Getting Right

I have long been a fan of Zapier and enjoyed reading this interview. Zapier CEO Wade Foster’s approach to building an AI-first company centers on blending agentic intelligence with deterministic workflows, advocating the “90% rule” where AI handles the majority of a process but humans supervise the crucial last mile. Foster emphasizes that perfect automation is a myth—most agent-generated outputs are about 90% correct, so designing systems with human oversight ensures quality and builds trust, especially in sensitive contexts like renewals and customer interactions. Companies should avoid rushing to full automation; instead, hybrid solutions using both AI agents for flexible tasks and deterministic protocols for predictable processes achieve the best outcomes.

Foster also stresses that AI fluency must be defined per role, not just company-wide, recommending regular hackathons and weekly show-and-tells to sustain AI literacy and cross-team inspiration. Prescriptive selling powers competitive advantage, as buyers now expect sales teams to leverage AI-driven context and recommendations. Critical mistakes that stall AI rollouts include believing agent hype, skipping functional definitions, choosing full automation prematurely, and failing to build an automation-ready culture. Ultimately, the most successful SaaS leaders will be those mastering intelligent hybrid workflows, tailoring AI adoption to specific functions, and keeping humans strategically in the loop.

The model faces several practical limits and risks that can challenge its adoption in other organizations or industries. Regulated fields like healthcare or finance may demand stricter accuracy than a 90% threshold allows, while scaling ongoing training and human oversight can strain company resources. Cultural resistance, workflow complexity, and unpredictable AI errors also pose significant hurdles, indicating that a one-size-fits-all approach is unlikely and that successful hybrid AI deployment depends on context, company readiness, and ongoing adaptation.

Zapier’s own public account highlights frequent missteps in workflows (such as AI hallucinations or broken JSON structures), showing that building reliable and context-aware agents remains a significant challenge even for companies deeply invested in the automation space