Tyler Phillips

Tyler PhillipsSep 9, 2026

Picture this: it's 1995. You're selling… penny stocks (oops you're Jordan Belfort 😂). You sit down at your desk, thumb through your Rolodex, pick up the rotary phone, and start dialing. Whoever answers, answers. Thirty years of new software later, most teams are still guessing who's worth reaching out to. Time to upgrade to Apollo.io

Tyler PhillipsSep 11, 2026

Sneak peak....but Apollo.ioNEXT, we will be announcing something I have been evangelizing internally called "agentic loops for GTM." And I've never had more clarity in my Apollo life when it comes to what we should build than agentic loops for GTM. Okay, so first - what do I mean by an agentic loop? My definition is: An agentic loop is a job the agent runs end to end, on its own, again and again, and gets better at every time. It has three characteristics: (1) a metric to optimize for, (2) a fast enough feedback loop, and (3) continually learns over time from experience. Most importantly, an agentic loop can only apply to areas where there is a right answer, not a taste call. This is why loops originally came out of the coding case (as everything seems to now in AI). An agent can score its code against unit tests, receives feedback immediately, and learn from the score to improve over time. This is what makes coding perfect for loops. One example of an agentic loop in GTM can be campaign optimization. The metric is reply rate, the agent can get feedback every 1 or two weeks, and learn from what changes it made to improve for the next time. Agentic loops are not a fit for all of GTM use cases (e.g. writing an initial personalized message), but for some, it makes perfect sense. Agentic loops in GTM, in my POV, is the only path forward for a differentiated value proposition vs. LLMs and MCP/APIs. It compounds with usage only your product owns, it survives the habit problem (AI products inside products struggle with retention - but this one just works on your behalf), and it has unlimited upside (both for the user in terms of more pipeline and the business in terms of expansion revenue). And don't take my word for it: one of our early beta testers, Rushpal Singh, 5x'd his reply rate with our campaign optimization loop. We are looking for more beta testers in case anyone is interested, comment or DM me! Or join us at ApolloNEXT to learn more (registration in comments)!

Tyler PhillipsAug 19, 2026

Three things I’ve learned building Apollo.io’s API products (as a total newbie API PM): 1. Customers who get throttled actually spend more and retain at a higher rate. You don’t have a churn problem, you have a rate limit and a billing problem. 2. The problem is almost always in the 1-3 API calls. If there is an error after 2-3 retries, there’s gone. If you use Codex or Claude Code, you’ve already seen it (eg API error, retry, error, move on to next API). 3. Docs are your best friend. If your docs suck, your user’s agent will suck using it as well. Used to humans, now it’s Claude Code or Codex. That’s all, more to come soon. P.S. let me know if anyone has any feedback on Apollo’s APIs 🙏