Create structured playspace: Run hackathons, show-and-tells, or Friday afternoon experiments to let teams get hands-on with tools
• •Fear dissipates with usage: Once people actually use AI tools, they see both the capabilities and limitations, making adoption more pragmatic
• •Measure and reward AI fluency: Build rubrics for different roles (like PM levels) to make AI skills concrete and career-advancing

Wade demonstrates using Granola's meeting transcripts with a 'recipes' prompt to extract the company's actual unspoken culture versus stated values. This workflow analyzes aggregated meeting data to reveal how the organization truly operates, providing material for job descriptions, hiring rubrics, and performance evaluations.

• •Use Granola's 'unspoken company culture handbook' recipe to analyze all your meeting transcripts and extract real behavioral patterns
• •Compare extracted culture against stated values to identify alignment gaps and undocumented cultural norms worth reinforcing
• •Turn culture insights into actionable hiring materials: job descriptions, interview rubrics, and candidate evaluation criteria
• •CEOs should analyze their own meeting patterns to understand how they actually operate versus how they think they operate

Wade shows his Zapier agent that automatically evaluates interview candidates against job descriptions and company values. The agent triggers when Granola adds interview notes to a folder, analyzes transcripts, and emails a yes/no/maybe recommendation with reasoning—acting as both a bias check and thought partner across diverse roles.

• •Set up automatic triggers: Agent activates when interview notes are added to a specific Granola folder
• •Upload knowledge sources: Include job descriptions and values rubrics as context for the AI evaluator
• •Get structured recommendations: Agent provides yes/no/maybe decision with 3-5 sentences of reasoning based on functional expertise and values alignment
• •Iterate on prompts: Use actual evaluation outputs to refine what constitutes 'good' vs 'bad' answers, making the agent increasingly accurate
• •Add interviewer coaching: Evaluate the quality of your own interview performance and get suggestions for improvement

Wade demonstrates using Zapier's Copilot feature to modify agent prompts in real-time, showing how to remove PII from candidate evaluations and incorporate new suggestions. Copilot helps users write better standard operating procedures for agents without requiring deep prompting expertise.

• •Use Copilot for rapid iteration: Give basic instructions and let it generate proper SOP-style prompts
• •Implement suggestions in seconds: Adding interviewer feedback and subject line improvements took under 60 seconds
• •The real challenge is ideas, not implementation: Coming up with what to improve is harder than actually building it with AI
• •Think beyond current capacity: Ask 'if I had more time' or 'if I had three interns' to imagine fuller workflows

Wade explains how AI agents enable tasks that were previously too expensive or tedious for humans to do consistently. He emphasizes that the real value isn't just automating existing work, but doing the 15 things you'd do with infinite resources instead of the 3 things you have capacity for now.

• •Many valuable tasks simply don't happen today because they're not economically viable for human labor
• •AI agents excel at consistent, tedious work that humans find annoying but would be valuable if done reliably
• •Expand your thinking: Don't just automate what you do—imagine what you'd do with unlimited time and interns
• •Example: Marketing teams could execute 15 follow-up actions from sales calls instead of just 3

Wade reveals his unconventional recruiting strategy using Grok to find 'diamonds in the rough' on X and YouTube. By searching for people who post about Zapier, no-code, and automation with modest followings outside the Bay Area, he discovers talent pools that traditional LinkedIn recruiting misses entirely.

• •Use Grok for natural language talent searches that are difficult in LinkedIn's Boolean interface
• •Search for specific behaviors: fans of your product, tutorial creators, education-focused posters with modest followings
• •Filter by geography and authenticity: Look for real profile photos, US-based candidates, avoid bot-like accounts
• •Discover unexpected talent pools: Found concentrations of no-code talent in India and Nigeria
• •Expand beyond hiring: Use same technique for influencer marketing, product feedback, and community building
• •Iterate rapidly: Recruiting is a numbers game—use Grok to increase surface area for finding interesting people

Wade addresses which roles are still highly competitive despite AI transformation. While top talent is in demand everywhere, hyper-specialized analyst roles focused on single tasks are most at risk. The key is elevating yourself by learning to build agents and orchestrate AI rather than being replaced by it.

• •Engineering and engineering leadership remain insatiably in demand
• •Top talent is competitive everywhere: designers, recruiters, PMs, marketers, sales reps all still needed
• •At-risk roles: Hyper-specialized analysts doing tasks that are now 'just a prompt'
• •Elevation strategy: Learn to build agents and orchestrate AI to redeploy yourself in higher-impact areas
• •Organizations with 'fleets' of specialized workers face the toughest transitions
• •What it means to be 'top talent' has fundamentally changed—AI fluency is now table stakes

Wade describes Zapier's AI-transformed workplace where feedback bots are everywhere, emoji reactions trigger automation suites, and new employees accidentally launch workflows. He sees the next frontier as breaking down silos between job families, though he's realistic about the timeline for roles fully merging.

• •Feedback bots provide constant coaching across the organization, even for the CEO
• •Slack is 'unhinged' with emoji-triggered automations—new employees often accidentally trigger workflows
• •The 'don't be a robot, build a robot' value predisposed Zapier to aggressive AI adoption
• •Next frontier: Breaking down silos between job families (PM, designer, engineer becoming 'builders')
• •Realistic timeline: Full role convergence is more likely by 2030 than 2025