Lighthouse vs Landgrab

Two playbooks

Joe Schmidt IV's framework splits enterprise AI go-to-market into two modes, set by what is being sold and who is buying it. The question it answers is one of market structure rather than of timing: which motion a given market permits at the moment of entry, not how long a company should stay in the one it chose. Lighthouse fits category creation, work AI newly enables where no incumbent exists to displace and the buyer has no existing mental model, so a few credible peer adopters function as the product's real evidence. A legal AI company needed two marquee law firm clients before risk-trained lawyers would believe the category was real; another broke into finance through the largest private equity firms and hedge funds before expanding across asset managers. The mode is founder-led and high-touch, six to seven figure contracts, cycles running three to six months or longer with proof-of-concept work, deliberately expensive and unscalable.¹

Landgrab fits markets where the buyer already understands the problem and a wrong choice is recoverable: win on math, move fast, sign the largest number of customers possible. The pitch is a direct cost or outcome comparison against an incumbent, and speed matters because the incumbent already owns the customer relationship and adds its own AI every quarter. Alex Rampell's framing of the underlying race applies directly here: get distribution before the incumbent gets innovation. One accounts-receivable automation company deployed in under a week against traditional rollouts running six to eighteen months; another went from zero to eight figures in annual revenue in eighteen months on rapid deployment and immediate return on investment, implemented through the kind of Forward-Deployed Engineer Model built for volume. The mode is demo-driven and standardized, and it only works if the unit economics hold at scale, the landgrab complement to Selling Outcomes Not Tools.¹

Because the two modes are treated here as a choice, the sequencing case sits outside this file: a company that won a category on a handful of marquee customers eventually earns the right to switch to volume, and the question of when that right has been earned and which vertical to point it at next, worked through Affirm's move from Casper to every other mattress company and then to exercise equipment, is treated in the file on lighthouse to landgrab sequencing.¹

The map: two questions

Two questions locate a given market on this grid. How exposed is the buyer who signs: exposure climbs with regulation, with replacing a system of record rather than adding a tool alongside one, and with output that faces the outside world rather than an internal draft someone reviews. For a highly exposed buyer, the return-on-investment math becomes beside the point. And does social proof travel: in law and financial services, firms watch each other closely and status is legible, so two marquee adopters have effectively completed the risk assessment for everyone behind them, while in fragmented markets buyers do not watch each other and each sale starts from zero. High exposure paired with proof that travels calls for lighthouse; recoverable mistakes paired with proof that does not travel calls for landgrab, where Velocity as Design Principle and Time Favors the Fast become the whole strategy. The common mistake in 2026 runs the wrong direction by default, assuming buyers need lighthouse-style proof simply because the technology is AI and feels unfamiliar, then showing up with a marquee logo to a buyer who only asked what it costs.¹

The traps

Lighthouse fails by becoming hostage to the logo, since the same few hundred marquee accounts know they are being pursued and can extract concessions, or by winning prestige without payback, a reference customer that will not pay recurring, real contract values, or collaborate on repeatable software. It also fails through pilot purgatory, proof-of-concept work that was never going to convert, and through over-fitting the product to a single marquee customer so no other business ever follows the same ship. Landgrab fails by dying of indigestion, growing to hundreds of customers without qualification discipline until a large share sit underwater, or by grabbing land the product cannot hold, scaling coverage faster than the product is ready and creating detractors at volume.¹

Open question

The two axes, buyer exposure and whether proof travels, tend to correlate in practice, since regulated industries are usually also the ones where status is legible, which may be why the framework has little to say about the markets that mix high exposure with proof that does not travel: those are simply hard markets, and the companies stuck there are not the ones anyone has heard of.