Jason M. Lemkin recommends starting with a simple horizontal agent like Delphi that handles basic tasks with minimal setup, then building to more specialized agents once you've proven it works. Early wins create internal buy-in for bigger AI deployments.
Chapters
The learning for folks that watch, I didn't know this at the time, stair-step it. So the fact that we went from Delphi, a horizontal agent, which basically can self-ingest your content with no training, like you do need to, but it's the least amount of training of the tools in our stack. Start easy, start broad, get a few, get a win under your belt. Even if you started with say qualified, I'm not a total qualified, start for the simplest use case, okay? Because if you fail with your first agentic thing, not only will it, is it frustrating, you won't know, you won't know, you just won't know because we had the early win that that was better than expected. Like, our general digital Jason could do some sales, it could do some qualification, and it would start to say, you know what, Craig, you're not big enough for SaaStr Annual this year, don't come. That was like great for us. Like, it would do it on its own. So we gained confidence. And if you can stair-step this stuff, you will build confidence by your second, third agent or deployment. And we didn't intend— but that's what we ended up doing. That's how even us were successful. We stair-stepped.
I think every company is probably a little bit different, but like the stair-step We always talk about you gotta, you gotta crawl, walk, run. Everybody like sees the prize of like, I'm gonna be sprinting. And you're like, whoa, you know, where do I start? And you gotta get the quick win. You gotta build the confidence internally.
