I recently sat down with Austin Hay for a conversation about how founders are using AI to accelerate their GTM motions—where it's creating real leverage, where it's just changing existing workflows, and which use cases go beyond basic content generation.
If you haven't seen it yet, here's the full session on YouTube:
About our guest
If you don't know Austin Hay, he's an operating partner at Khosla Ventures who works with founders and GTM teams every day, and he's previously helped build the GTM systems at Ramp, Notion, and Replit.
He also writes the Growth Stack Mafia newsletter, a newsletter I love, about the GTM systems he's building with AI.
Basically, he has a front-row seat to what's working and what's hype. Here's what stuck with me from our conversation.
Solve the boring stuff first
The single line I keep coming back to: "All the stuff that's boring is the stuff you should solve first with AI, not the stuff that's hard."
Austin sees a lot of founders trying to build an autonomous, do-everything growth machine on day one. His advice is the opposite. Start with the boring, administrative, process-driven work—expenses, campaigns, list cleaning—the discreet tasks that aren't probabilistic.
Those are actually things where you're gonna get a lot of value and time back, because AI is very, very good at doing these things.
The flip side is content. Austin doesn't think AI is good at writing genuinely good content — and he has the receipts: "If AI is good at writing really good content, we wouldn't see so much slop on LinkedIn right now."
His take is that writing serves two people at once. "The part for you is for learning, part for the reader is understanding." So he writes his own articles from the heart and only brings AI in afterward to copy edit and SEO edit. A much better use of the tool.
Every job is a collection of skills
The bigger idea underneath all of this: "Every job at a company can be broken down into skills."
A skill is a markdown file that describes a job to be done. If you want to future-proof how your team works, he suggests everyone take each piece of work they do and skillify it. Turn a problem into a skill as you solve it, rather than asking AI to rote-generate a pile of skills you'll never use.
Do something once yourself, keep doing it, and then once you've proven that it works, then promote it to something autonomous.
Skillify first, automate second.
Austin also recognizes the tooling to share all this is still messy. GitHub works as a "harness" for engineers, but the whole idea of GitHub as a harness just totally falls down when you remember that many people don't know how GitHub works. The marketers who pick up technical skills will get more powerful, but you still have to design systems around the way your team actually works.
AI is great for reporting
On reporting, Austin's most useful framing was splitting business reporting from directional reporting.
- Business reporting should be done one way by everyone, off the production database using agreed-upon queries.
- Directional reporting is where teams can move fast with whatever tools they have, as long as they don't burn weeks reconciling numbers.
"Every tool will diverge at least five to ten percent" for a dozen boring reasons, and chasing that gap is where organizations spin their wheels.
The founder-led sales unlock
Austin's most concrete tip for founders: build a meeting follow-up bot. A skill that runs at least once a day, reads your call recordings from whatever tool you use, summarizes the action items, and drops a drafted follow-up email into your inbox. Feed it a couple years of your own emails first so it learns your voice, add a "humanizer" layer, and after a few days "that skill will be writing amazing draft emails for you."
The hard line he drew: don't let it hit send. One bad interaction can sink a deal.
AI is not good enough to send emails for you automatically. And founders would be insane to ever let AI send an email to a customer.
The myth of "auto mode"
The conversation led to an audience question a lot of founders are quietly wrestling with: can't I just flip GTM to auto mode?
Austin's answer was blunt: "The idea of auto mode right now is like not real." You can get close, but you can't remove the human who reviews or pushes the button.
Less is more
We closed the discussion on tooling. For a founders and operators building their stack from scratch, Austin's mantra was simple: "Less is more, baby."
Your job is to drive engagement and close deals; anything else is a waste of time. Find your most repetitive, time-consuming tasks, do them manually five to ten times, turn them into skills, then promote it to a cron. You'll discover the tools you actually need along the way—as Austin puts it, "tools are just meant to solve problems."
Thanks to Austin for a great conversation. If there's one thing to take back to your own GTM systems, it's this: skillify the boring work first, keep a human in the loop, and let the fancy stuff come later.
Agent-led customer engagement
The same playbook Austin describes for GTM—skillify repetitive work, keep a human in the loop, and promote what works—applies to customer engagement.
Knock gives founders and growth teams agent-led infrastructure to build lifecycle messaging, onboarding flows, and product notifications across email, in-app, push, SMS, and Slack, with the guardrails to review before anything goes out.
Create your free Knock account to get started, or talk to our team to learn how Knock can power AI customer engagement at scale.