Two questions tell you which way to sell AI
The person signing the deal is weighing personal risk. Whether industry chatter reaches them decides the rest.
- Chasing a marquee customer could cost you the whole market.
- Two questions decide your path: how much risk the signer carries, and whether reputation travels in the industry. When the answers clash, risk wins.
- Regulated industry doesn't automatically mean Lighthouse. Samsara, in a mandate-driven trucking sector, picked Landgrab.
- Plus practical rules you can use now, from framing a POC to setting quota targets for your early team.
Chasing marquee names costs you the market
In late July, a16z partner Joe Schmidt published an article on how AI companies should sell. Two weeks later, he and Andy McCall recorded a podcast of the same name, unpacking the framework step by step.
Andy's background is worth pausing on. From 2017 to 2023 he was CRO at Samsara, taking revenue from single-digit millions to $1B ARR and an IPO in 2021. Earlier, as a VP at Meraki and Cisco, he scaled Meraki from the same single-digit millions to hundreds of millions. So you have the author of the framework and someone who's actually fought on the front lines.
The problem this framework solves can be stated in one sentence: The marquee customer you're chasing is losing you the market.
Here's how it plays out. A founder with a great product spends months chasing a first Fortune 100 customer. They burn through a funding round. Their team bleeds time on deals that never close. It's not for the revenue, which is negligible. They want a recognizable name on the sales deck, betting it will make every future deal easier. So they offer steep discounts, sometimes even paying the customer to use the product. Meanwhile, the buyers who actually need the software and would pay full price have never heard of the company.
The framework starts with an observation from a drive. Joe was on Highway 101 and noticed two competing companies, one on each side of the road, selling identical software. Both were convinced the only customers worth selling to were along that route in San Francisco. Now, it's even more extreme: buses wrapped in ads, planes towing banners. All clever ideas, but they're targeting the same sales motion.
When it comes to selling AI to enterprises, there are really only two paths. One is Lighthouse: focus all your energy on winning a few high-profile customers that everyone recognizes, and use their endorsement to pull the rest of the industry along. The other is Landgrab: skip the big names, give your prospects a clear-cut business case, sign as many as you can, and win on coverage.
Both paths have proven themselves, and picking wrong leads to the money-burning name-chase we started with. So the real question is just one: which path does your business belong to? Asking which path is better is a false question. Let's first flesh out what each path looks like, then get to the decision method.
How Lighthouse and Landgrab actually work
Let's get the terms straight. Lighthouse means focusing your firepower to win a few renowned reference customers and using their endorsement to validate the whole industry. Landgrab means skipping the big names, winning deals with a clear-cut business case, and signing as many regular customers as possible, as fast as possible.
A lighthouse doesn't carry cargo, but when it shines, the ships behind it know the channel is safe. Landgrab is different: whoever claims the land first owns it. By the time your competitors react, there's no empty ground left to plant a flag.
When should you go Lighthouse? When you're doing category creation. AI is making possible what wasn't possible before. There's no precedent in the buyer's organization, no entrenched incumbent to replace, and no mental model to fall back on. You're inventing something new, not swapping out an existing piece of software. This means you're asking the buyer to take a leap of faith.
That's how Harvey started. Law firms used to buy research tools from Thomson Reuters and LexisNexis, which surface information for paralegals to interpret. Harvey does the work itself: drafting, research, due diligence across thousands of documents. Lawyers are trained to avoid risk, and no one wants to be first. At the end of 2022, Allen & Overy signed up. Paul Weiss followed in early 2023, and the whole peer group took notice. Today it has hundreds of millions in ARR and an $11B valuation. Hebbia, on the finance side, took the same path: client teams spent 60+ hours a week digging through high-stakes data rooms, where confidentiality and reputation rule and no one wants to move first. Hebbia broke through with the largest private equity, hedge fund, and consulting firms, and later landed more than 40% of top asset managers by AUM.
Landgrab is the opposite: when the buyer already knows the problem, and making a mistake won't cost them their job. The pitch is simple: I can replace what you're using with better results, or at a lower cost. You don't need the CTO of Stripe to vouch for you to get a meeting. You get the meeting by showing the VP of Customer Support what they're spending today, and telling them you can cut that in half.
Category creation, no existing mental model for the buyer
>$100K ACV (annual contract value), often $1M+ in early days
3-6 months or longer
Small team, founder-led; the team closing deals is often the team delivering them. Expensive and unscalable by design.
Buyer understands the problem; errors won't end careers
Much smaller, banking on volume
Short, emphasis on rapid deployment
Demo-driven, larger team; product must be so standardized that customers can get going quickly, supported by a specialized pre-sales deployment team