Evidence receipt / evaluation
Published · transcript-backedJessica Fain: evaluation
22 Mar 2026 Lenny's Podcast The art of influence: The single most important skill that AI can’t replace | Jessica Fain (Webflow, ex-Slack)
“We talk about this a lot, especially when we think about identity, security, compliance, that agents are a teammate for accelerating work, and the risks that that also provides because of the places where agents fall down.”
Source trail
Everything needed to verify it.
- Speaker
- Jessica Fain
- Attribution
- Verified speaker
- Claim type
- evaluation
- Recorded
- 22 Mar 2026
- Publisher
- Lenny's Podcast
Transcript context
…Okay. Maybe one more question along these lines, just agents now running as colleagues. Everyone's got all these Claudes running, Codexes, all kinds of agent platforms. Now that agents are a part of the workforce and increasingly, is there anything there around influence that is valuable for people to start thinking about to learn how to do? I love that question. We talk about this a lot, especially when we think about identity, security, compliance, that agents are a teammate for accelerating work, and the risks that that also provides because of the places where agents fall down. I think agents are absolutely something we have to think about with relation to influence, because we are basically all directors of work now, right? If we have an army of agents or this hundred new colleagues that joined our team, how would we onboard new junior team members who don't already understand our product philosophy, what's important to us, and how do we codify that for ourselves? How do we take the time to say, "What's important to me as a product leader that I want to inculcate into anything that's developed for my team? What do I believe about product market fit? What do I look for in terms of what success looks like? How do I set metrics?" And if we can do that upfront work to actually analyze ourselves and where we've seen success, where we've seen downfalls, and then continue to train our agent models with that data, that history, that context, they will be better off. But I also think the flip side of this is guardrails. Where do you need to be involved? Where do you need to catch hallucinations or missteps? Where do you uniquely have taste, judgment, frame of reference, something intangible where you need to say, "Don't do this without me." What's so funny about this, essentially we're helping the agent learn the skills we've been talking about to influence you. Here's context on my goals, here's what I understand about the customers, here's our current priorities. Help me. Tell me what to do. And it's like influencing you to agree.…
Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.