Founder-led GTM in the AI era

We had a fascinating discussion for this month's AI2 Incubator founder roundtable, focusing on a question a lot of early-stage teams are wrestling with right now: what actually works in go-to-market today?

A big theme was that generic outbound is getting less effective. The old playbook of building a broad list, writing a decent message, and hoping volume carries the day just doesn’t go as far as it used to. People are overloaded, inboxes are crowded, and everyone has access to roughly the same tools. The bar for relevance is much higher now.

One framework that stood out came from Dan Moore, co-founder at Ergoly.ai and Tarka and a longtime entrepreneur. Instead of thinking only in terms of ICP, think in terms of a pain-qualified segment. In other words, don’t just ask whether a company looks like it should be a fit. Ask whether there are visible signs that this company is actually dealing with the problem you solve. That might show up in hiring, product launches, tooling choices, gaps in their team, or other signals you can pick up from public information online.

That naturally leads to a second idea: if you’re reaching out, give the person something useful right away. Dan described this as a permissionless value prop — creating value whether the prospect replies or not. The example he shared was from Tarka: identifying startups that seemed to be underusing PostHog, analyzing their setup, and sending a company-specific set of recommendations. That’s a very different kind of outreach from “saw your profile, would love to connect.” It’s real substance, and it shows the product or service through the outreach itself.

Another point that came up a few times: list quality matters more than people want to admit. There was a lot of emphasis on the idea that most of the work is upstream. Defining the right audience, pulling the right leads, cleaning the list, and being honest about who actually belongs on it all matter more than tweaking copy over and over again. A weak list makes everything downstream harder.

The conversation also drew a useful distinction between discovery mode and scale mode. For discovery, the advice was to keep it simple: use your own LinkedIn, use your network, and focus on talking to the right people as directly as possible. Second- and third-degree connections were called out as especially useful because warm context helps a lot. Once you know who responds, what resonates, and what pain is real, then you can start thinking about scaling.

LinkedIn came up as one of the best channels for this early-stage work. Not because it’s perfect, but because it gives founders a practical way to reach relevant people with some amount of trust built in. A few tactical ideas came up too: keep the connection request short, don’t overcomplicate it, and consider using voice notes in follow-up since they still feel more personal than standard automated messages.

When the conversation shifted to email, the tone changed a bit. The takeaway there was basically: if you want to do it at scale, you need to treat the infrastructure seriously. Separate domains, inbox warmup, verification, volume limits, reputation management — none of it is exciting, but it matters. If you skip it, the channel breaks.

One of the more interesting themes from the roundtable was that hard-to-copy GTM is getting more valuable again. As more teams use the same automation tools and the same playbooks, the advantage shifts toward things that take real thought and effort: custom research, tailored assets, smart event strategy, curated dinners, and other kinds of outreach that don’t feel mass-produced.

That was probably the broader message of the session: AI can help speed things up, but it doesn’t replace good judgment. The teams that stand out are still the ones that understand their buyers well, know how to spot real demand, and put something genuinely useful in front of the right people.