Usman Salis
Artificial intelligence is changing how companies automate work—but maybe not in the flashy way headlines suggest. While "AI agents" dominate tech conversations, Zapier CEO Wade Foster thinks the real opportunity is elsewhere: agentic workflows. These are structured, predictable automations that use AI but stay firmly under human control.
Foster's Five Principles for Practical AI Automation:
Recently, Peter Yang, an AI commentator and founder of Creator Economy, shared five key insights from an interview with Foster. His takeaways offer a grounded roadmap for businesses trying to adopt AI without getting lost in the hype around fully autonomous agents.
1. Start with Workflows, Not Agents Most teams think they need "an AI agent," but what they actually want is an agentic workflow—a well-defined automation powered by AI but guided by logic and human decisions. As Foster puts it: "When people say they want an agent, more often than not they just want an agentic workflow. The latter gives you more determinism, reliability, and cost advantages." Predictable automation beats experimental autonomy every time.
2. Keep Each Agent Focused on One Task The best AI agents do one thing really well. Foster shares an example from Zapier: an internal email agent that sifts through 100 emails and surfaces just 10 priorities, sorted into categories like "Action Required" or "FYI Only." Narrow specialization makes agents more reliable and easier to improve.
3. Build Complexity Gradually by Chaining Tasks Foster recommends starting small—maybe with an AI that generates a prep doc—then linking multiple automations together over time. "You'll realize maybe I could actually chain these things together and do something way more impressive." This modular approach keeps systems manageable while unlocking sophisticated multi-step workflows.
4. Treat AI Like a Junior Employee You're Training One of Foster's most practical tips is conceptual: think of AI as a junior team member, not a perfect machine. "Instead of me answering all these emails, I'm thinking what new instructions I can provide to the AI so that it does a better job." By reviewing outputs and refining prompts, you turn AI into an assistant that improves over time.
5. Use AI as a Thinking Partner, Not a Replacement Foster draws a clear line: AI should enhance human creativity, not substitute for it. "You're using it in a smart way—you're going back and forth with it and providing context. That's very different from somebody who just turns in AI's initial output as their homework." The best results come from iteration and collaboration, not delegation.
Why This Approach Works
Zapier's strategy offers something rare in AI discussions: a realistic blueprint. According to the company, nearly 90% of its internal teams now use AI tools daily. But scaling that success requires structure—exactly what agentic workflows provide.
They deliver reliability (predictable results without AI "hallucinations"), efficiency (lower costs per task), and control (clear oversight and audit trails). Foster's framework turns hype into discipline, transforming AI from an experiment into a repeatable business system.
Usman Salis
