Are Lighthouse’s customers still capturing the market? How should AI companies choose their sales strategies

Authors: Joe Schmidt, Julian Marx, a16z

Translated by: Omnitools

That famous client you’ve been chasing might be making you lose the whole market.

Many startups with excellent AI products spend months trying to win their first Fortune 100 customer: the funding gets burned quickly, and the core team gets stuck in deals that never convert. Founders pursue these customers not necessarily because the contract value is high, but because they believe that if a known brand appears on the sales materials, every subsequent transaction will become easier.

So they pour nearly all their effort into these companies—even with steep discounts, or by paying customers to use the product. Whether revenue can cover costs doesn’t matter much; the goal is to get the name. Meanwhile, buyers who genuinely need the product and are willing to pay full price may never have heard of the company.

This choice is not without logic. AI is still a new technology, and founders often think they must first educate the market—so they default to a “lighthouse customer” strategy: win a small number of top customers, build social proof, and reassure buyers who fear making the wrong decision.

But for many AI companies, that instinct may be backwards. Their buyers already understand the problem, and they’re not worried that one procurement mistake will derail their careers—they just need to confirm whether the numbers make sense. Each extra week a startup spends chasing a marquee customer gives competitors another week to close deals in a larger market. For these companies, a “market capture” strategy may be more appropriate: win customers through a clear investment-to-return case, move quickly, and expand the number of signed deals as much as possible—without putting blind faith in customer fame.

When selling AI products to enterprises, there are broadly two go-to-market playbooks: lighthouse customers and market capture. The difference isn’t product quality or team strength—it’s what you’re selling and who you’re selling to.

Lighthouse customers: Prove that a new category exists with leading buyers

When AI makes work that used to be impossible achievable, companies are essentially creating a new category. These products have no precedent inside the customer organization; there’s no clear existing product that can be directly substituted, and buyers also lack an off-the-shelf mental framework to plug into. You’re not replacing Salesforce—you’re inventing a brand-new way of working, which means the buyer has to take a risky step.

In this kind of market, social proof is critical. Startups need to find “lighthouse customers”: their adoption behavior is not just a single order, but a signal to the entire industry that the category is real and that this company is worth betting on.

Harvey is a classic case. It provides AI services for legal professionals. Today, legal AI has become a mature category—but a few years ago, there weren’t many players in the market, and there were almost no customers either. Law firms used research tools from Thomson Reuters and LexisNexis: they would use the tools to find information, then hand it to legal assistants to interpret and process it. Harvey goes further, taking on drafting, research, and conducting due diligence across thousands of documents.

Legal work naturally emphasizes risk avoidance, so no firms wanted to be the first to try. Until Allen & Overy signed at the end of 2022, Paul Weiss followed in early 2023, and then other peers started paying serious attention. Today, Harvey’s annual recurring revenue is in the hundreds of millions, and its valuation is $11 billion—but before it reached that point, it first needed these two leading law firms to prove to the entire industry that parts of legal work can be done with AI.

Hebbia used a similar strategy in financial services. It builds an AI intelligence platform for teams that spend 60 hours per week or more handling high-risk data room materials. In transaction environments that are highly confidential and extremely reputation-sensitive, no funds were willing to be the first to take the risk either. Hebbia first won global large private equity firms, hedge funds, and consulting companies, and then expanded to more than 40% of top asset managers ranked by assets under management, including KKR and BlackRock. Like Harvey, once well-known institutions made the first move, the market began to follow.

Winning lighthouse customers typically relies on small teams, direct founder involvement, and high-touch service. In the right markets, this playbook can generate large orders: early annual contract values are at least six figures of dollars, and often reach into seven figures. Because it involves proof of concept, custom development, and convincing buyers who fear costly mistakes, the sales cycle can last 3 to 6 months or longer.

The teams that land the orders are often also the teams that deliver the product. This approach is expensive and hard to scale—but it’s designed precisely to build a new category. For early buyers, being first can also bring outsize returns. That means great lighthouse selling needs to make the customer feel that this “risk” is more like getting a competitive advantage early, not like taking a chance at making an error.

Market capture: Use the economics and speed to broaden coverage

If the buyer already understands the problem, and a single procurement mistake won’t cost decision-makers their jobs, the rules are completely different.

These buyers know what you’re selling, and your sales pitch can be very direct: “I can replace the current solution at lower cost, or deliver better results.” You don’t need a star company’s technical leader to endorse the product. Just lay out the current costs for the customer support leader, then tell them, “We can cut that expense in half,” and that’s enough to earn a meeting.

Social proof can still speed up deal-making, but the market doesn’t need to be persuaded that the problem exists. In this kind of market, sacrificing speed is often fatal, because startups aren’t only competing with other startups—they’re also up against traditional vendors adding AI capabilities to their existing products.

For example, Zendesk adds AI assistants, and Decagon uses agents to replace most customer support functions—there are fundamental differences in product form. But Zendesk already has customers. The more AI capability it adds each quarter—whether through in-house development or acquisitions—the harder it becomes for startups to win customers away from it. Therefore, speed itself is the core competitive advantage. As a16z partner Alex Rampell put it: companies must secure distribution and customer coverage before traditional vendors catch up with innovation.

Stuut provides a receivables automation example. Its product covers collections, payment processing, cash application, and dispute handling. Enterprise resource planning suites from SAP and Oracle already include receivables modules, and companies like HighRadius have sold related point solutions since the early 2000s—but enterprise teams still waste huge amounts of time chasing invoices.

Stuut’s outcomes are very clear: customer cash flow increases by 40%, manual tasks decrease by 70%, and the average days to collect drops by 37%. The company chose broad customer acquisition early on, prioritizing the mid-market and lower-end enterprise segments instead of chasing Fortune 100 brands. Today, it serves manufacturers, distributors, and logistics companies across places like Michigan, Ohio, and Texas. Traditional solutions often take 6 to 18 months to deploy, while Stuut can go live in a week.

Decagon uses a similar playbook in customer support. Before developing the product, the founders interviewed roughly 100 customers within a month, then positioned it around fast deployment and immediate ROI. Within 18 months, the company’s annual recurring revenue grew from zero to eight figures. In 2025 alone, Decagon signed more than 100 new enterprise customers across travel, financial services, healthcare, and retail; in less than 6 months, its valuation tripled to $4.5 billion.

Market capture sales driven by demos typically require a larger team. The product must be standardized enough so customers can get onboard quickly and see value immediately. Implementation work should be handled by front-line delivery teams that are good at delivering—not by restarting discovery with each customer. Since scale is the strategy itself, the unit economics must support large volumes of customers.

How to tell which strategy you’re in

Enterprise selling is fundamentally a tradeoff between risk and reward. Behind every deal is someone who has to sign off on the procurement decision. They want the approved product to work—and they want to still have their job next year.

Buyers aren’t evaluating the product in the abstract; they’re assessing: how much personal risk will this purchase create for them, and what evidence is needed to make that risk acceptable.

So startups can choose between the two strategies using two questions.

First, how much risk does the signing buyer bear

The cost of different mistakes isn’t the same. In customer support or receivables automation, a wrong reply or an invoice with an incorrect amount annoys the customer, but it can usually be corrected. A decision-maker might experience a bad quarter as a result, but it doesn’t necessarily end their career.

Buyers’ risk exposure is mainly influenced by three factors:

2. Is the industry tightly regulated, such that a vendor’s error becomes the buyer’s compliance responsibility?
3. Is the product replacing an enterprise’s core record system, or is it only an add-on tool?
4. Does the product output go directly to the outside world—for example, formally submitted documents or answers sent directly to customers—rather than internal drafts that still require human review?

In legal and financial services, these factors are often all high. A single piece of fabricated data can lead to incorrect asset pricing, or even cause the entire transaction to fail. For these buyers, calculating ROI isn’t the point; what they need to control is personal risk that discounts can’t offset.

Second, can social proof spread in the market

Some markets have high reputational diffusion efficiency, and others don’t.

Financial and legal institutions closely watch peers, and industry standing is usually easy to recognize. Winning two top institutions can influence the entire market, because first adopters force a risk assessment that latecomers can benefit from.

But in fragmented mid-market receivables markets, a finance leader in one region may not care whether a famous brand uses your product—or may never even come across that information. There’s little tight observation among buyers, so each sale has to re-prove the value from scratch, and the help from top-customer names is much more limited.

In other words: centralized, emphasis-on-industry-stance markets can transmit social proof; highly fragmented markets require sellers to re-calculate the economics in every deal.

Put the two questions together, and you get a strategy map:

• When buyer risk is high and social proof spreads quickly, it’s a lighthouse customer market. A small number of credible buyers adopting first can open up the whole market—legal work and financial research are examples.
• When errors are easy to fix and social proof doesn’t spread well, it’s a market capture market. Clear economic logic drives deals; broader customer coverage wins the market—customer support and receivables automation are representative examples.
• When social proof spreads quickly but buyers don’t rely on proof, the product can spread through user-driven adoption. For example, developer tools can spread among engineers, and a category can form bottom-up; in such cases, you may not need a huge sales team from the start.
• When buyers need proof but the influence of top customers can’t reach other buyers, this is the hardest market. Companies stuck in this quadrant also tend to be hard for the outside world to see.

A few surface-level heuristics can help too: smaller businesses tend to lean toward market capture and dislike risk; risk-averse industries tend to lean toward lighthouse customers; and add-on tools often sell faster than products that replace core record systems. But if you trace these rules back, they always come back to the two variables: buyer risk and whether proof can spread.

Markets won’t always be clearly classifiable. For ambiguous cases, you can also look at existing budgets and rough ROI calculations. But only when buyer risk is already controllable can these signals confirm that the company is in a market capture strategy. Buyers may have budgets and can understand the economics, yet still wait for a trusted institution to adopt first. When risk and economics point in different directions, prioritize risk.

The sales cycle can also serve as a quick check. If the deal lasts longer than 60 days, if the team needs extensive customization to prove the concept, and if the buyer asks “Is it safe?” rather than “How much does it cost?”, then they need social proof and the company should use a lighthouse customer strategy.

Right now, the most common mistake is actually the opposite: because AI is new, assume every buyer needs to be educated and needs top-customer endorsement. But if the buyer’s worst case is merely an incorrectly billed invoice that can later be corrected, then they aren’t managing career risk. In that situation, showing up with a list of famous customers is essentially answering a question the buyer never asked.

Traps that both strategies are most likely to fall into

Four traps of the lighthouse customer strategy

First, being hijacked by famous customers. Every startup is competing for the same top brands, leading to brutal competition that ultimately revolves around roughly 500 large customers. Big customers know startups are under pressure to close, so they keep demanding concessions. Lighthouse customers are a tool to open the market—not the end goal. After winning a few trusted brands, the company needs to move quickly into a broader market, because the vast majority of revenue still comes from companies that aren’t well known to the public.

Second, only fame, no returns. The wrong customers may refuse to help build replicable software, slowly turning the startup into a consulting firm; or they may refuse to keep paying, causing the business model to become unsustainable; or the contract value may be too low, making unit economics fail. Worst case: the company serves a famous brand, but learns no replicable playbook, and revenue isn’t enough to cover the investment—leaving only a vanity metric.

Third, getting stuck in a pilot purgatory. Large companies love pilots. A proof of concept that lasts 6 months but never converts into a formal contract can burn through a startup’s best talent. The fix is to set a clear duration for the pilot, define milestones, and include in the contract an automatic conversion to a formal engagement once conditions are met.

Fourth, illuminating only one ship. The team over-responds to a lighthouse customer’s bespoke needs, ultimately producing a product that fits only that customer and has no value for other customers. The company may indeed win the lighthouse—but other ships in the market don’t move forward along that beam of light.

Three traps of the market capture strategy

First, dying from indigestion. When a product can be sold to almost anyone, the real discipline is learning how to say no. Some customers are hard to onboard and hard to produce results for, yet the contract values are low. Without strict customer filtering and transaction approval mechanisms, a company might wake up with 200 customers, 50 of which are losing money.

Second, taking over a market you can’t defend. If you expand your coverage before the product is mature, you’ll produce dissatisfied customers in batches. 50 unhappy customers means churn in renewal revenue; 500 unhappy customers can become a reputational crisis.

Third, mistaking a local market for the whole market. Running through every bus stop in a big city doesn’t mean you’ve truly won market capture. The bigger opportunity is to find a way to show a clear investment return to 50k companies outside of traditional relationship networks.

Best sequence: move from lighthouse customers to market capture

The best companies don’t stay forever in one mode. They intentionally transition from lighthouse customers to market capture: win benchmark customers in a vertical first, dominate that vertical, then look for adjacent industries with similar characteristics.

That’s exactly how Affirm’s path unfolded. Its early breakthrough came from mattress brand Casper. After winning one mattress company, Affirm continued to pursue nearly all mattress companies, then moved into fitness equipment, and later expanded to products that looked similar to fitness equipment but actually belonged to other categories.

Mattresses and Peloton bikes don’t have much in common on the surface—but they’re both high-priced consumer goods that customers want to pay for in installments. Affirm saw that earlier than the rest of the market.

Once lighthouse customers define and validate a new category, that category eventually becomes a recognized market. But the company must first earn the qualification to transition from lighthouse wins to scaling. Fully rolling out before the category exists consumes cash and credibility at the same time.

A clear conversion signal is: buyers start proactively coming to you with pre-allocated budgets, asking to see a product demo—not repeatedly asking “Who else is already using it?” If that happens, it indicates that lighthouse customers have done their job and the market capture phase has arrived.

Conclusion

Every AI founder believes they’re inventing the future—and they may indeed be doing so. But buyers don’t buy the abstract concept of “the future.” They buy proof, or a piece of economics that makes sense.

If buyers need proof, then go win the top customers who can provide it. If buyers need economics, put the investment-to-return ratio in front of them as soon as possible and get the contract signed before competitors or traditional vendors.

Founders who pick the wrong strategy might not fail because the product is wrong or because they selected the wrong option from a strategy checklist. They’re more likely to fail because they never truly asked: which game are they actually playing?

In a market that changes this fast, companies may only get one chance to answer that question.