Thoughts on GTM Engineering
A few months ago I went to Gradient’s GTM Engineering Dinner, where they opened up with “All 40 of you applied, and we chose the best 20 of you.”
That’s how unknown this role used to be. Now every company seems to be hiring GTM Engineers.
I've now been a Founding GTM Engineer for a little over six months. Here are my 5 takeaways:
1. If it doesn’t work manually, AI won’t save it
Before you build an agent or a workflow, the motion has to work without one. You shouldn’t aim to automate everything. It should first work on its own, and then you should scale it with AI. Don’t try to over-engineer things that might not be even useful.
2. Growth is science - really
I always say that you really have to treat growth roles more like scientists who conduct experiments. Especially early on, you really don’t know what will truly work. You can have hypotheses, but until you actually test them, you don't have the full picture.
3. Personalization >>> Volume
I have been there. Beautifully enriched lists via Clay, multiple A/B tests, and… only got a couple of replies. I was still lucky to get these, but it felt like throwing darts into the dark.
Broadening the top of the funnel is important, but the real key lies in personalization.
Narrow the list as much as you can and lead with value plus proof you did the research. Instead of “would you have 15 minutes for a demo,” say "you have problem X, here is why that's expensive, here is how this fixes it at your company." Then scale.
Also, especially in early stages, seeking feedback always beats selling.
4. Track data like your company depends on it - because it does
PostHog has become my bible. Google UTM builder has become my best friend. With every change I make in the onboarding, features, design, or in GTM initiatives, I make sure that we have appropriate tracking in place.
PostHog does take time to properly set up. However, it will be one of the best investments you can make for your company.
5. Start with 1 channel and 0 agents
If your team is small, don’t try to do everything. Start with 1 channel, and once you nail it, experiment with others. You don’t have unlimited resources.
And please, don’t write everything with AI.
It's pretty easy to distinguish when a LinkedIn or Reddit post was generated, and I won't read one if I spot the em dashes. I was guilty of that one too - my OpenClaw Reddit accounts were banned. It taught me that going organically and learning about the community and how they interact is always better than relying on agents. You can learn a lot about your users in the process.