<aside> 👋
I’m Charles, a 0-1 PM / ex-founder. I’m excited about the AI capabilities product role at Zapier.
I’m writing this document to highlight my fit. I chose to use Zapier Agents and write a friction log because I wanted to build empathy as a user, and solve a problem I was actually having: writing personalised emails to attendees of a meetup I’m hosting.
</aside>
Author: Charles Dupont
Date: December 14th 2025
About me: I have built and deployed agents for personal, local workflows, but I am new to Zapier Agents. This test is based on personal (messy, small) data**,** not enterprise scale use cases.
My goal: Send personalized messages to Lenny’s community meetup attendees. I have been doing this manually, but it’s taking longer as the group grows. It’s important that people I know get genuine outreach, not AI slop. Think: Zapiers’ ‘Sales Email Creator’ agent template.
My approach: This doc goes from idea → emails sent, noting friction points as a user. I’m following best practices from Stripe friction logging, with a feature pitch at the end
For Zapier Agents, I assume most usage will be new users who are looking to try it out by completing a specific task, and that the team wants to convert these testing users into regular users. The longer it takes for setup, the more likely the user is to churn.
<aside> ⭐
Goal: Faster time-to-successful-run for AI agent creation
</aside>
We can extract prioritized themes from the friction log, with drivers for prioritization.
| Priority | Scale | User Impact on time-to-successful-run | Related observations |
|---|---|---|---|
| P1: Mental model for ‘agents’ is unclear | High Affects all new users. No standard exists. | High I can’t use a tool I don’t understand it | (1-3), (6), (8), (9) |
| P2: Agent configuration makes me think and mistakes still happen | Medium. Affects most users, eventually. | High. Each step costs time, especially fixing a mistake. | (6-7), (9), (11) |
| P3: Black box process + wide-ranging permissions creates anxiety | Low. Affects few users; most already have high-trust and existing permissions setup with Zapier | Medium. I can’t trust a powerful system I don’t understand, but this is well mitigated by agent preview. | (4-5) (10) |
Planning mode questions is the common pattern of clarifying user intent when creating a plan. Users questions before implementation create alignment, and brings forward all of the right context. That means users quickly ‘get’ the Zapier agent mental model (P1), and there are fewer opportunities for mistakes (P2)
Specifically, I recommend experimenting with quick multiple-choice clarifying questions focused on scope, evaluating the impact on time-to-success as measured by time from new agent → run.