Homebrew’s Hunter Walk has a long-standing habit of replying to every cold email. On September 24, he wrote that he “might need to” give it up. The reason: “clearly AI-authored and automated cold pitches” that cite blog posts he never wrote and “the most random of our portfolio companies.” He calls it “a digital tragedy of the commons,” one that’s “poisoning the channel for people actually doing thoughtful outreach.” He still ends with: “I’ll try my best.”
That’s one inbox. But it’s a clear picture of what happens when AI sits on both ends of the pitch. Founders use agents to find investors and write to them. Investors use AI to sort what comes in. Let’s look at what’s changing, what the data actually shows, and what still gets a founder noticed.
Founders Are Sending Their Agents
On October 7, Bonfire Ventures, an LA seed firm, published data from more than 1,000 website proposal responses over the past year. 25% of founders named an LLM as how they found the firm, ahead of LinkedIn (20%) and VC databases and Google (18% each). ChatGPT first showed up as an answer in May 2024.
Bonfire is careful about scope. These figures cover only founders who came through its website. They don’t include its main sourcing, which runs through investor relationships, referrals and companies its team seeks out.
The write-up also shows how fast the behavior changed:
- May 2025: a founder “uploaded my deck to ChatGPT and asked it for firms that might be interested in investing.”
- July 2026: a founder said they had “Asked Claude (Anthropic) to find and apply to the best seed funds.”
- September 2026: a founder had ChatGPT analyze more than 1,000 investors and rank them by fit.
Another founder built a system that read a Bonfire partner’s essay and scored funds on thesis fit, check size and conflicts. And Bonfire isn’t against any of it: “Do we care whether a founder fills out our proposal form or their agent does? Not really. We care whether we’re a fit.”
Everyone Uses AI, and It Shows
The WISH study of 100+ startups and about 40 investors and experts (reported by Startups Magazine on September 30) found that up to 95% of startups use AI in their communications. The top risks founders named were loss of personal touch or authenticity (64%), weaker critical thinking (62%) and messages that sound identical (54%).
You can see the sameness in the data. At Y Combinator, the share of applications using the word “wedge” went from under 1% in spring 2025 to over 20% in summer 2026, and applications have grown 60% longer over three years (Business Insider, August 6). YC partners point to AI-assisted writing as the reason, though word counts can’t prove who wrote what.
Investor Hemant Mohapatra estimated in July that almost 90% of the 300–400 cold pitches he gets each month are “very clearly AI generated.” That’s one investor’s estimate, not a measurement, but it echoes what Walk describes.
Investors Are Automating the Screen, Too
The Data Driven VC Landscape 2026 (released July 9) tracked 345 data-driven VC firms. In 2024, 86% of them were “Augmented” (human-led, AI-supported). In 2026, the split is 61% Augmented and 39% “Agentic,” the report’s own term for AI-led work with a human in the loop. That’s a self-selected group of data-driven firms, not all of venture. Within it, 45% plan to cut junior investor roles, and 49% say their top bottleneck is time and bandwidth.
Emerging manager Freesbee Ventures launched a $15M Fund II on September 29 and says, “Sourcing, screening, portfolio monitoring and investor reporting go through our own AI tools.”
What the Data Says Still Works
The most detailed comparison available is a vendor’s own data. CherryPitch, a fundraising-tools company, tracked 1,336 outreaches from 108 founders at pre-seed, seed and Series A from April to June 2026. Outcomes were partly tracked on its platform and partly self-reported. It’s one company’s cohort, so read it as a signal, not a market-wide rule.
- Fit: a curated, well-fitted list drew 28% replies vs. 12% for a self-sourced list.
- Personalization: personalized outreach drew 30% replies vs. 16% for generic sends (founders whose decks were at least partly clear).
- Warm intros: among founders whose decks stated their facts plainly, a warm intro lifted the reply rate from 24% to 38% and the meeting rate from 14% to 24%.
- After the meeting: 7.7% of warm-intro meetings advanced toward a round, vs. 7.0% after cold outreach. That’s about the same.
So in this cohort, a warm intro helped founders get the first meeting. After that, the company had to stand on its own.
An Agent Ran the Raise, and Investors Still Wanted the Founder
Polsia, an AI agent-orchestration company built to “run your company,” raised $30 million at a $250 million valuation in May 2026 (Pulse 2.0). Polsia’s founder said its agent “handled the data room, briefed investors, ran the back-and-forth on diligence. I joined the final calls.”
PitchBook’s May 29 reporting adds the friction. He still negotiated terms, adjusted the outreach and met the most promising investors himself. “Some [investors] got frustrated about access, they wanted to talk to me directly and the agent was holding the line.” At one point, after he agreed by email to meet a VC in person, the agent wrote back that it was leading the fundraise and he didn’t have time to meet.
In the same piece, Clarasight founder Adam Braun said he picked AlleyCorp’s Marshall Porter as his lead after a nearly two-hour coffee-shop conversation: “That human touch is actually a remarkable advantage.”
How to Get Noticed When Every Pitch Is Polished
Founders:
- Let AI research, and write the first line yourself. Use agents to build and check a well-fitted shortlist.
- Build a visible footprint early. WISH found 75% of startups wait for traction or revenue before communicating publicly. Short, regular updates give investors something real to find.
- Make fit obvious fast. Show why this investor, this thesis and this check size in your first two sentences.
- Spend warm intros where fit is strongest, then prepare as if the meeting is all that counts.
- Model, test, adjust. Track replies by list and by message, and change what isn’t working. Don’t automate things that aren’t working.
GPs:
- Publish a clear thesis. Founders’ tools score funds on what partners publish. A vague page draws the wrong pitches, or none.
- Be easy to reach and quick to reply. Freesbee promises a reply within 5 working days and a decision within 6 weeks.
- Compete on what a model can’t scrape: co-investor trust, founder referrals and being someone’s first call.
The Signal That Can’t Be Automated
When a convincing cold email takes seconds, the email tells an investor very little. What’s left is what’s hard to fake: a track record people can see, a thesis that actually matches, and a credible person who’ll vouch for you.
Those trust paths can be mapped. Co-investment histories (SEC Form D related persons), adviser filings (Form ADV), shared nonprofit boards (IRS 990s) and business registries show where a founder and an investor already overlap, often through someone a step or two away. ConnectLab.live maps those paths, turning “who should I pitch?” into “who already knows them?”
Connection is the currency of the 21st Century, and it runs both ways: founders find the investors who fit, and investors find the founders already one introduction away.
When an investor opens your email, what’s the one signal in it that a machine couldn’t fake?
If you’d like help building that signal, here’s an open invitation:
- Founders: sign up for ConnectLab’s connection system for founders at https://connectlab.live/founder
- Fund managers: email ken@connectlab.live to claim your fund on ConnectLab, so well-fitted founders and co-investors can find you.
All the best,
Ken McArthur
Best-Selling Author and Producer