Palantir Sells AI by Sending Engineers, Not Salespeople. Why?
Deep thoughts on AI and aspirations —— ByteDance Deep Thinking Circle
For most people, enterprise software sales conjures the same picture: build a sales team, cultivate client relationships, pass the POC, sign the contract. Palantir is an exception. This company, which at one point approached a $500 billion valuation, has barely built a traditional sales army. Instead, it sends engineers to the front lines. In 2025, the total value of its new contracts exceeded $10 billion (company disclosure).
In an enterprise software industry where customer acquisition costs climb year after year, this playbook deserves closer examination. What it sells is no longer software, but a sales machine built on live demonstrations.
Enterprise AI Sales: The Trust Bottleneck
Buyers of enterprise AI have already heard about large language models and agents. Information isn’t what’s missing. What’s missing is trust, for two reasons.
First, it’s hard to imagine what an AI product can do before using it. What can a large language model do for a factory’s supply chain? No amount of explanation creates conviction—the client needs to see it run on their own data. Second, enterprise software purchase decisions have traditionally relied on two forms of trust: relationships built by salespeople, and social proof from peer adoption. Both are slow variables. Relationships take time to cultivate, case studies take time to accumulate, and in the AI era where product forms change every six months, yesterday’s case study may not explain today’s product.
Palantir’s approach replaces these two slow variables with one fast variable: send engineers directly to the client site, use the client’s real data, and within days solve a problem the client is demonstrably struggling with.
This approach has a straightforward internal name: Bootcamp. Clients bring real data, engineers help them build a working prototype on the platform in a few days, and decision-makers see results on the spot. Public data tells the story: in 2022, Palantir ran 92 pilots all year; after scaling this approach in 2023, they held over 500 sessions in one year, covering more than 400 organizations. The typical six-month enterprise software sales cycle compressed to around three months.
The mechanism isn’t complicated. Persuasion builds trust through language; demonstration builds it through results. The former takes time, the latter takes engineering capability. In an era where AI capability itself is rapidly appreciating, the latter is getting cheaper.
A Three-Stage Machine: Get In Cheap, Expand Deep, Exit Clean
Replacing persuasion with demonstration is just a sales tactic. What’s really worth unpacking is the complete growth model behind it. I break it down into three stages: entry, expansion, and exit.
In the entry stage, target problems that would be catastrophic if left unsolved: supply chain disruption, fraud risk, resource allocation during a pandemic. Then enter at extremely low prices or even free. The UK NHS is the most typical example. At the start of the 2020 pandemic, the British healthcare system faced challenges with resource allocation and vaccine rollout planning. Palantir signed on for a symbolic £1, and within days, engineers integrated thousands of disparate medical databases and built a tracking dashboard. That one pound bought something very specific: the fastest ticket into a real, high-value scenario.
The expansion stage comes after solving one pain point, then spreading outward along the business chain. This relies on a mechanism called the Ontology: translating business concepts like parts, suppliers, aircraft, and routes into a unified data structure that AI and systems can operate on. Airbus is a typical case. It started with improving efficiency on the A350 production line, increasing A350 delivery speed by 33%, then expanded to supply chain, quality control, and flight data. Eventually, thousands of Airbus suppliers and hundreds of airlines worldwide connected to the same platform. In 2024, the two parties signed a roughly $1 billion ten-year contract.
Why does this expansion create strong lock-in? Because the essence of the Ontology is translating the client’s business language into your system’s language. The deeper the translation, the higher the switching cost. Changing vendors means retranslating the entire business from scratch.
The exit stage comes after the platform matures: the client’s own employees can use it, engineers withdraw from the front lines, marginal costs plummet, and gross margins explode. BP has been using it for over a decade. Tens of thousands of employees run on the platform making decisions, Palantir invests less and less manpower, yet revenue climbs year after year.
The financial results of this machine are quite transparent. After 19 consecutive years of losses, the company turned its first annual profit in 2023. Net profit more than doubled in 2024 and climbed another level in 2025. Net retention rate reached 139%, and the industry-standard Rule of 40 (growth rate plus profit margin) exceeded 100. All figures are from company disclosures or public reports.
But This Machine Has Three Prerequisites
A dose of cold water is needed. This model looks elegant, but it’s not universal. It rests on three prerequisites.
First, you need surplus engineering capability. Entering for free means you can actually solve the problem with engineers in a few days. Without the capability, free demonstrations become paying to do outsourcing for clients—worse, publicly proving you can’t deliver. Palantir spent over a decade in high-difficulty scenarios like intelligence and defense to build this surplus. Most startups don’t have it.
Second, results must be verifiable in the client’s own business language. Demonstrations work because results are concrete: 33% improvement in delivery speed, thousands of databases integrated, a prototype that can go directly to the CEO. If your product value can only be described as “good atmosphere” or “smooth experience,” no amount of flashy demonstration produces persuasive results, and the first stage of the machine won’t turn.
Third, you need an exit mechanism. This is the dividing line between service companies and platform companies. If every client requires engineers on-site, no matter how big the business grows, it’s just high-end consulting. Labor costs cap gross margins. Whether delivery experience can be consolidated into a platform that clients can use themselves determines whether you can reach the third stage.
Beyond these three prerequisites, two facts must be acknowledged. First, part of Palantir’s success is unreplicable: decades of accumulated security clearances, government relationships, geopolitical positioning, and brand narrative built around that positioning. This anti-Silicon Valley posture is medicine for Palantir; for ordinary companies, it’s usually poison. Second, some circulating numbers are suspect. For example, the claim that bootcamp conversion reaches 92% at its highest—that’s the best-case scenario under a specific definition. It’s safer to treat it as a ceiling than an average.
Implications for Action
For companies building enterprise AI, three self-examination questions suffice.
First, in your product, is there a problem that will cause disaster if the client doesn’t solve it? If not, go back to finding scenarios. Don’t rush to copy the free demonstration model.
Second, given five days and the client’s real data, can you produce a result the client’s CEO can understand? If yes, move demonstration to the very front of the sales process and compress the cycle. If no, solidify the product first. Don’t learn to give things away.
Third, does each delivery consolidate something reusable: templates, ontologies, industry components? If yes, you’re moving toward a platform. If no, you’re moving toward consulting, and pricing should reflect that.
For buyers, the judgment criteria are simpler. When selecting an AI vendor, see if they dare to run live demonstrations with your real data. Don’t look at how polished the PowerPoint is. Those who dare at least have confidence in delivery. Those who don’t, probably don’t.
Persuasion hasn’t expired. It’s just that in the enterprise AI market, it’s giving way to demonstration. This shift actually demands more from companies: sales skills can be trained, but demonstration requires real capability.
Key points: The bottleneck in enterprise AI sales is trust; demonstration replaces persuasion. The three-stage machine: enter cheap, expand deep, exit clean. Three prerequisites—surplus capability, verifiable results, exit mechanism—missing any one makes imitation suicidal. Palantir’s clearances and positioning are not replicable; some marketing figures have questionable definitions.