Why AI Hasn't Sped Up Go-to-Market Strategy

AI has made go-to-market execution faster. Building the product, drafting the deck, sending a thousand messages, all of it moves at a speed that didn't exist three years ago. What it hasn't sped up is go-to-market strategy: which customer to bet on, which story to lead with, when to move resources off the plan you raised money on. At AI Infra Summit 2026 in Santa Clara, where I spent three days as a media partner recording conversations on the floor and in the Podcast Zone, the investors and founders working through those decisions described the same thing. They didn't figure it out alone, and they didn't figure it out from a deck.

Three things they said, and what each one cost someone.

Higgsfield's strongest engagement wasn't where the revenue was

Ray Wu is founding managing partner of Alumni Ventures' AI First Fund. Alumni Ventures is an investor in Lambda, Groq and Cohere, and PitchBook ranks the firm among the most active investors in AI infrastructure, so he's seen more infrastructure companies find and lose their footing than almost anyone I could have asked. He described Higgsfield, an AI video company he backed at Series A.

They started with enterprise and consumer. Enterprise engagement was strong, but revenue grew slowly.

The product wasn't wrong. Enterprise wasn't ready to buy. Rather than treating interest alone as product-market fit, the team kept testing where the strongest willingness to pay existed. That's harder than it sounds. Strong engagement looks like traction from the inside, and moving resources off the customer you planned for is not a decision founders make by looking harder at a dashboard. When Higgsfield turned to creators and prosumers, revenue moved by a multiple.

Ray's point about crowded categories applies to most of the AI infrastructure floor: a market full of demand doesn't mean the demand is yours, so focus first on the segment that's ready to move now. The difficulty isn't hearing that. It's telling which segment that is in your own numbers before the quarter is gone.

A cooling founder on the expo floor had the same lesson at a higher price. He'd tried his current technology twenty years ago, tied to another company, and both failed. The technology worked. Nobody needed it yet, and no one in the room could tell him.

Same product, different buyer, different proof

Stefan Krassin is co-founder and CEO of yasp, a German deep-tech company that spent two and a half years in stealth building an intelligence layer that automates the work of optimizing AI models for different hardware. The engineers who can do that by hand are scarce, a few hundred in the world by his estimate, and it takes them months. In his summit talk, Stefan described yasp's agents doing it in hours. The company came out of stealth this year, and when we spoke he'd just signed two end-customer license agreements.

Stefan ran an FPGA training and consulting firm for eleven years before this, and it shows in how he came to the summit. Not expecting contracts, he told me, but to build relationships. yasp already has a small number of close design partners he calls multipliers, companies through which it reaches many more end users, and he came to add two or three more. It shows in what he sells, too. Not just a smaller compute bill, but time: a team that gets an optimized model in a day instead of a quarter reaches its next milestone sooner, and that's the number buyers care about.

And it shows at the border. In Europe, the conversation about yasp is energy efficiency. In the US, it's speed and getting the most out of the hardware. Which buyer to lead with in a new market isn't a message you rewrite. It's a bet on where the company's next year goes, and yasp is placing it through those design partners, over months, with evidence.

Ray put the same problem structurally. In an enterprise sale there are four people in the decision: the buyer, the influencer, the beneficiary and the user. Each needs a different message.

AI multiplied the outreach. It didn't answer the question.

I expected an investor in AI infrastructure to say AI had compressed go-to-market. Ray said the opposite: it hasn't accelerated go-to-market, it's multiplied the volume. His phrase: AI scales outreach, but increasingly it's my AI talking to your AI. It's the instinct Ray described in technical founders generally: when you have deep expertise, you see everything as a problem and want to hammer it. More outreach is an easy hammer.

A Czech networking company I met is opening a US office because of a question no outreach tool can answer. Hyperscalers told them the technology was good, but they needed to know the company would still exist in five years. Stefan is planning a move to the South Bay by the end of the year for a related reason. Online meetings are useful, he told me, but the relationships get built in person. The decisions that matter get made there too, with the people who'll pay.

What Founders Getting Go-to-Market Strategy Right Have in Common

Not one of them described a shortcut. Higgsfield found its customer by noticing where money actually pulled, then making the uncomfortable move. Stefan holds direction with quarterly objectives the team freezes on purpose, so a startup that could change course every day doesn't. None of them decided alone, and none of them decided from a slide. They put the choice in front of evidence and other people before they spent on it.

That's the work I do with founders whose companies have real customers and are launching, entering a new market or changing how they sell. And it's what two rooms during Tech Week are for.

In San Francisco on October 6, with Sequoia, a private breakfast for twenty founders on what to know before your next sales, marketing or growth hire. Applications close September 28. Apply here.

In Los Angeles on October 13, with Worldpay, a working session for twenty-five founders who have a real growth choice in front of them and need to decide what deserves resources next. Bring the choice. Applications close October 5. Apply here.