What a16z's Lighthouse-or-Landgrab framework gets right, and the part it leaves in the dark.

Last week, a16z's Joe Schmidt IV and Julian Marx published one of the sharpest go-to-market pieces we've read this year. The argument is clean. AI companies selling into the enterprise face a choice between two playbooks, and the market, not the product, decides which one wins.

The line in the blog post is the one worth keeping. “Your buyer doesn't purchase the future; they purchase either proof or math.”

We agree with almost all of it. However the 2x2 quadrant quietly drops the part that matters most to any startup trying to swim upstream into the enterprise. Proof doesn't travel on its own. Someone has to carry it. And that carrying is the work most founders underrate, right until the deal they were counting on goes quiet. Or worse, when they have no customer references or evidence on the heels of launching their new enterprise product.

Lighthouse vs. Landgrab, briefly

a16z lays out two plays. The Lighthouse: win a few marquee customers in a brand new category, and let their credibility unlock the rest of the market. As their post highlights, think Harvey in legal or Hebbia in finance. Long sales cycles, founder-led, heavy custom work, and a large payoff when the logos land. The Landgrab: skip the prestige race, win on math and speed, and close on ROI before an incumbent bolts AI onto its product to compete. a16z points to Decagon and Stuut here, companies that win a high volume of customers fast with a standardized product, rather than chasing a few prestige logos.

You choose between them with two questions. First, how much does the buyer risk when they sign? Risk runs high in regulated industries, when you replace a system of record, or when the output faces outside scrutiny like a compliance review. It runs low when a wrong call only annoys someone internally. Second, does social proof travel? In concentrated, status-driven industries like law, private equity, and hedge funds, landing two marquee firms can move the whole market. In fragmented ones, as a16z puts it, the controller in Des Moines doesn't care which blue-chip brand signed on.

It's a genuinely sharp map. It doesn't, however, go far enough on another key facet that gets left in the dark.

We propose a staircase, not a fork in the road

The graphic makes the Lighthouse and the Landgrab appear like a fork in the road. Pick a lane. But in practice, it's usually a staircase.

Win and document two or three real marquee deployments. Use them to earn enterprise trust and category legitimacy. Then convert the repeatable workflow and the proven ROI into a broader landgrab. a16z's own essay actually allows for this. It points to Affirm as the example and how they landed Casper and then went on to secure every mattress company before moving on to target exercise equipment like Pelotons. It's the reshared graphic that flattens the choice into either-or. The founders who read only the grid miss the sequence tucked inside the full piece.

Proof doesn't travel on its own

Here's where we propose a different facet with the AI Sales Strategy map. a16z treats “does your proof travel?” as a fixed property of your market. In reality, it's not. Proof travels because someone makes it travel, or it sits still because no one did the work to spotlight the proof.

There's another trap to keep an eye on early that startups can fall into. Say they land a brand-name pilot, with a company that everyone would recognize. Then it sits collecting cobwebs under a stalled POC or an NDA. Unnamed, invisible and in the dark.

That, my friends, is a lighthouse with the bulb turned off. The proof exists. It simply can't do the one job a lighthouse has, which is to be seen from afar.

Not every lighthouse market lets you turn the bulb on

Now let's push on the “does proof travel?” question and go one more level deeper, because the Lighthouse quadrant isn't one thing. Even inside the markets where proof is supposed to travel, a hidden variable decides how far it actually gets to proceed.

How nameable is the proof allowed to be? We see it as a range: a “Name-ability Spectrum.”

Three industries sit at three different points on exactly that Name-ability Spectrum axis.

Legal is the lighthouse shining brightly in the open. Firms will go on the record as a reference, because using the most sophisticated tool is itself a status signal. Harvey's marquee wins get talked about precisely because the buyers want to be seen making them. The light shines out for everyone.

Pharma is the lighthouse behind frosted glass. Nobody announces early. Often, nobody announces ever. Proof moves first through gated channels: key opinion leaders (KOLs), scientific sessions, published validation, and people spreading the word as they move between companies. Then the moment one major is known to be in, the herd moves fast, because they're all equally risk-averse. They all want the choice that someone else already stuck their neck out and de-risked. In this bucket, proof travel is delayed, and then it cascades.

Oil and gas is one of the hardest sectors, the light bulb is 5W and barely there. These operators are so risk-averse and so disclosure-averse that even after they adopt, they often won't let you name them at all. The proof exists yet still can't advance, because the reference itself is boarded up. The same DNA often runs through aerospace and defense, banking and insurance, and utilities. Concentrated markets, high stakes, everyone watching everyone, and a buyer who refuses to go first or reveal their hand.

Here's the rule that falls out of the Name-ability Spectrum. The less nameable your proof, the more the communications craft is worth. When a company can't secure a named, on-the-record customer reference, the job becomes manufacturing proof, which can still travel. Here manufacturing proof carries its weight. This proof will need to come in a different form: an anonymized but verifiable “top-five major” case study, analyst validation, a safety or regulatory milestone told as a story, executive vision, and published data. That's exactly the work a founder can't do from inside a sales deck.

How to turn the light bulb on

Turning that light bulb on lands with Marketing's function and namely PR. It's not something a GEO dashboard can do, because a GEO scoreboard reports your number. It can't zero in on a company's core distinctions, craft the differentiated messaging, shape a founder's vision into a compelling narrative that buyers repeat, earn the news coverage, or fix how AI engines describe your brand. Optimization was never built to do that.

This isn't a hunch. Muck Rack's Generative Pulse, an analysis of more than 25 million citations across ChatGPT, Claude, and Gemini, found that 84% of AI citations come from earned and reputational sources like journalism, research, and third-party coverage. Paid and sponsored content accounted for 0.3%. The answer a buyer reads is indeed built almost entirely from what other people say about you, not from the pages you write and own.

A logo sitting inside a private sales deck doesn't travel beyond the file where it's stored. Turning one buyer's decision into market-wide reassurance takes deliberate work across multiple channels. Together they are what makes proof move.

Here's four key functions that show how PR makes this work.

2026 rewrites the last mile

Proof used to travel through analyst reports and quiet conversations between two CIOs. It still does. But increasingly it travels when ChatGPT, Perplexity, Gemini, and Claude repeat it in the answer that a prospective buyer reads before they ever call you. And that's assuming you show up in their query.

Note, if the AI engines don't have your proof to cite, it doesn't matter how strong that proof is. It won't surface where the buyer journey and their decision now starts. Making a company legible to those AI engines, so the right proof shows up at the right moment, is the discipline we call Machine Relations. It's the difference between a proof point that exists in a deck and a proof point that arrives in a citation.

The part worth adding

a16z is right. Your buyer buys proof or math. Here's the part worth adding to their map. Whether your proof travels isn't decided the day you win the customer. It's decided afterward, in whether anyone, human or AI engine, is ever allowed to see what you did.

A lighthouse only works if someone can look at it. PR is what makes proof travel.

Get Found. Get Cited. Get Chosen.

Sources: a16z, "Lighthouse or Landgrab? How to Pick Your AI Sales Strategy" (Joe Schmidt IV and Julian Marx); Muck Rack Generative Pulse (May 2026).