Starting an AI GTM agency from scratch in San Francisco at 35, while also caring for a newborn—I have to hand it to you. During the day, you’re dealing with clients, breaking down sales processes, and keeping product implementation on track. At night, you’re getting up to feed the baby and change diapers. A lot of decisions get made in those stolen moments at two or three in the morning. Living and hiring in the San Francisco Bay Area is already expensive, and a startup can’t bring on a big team right away, so every dollar and every hour has to be carefully budgeted. At this stage, the biggest danger isn’t being busy—it’s being busy in the wrong direction: chasing one flashy feature today, switching to a different acquisition channel tomorrow, and eventually pulling the team apart.

Using AI to find customers for B2B companies sounds simple, but the execution is full of details: where the leads come from, how to decide whether they’re worth pursuing, and how to put the messaging into practice. Lead sources might include company career pages, funding news, changes in tech stack, industry event attendee lists, or dormant leads in a CRM. But once you’ve found them, you still have to clean and deduplicate the data, fill in contact details, and figure out whether the company has a budget and a pressing need right now. Deciding whether a lead is worth pursuing takes more than looking at its industry tag—you need to weigh the strength of the signals. Did the company just raise funding? Is it hiring for relevant roles? Did a key executive just change? Has it been visiting the product page frequently lately? And putting messaging into practice isn’t just a matter of having a model write a bunch of emails at once; you need to tailor them by role, industry, and scenario, then keep refining them based on reply rates. A common observation in the industry is that salespeople spend a lot of time finding leads, organizing records, and following up repeatedly, leaving relatively little time for in-depth conversations. And when B2B leads don’t get a prompt response, conversion rates can drop significantly. So the real challenge isn’t the model—it’s breaking down all the messy work of sales into tasks a machine can handle, then making sure it does them reliably, with clear tracking and a way to review what happened.

One observation: these tools ultimately come down to closing the data loop. The companies that first figure out how to “find leads and follow through” are the ones that survive. “Find leads” means having broad lead sources with quality you can control. “Follow through” means data from outreach, replies, meetings, and closed deals can all feed back into the system, making the next recommendations more accurate. A polished lead list without follow-up and feedback will quickly hit its ceiling. After all this time, I’ve learned that managing position sizes is always more important than calling the direction of the market. Startups are no different: discipline is worth more than ambition. Wishing you the best as a new dad at a startup—hope you don’t burn out on both fronts. Maybe I’ve got it wrong. I’ll wait to be proven wrong.