Key Takeaways

  • Andreessen Horowitz separates enterprise AI sales into lighthouse and landgrab motions based on buyer exposure and the portability of customer proof.
  • Legal and financial research applications tend to require credible reference customers, while customer support and accounts-receivable automation can favor rapid coverage.
  • Governance evidence, measurable economics, and disciplined customer selection can determine whether either approach scales.

Andreessen Horowitz has set out a two-track model for selling enterprise AI, arguing that go-to-market strategy should reflect the risk buyers assume rather than treating artificial intelligence as one uniform market. Its "lighthouse vs. landgrab" thesis divides AI applications according to how much confidence customers require and how readily evidence from one deployment transfers to another.

The distinction arrives as AI moves from experimentation into established corporate budgets. A McKinsey survey published in 2024 found that 65% of organizations were regularly using generative AI, up from 33% in 2023. That breadth changes the sales question. Vendors are no longer selling only the idea of AI; they are selling a particular operational outcome, risk profile, and path to deployment.

Under the lighthouse model, a vendor concentrates resources on a small number of credible customers whose adoption can reassure the rest of the market. Legal work and financial research fit this pattern because errors may affect client advice, investment decisions, confidential information, or regulatory exposure. Buyers in these areas are likely to scrutinize accuracy, data handling, auditability, and human oversight before focusing heavily on price.

A recognizable customer logo helps, but the underlying proof matters more. Prospects may want evaluation results, escalation procedures, security documentation, model-monitoring practices, and evidence that employees can review or override outputs. One successful deployment can reduce perceived risk elsewhere, particularly when the reference customer operates under comparable obligations.

A lighthouse account can also consume enormous attention. Product changes, custom integrations, procurement reviews, and executive involvement can turn one customer into a lengthy engineering program. The logo has value only if the resulting capabilities and evidence can be reused. Otherwise, the vendor may have landed a prestigious account without creating a repeatable sales motion.

The landgrab approach looks different. Customer support AI and accounts-receivable automation often have clearer unit economics and a broader pool of potential buyers. A seller can emphasize ticket deflection, handling time, collection costs, payment speed, or staff capacity. Faster outreach and standardized onboarding may outperform a highly selective pursuit of marquee brands.

That does not make these deployments risk-free. Customer support systems can still mishandle personal information or produce inappropriate responses, while AR automation can affect customer relationships and financial controls. Yet the downside can often be bounded through approval thresholds, workflow constraints, limited permissions, and phased rollout. The sales conversation therefore tends to move toward measurable economics sooner.

Governance can shorten the distance between those two motions. The NIST AI Risk Management Framework organizes AI risk work around govern, map, measure, and manage functions. For vendors pursuing lighthouse customers, evidence mapped to those functions can make safety claims more concrete. It gives procurement, compliance, and technical teams a shared vocabulary instead of leaving them to interpret broad assurances.

Likewise, ISO/IEC 42001:2023 provides an AI management system standard covering organizational controls for developing, providing, or using AI. Certification or alignment does not remove deployment risk, but it can offer enterprise buyers a recognizable governance signal. In some sales cycles, that may reduce repetitive diligence and help proof travel between customers.

So which motion should an AI vendor choose? The answer can vary even within one product. A company might use lighthouse selling for a sensitive decision-support module while applying landgrab tactics to a lower-risk workflow feature. Geography, industry regulation, data sensitivity, and the cost of a wrong output can shift the balance.

For enterprise leaders, the practical lesson is to design sales capacity around the buyer's exposure. Lighthouse motions tend to reward concentrated account selection, technical validation, and reference development. Landgrab motions tend to reward segmentation, fast qualification, standardized implementation, and clear return-on-investment calculations. Andreessen Horowitz's central contribution is a sharper way to decide between them before hiring a large sales team or committing scarce engineering resources.