Flo Crivello from Lindy shows how to build agents for scheduling user research interviews and researching sales leads at scale.

​
First use case: UXR Scheduler

Flo demonstrates building an agent that schedules user interviews from a CSV list. The clever part: he adds delays between outreach to prevent double-booking. His CSV has columns for user info plus “how long to wait before sending availabilities” - ensuring the agent sends availabilities sequentially as meetings get booked. Here’s Flo’s workflow: First, set up the trigger: “You’re going to receive a message with the information about the user. You are going to set a timer and wait.” Second, configure calendar search with minimal settings - just specify 30-minute meetings within the next 7 days. Flo emphasizes: “I’m going to do very little configuration.” Third, let the agent compose emails. Flo provides a simple template: “Hi [first name]. I am the founder of Lindy. We want to talk to users to ask some questions about a new feature we are working on. Would you have time at any of the times below?” He notes: “This is literally just prompting Claude.” Fourth, add AI-powered conditions: “Go down this path if the user accepted one of the times.” If they accept, create the calendar invite - again with no configuration needed. When Tal asks about the delay feature, Flo explains: “I am doing this on purpose to stagger the tasks so that Lindy doesn’t send availabilities to users that she has already sent to previous users.” The magic happens when you drag and drop a CSV with multiple users. Flo notes: “This is only 5 tasks, but it gets really magical when you do like 500 tasks at once, because agents obviously scale super well.”

​
Second use case: Lead Researcher

Flo’s favorite demo shows an even simpler agent: receive lead information → research on internet → reach out by email. The prompt: “Draft an email using the news you just found about them and asking whether they’re looking to expand their real estate footprint.” When Flo drops a CSV of 50 unicorn founders, the agent immediately starts researching each one. For Eric Lehman from Ramp, it drafts personalized outreach based on recent company news. Flo acknowledges: “Obviously, you will have to iterate on the prompt so it doesn’t feel as AI-generated.” Flo notes this is “very, very helpful for sales, and that’s one of our top use cases, for sure.” ➡️ Minimal configuration beats over-engineering. Give agents simple instructions and let them figure out the details - they’re surprisingly good at inferring what you need from context.