The Two Ways to Sell AI: Lighthouse or Landgrab?

Joe Schmidt developed the lighthouse versus land grab framework after observing competing AI companies on the 101 freeway selling identical software to the same San Francisco market. The framework maps go-to-market strategy on a two-by-two matrix with buyer exposure on the Y-axis and whether proof travels in a market on the X-axis. The top-right lighthouse quadrant suits regulated industries with a limited number of logos where buying the wrong software creates legal or reputational risk, and the core mechanism is proof. The bottom-left land grab quadrant suits markets where buyers already have established budgets and the seller can demonstrate through math that their solution outperforms the incumbent. Elena Berger describes the choice between these two strategies as the single most expensive question an AI founder makes.

Harvey and Hebbia are classic lighthouse examples. Harvey targeted select critical law firm accounts to establish that AI augmenting junior lawyer workflows was safe, and winning those accounts caused proof to spread rapidly among buyers with high reputational risk exposure. Stutt, targeting the accounts receivable market, is a land grab example, showing mid-market buyers mathematical proof that AI-augmented collections would outperform existing human teams and software on working capital, cost savings, and revenue. Pylon, an AI-native customer support company, is also cited as a land grab example, climbing the ACV ladder by replacing incumbents starting from modest deal sizes. A practical test for which quadrant a company occupies is whether prospects will get on the phone, move through a proof of concept, and buy without extensive education.

Samsara, founded in 2015, did not strategically choose between the two approaches. Cold calls to the largest trucking firms produced immediate rejections, pushing Samsara toward the mid-market, where customers required less social proof, had shorter sales cycles, and provided faster product feedback. The U.S. ELD mandate, implemented in phases between 2016 and 2019, forced the entire trucking industry to find budget for compliant devices simultaneously, creating a tailwind that benefited new entrants even against AT&T and Verizon, which were already in the hundreds of millions to half a billion in revenue. Andy McCall draws a parallel to the current moment, noting that enterprise AI boards are mandating AI adoption by specific deadlines, creating a similar wave of kinetic energy inside large companies.

Meraki, founded in 2006 and acquired by Cisco in 2012, pursued a land grab targeting the mid-market after finding that municipal Wi-Fi was not a viable business. Because Cisco and HP had the largest enterprises locked up, Meraki focused on mid-market customers who lacked large IT teams trained in command-line configuration and valued simpler cloud management. To drive adoption, Meraki offered a free access point to webinar attendees so prospects could experience the product's simplicity firsthand. Giving away hardware hurt gross margin but was judged worthwhile for customer acquisition.

AI proof of concepts carry a specific risk of becoming indefinite science projects because capabilities are advancing daily and vendors keep demonstrating new features on demand. Best practice is to set a hard end date of 30, 45, or 60 days and define success criteria upfront before the trial begins. Decagon is cited as a strong example of managing this by committing to specific benchmarks and hitting them within the trial period. Further AI, selling into insurance with a governance-first positioning, uses forward-deployed teams working alongside customer staff to get the product operational, and its wins with large incumbent insurers generate social proof that travels down the long tail of the insurance market.

Schmidt and McCall recommend that founders pursue whichever strategy produces the earliest and easiest sales at their current stage and then layer in the other approach later. Both Meraki and Samsara began with land grab strategies and later verticalized into lighthouse accounts, identifying the top five accounts in segments such as transportation, warehousing, and public sector. Meraki's earliest lighthouse vertical was school districts because districts within a state communicate heavily with each other about purchasing decisions, and winning the largest district in a state tends to pull smaller districts along. Schmidt notes that too few founders are willing to pick up the phone, get on a plane, or travel outside major metros, and that there is a meaningful opportunity right now to go sell large software deals in markets that are being overlooked.

Andy argues that the dominance of product-led growth over the prior 10 to 15 years was driven by the need to wedge into enterprises already locked into major cloud platforms for CRM, HR, ITSM, and security, most of which were founded between 2000 and 2010. He frames the current AI moment as a non-skeuomorphic shift where agents replace rote human work, creating an opportunity to sell large platform software again rather than incremental replacements. Andy also identifies sales operations as a role companies hire for too late, arguing that a single dedicated person thinking daily about territory alignment, name lists, commission structures, and sales constitution is sufficient early on, and that gaps in this infrastructure become speed bumps once a company enters scale mode. He views quota attainment rates of only 40 to 50 percent across a sales team as a sign that quotas are set too high or the wrong profiles are being hired, and considers this a disservice to the organization.