The companies that choose their value metric well will capture structurally more of the value they create than those that default to yesterday’s models.
In my work with early-stage startups, two topics come up in almost every engagement, and they’re rarely the ones founders want to talk about. The first is pricing. The second is sales. Founders, particularly those with a technology background, will happily spend months refining their architecture, their model pipeline and their product roadmap, and then set their price in an afternoon by looking at what a competitor charges and subtracting 20 percent. The result, almost without exception, is a company that underprices its product, undersells its value and wonders why it’s growing more slowly than its technology deserves.
The irony is that this is happening at precisely the moment when pricing has become one of the most strategically interesting questions in software. The rise of AI isn’t just changing what software can do; it’s breaking the dominant pricing model of the past two decades and forcing a rethink of how software companies capture the value they create. For founders building companies today, this is both a complication and an enormous opportunity.
The traditional SaaS model priced software per seat, on an assumption so deeply embedded that most people never noticed it was an assumption: Software is a tool that people use, so you charge for the people. This held for thirty years. Salesforce, Microsoft, Zendesk and the entire enterprise SaaS generation scaled on it. The model aligned neatly with customer growth: As they hired more people, they bought more seats, and vendor revenue grew with customer headcount. AI breaks this assumption at its foundation. When the software does the work rather than supporting a person who does the work, charging per person makes progressively less sense. Atlassian reported its first-ever decline in enterprise seat counts in 2026, driven primarily by AI agents replacing tasks that previously required dedicated human operators. The unit of value is no longer the seat; it’s the work performed.
The market response is visible in real-time, and customer service is the clearest battleground. Intercom prices its Fin AI agent at 99 cents per resolved conversation – not per conversation attempted, but per conversation actually resolved without human intervention. Zendesk has moved to charging for successful AI-driven resolutions instead of AI seat licenses. Hubspot cut its equivalent price to 50 cents per resolution in April 2026, an aggressive signal that at volume, outcome-based pricing can undercut per-seat models entirely. Salesforce launched Agentforce at two dollars per conversation and discovered the hard way that a poorly chosen value metric creates problems: Conversations can branch and linger without reflecting actual business value, forcing the company to introduce per-action pricing as an alternative. The buyer side has shifted just as quickly. Recent survey data from Futurum Research shows that fewer than one in five enterprise software buyers still prefer classic per-user pricing, with 43 percent preferring consumption-based models and 27 percent favoring outcome-based structures. Vendors offering only per-seat pricing increasingly risk disqualification before the evaluation even begins.
For founders, the lesson from these market leaders isn’t to copy their specific models; it’s to understand the underlying principle: The pricing metric should track the value the customer receives as closely as possible. When the metric and the value diverge, one of two things happens. Either customers pay more than the value they receive and eventually churn, or they receive more value than they pay for and vendors leave money on the table that they desperately need to fund their growth. The art of pricing is choosing a metric where these two failure modes are minimized.
This brings me to the practical advice, accumulated over years of watching technically brilliant founders make the same commercial mistakes. First, price on value, not on cost and not on competitors. The question isn’t what your product costs to run or what the incumbent charges; it’s what economic value your product creates for the customer: hours saved, revenue gained, risk reduced, headcount avoided. If your AI agent resolves customer service conversations that would otherwise cost eight euros each in human handling time, a price of one euro per resolution isn’t high. It’s a bargain, and you can defend it in every negotiation. Founders who can’t articulate the economic value of their product in the customer’s terms haven’t finished their homework, and no pricing model will save them.
Second, your price is a signal, not just a number. Technical founders systematically underprice because they’re intimately aware of their product’s limitations and price to their own anxieties rather than to customer value. But buyers, particularly enterprise ones, read price as a proxy for quality, seriousness and staying power. A price that’s dramatically lower than the value delivered doesn’t accelerate sales; it creates suspicion, attracts the wrong customers and starves the company of the margin it needs to build a real business. In my experience, the advice that early-stage founders most need to hear about pricing is simply: Double it. Almost nobody loses deals primarily on price at the early stage; they lose deals because the buyer didn’t believe the value story.
Third, start simple and iterate. Pricing isn’t a decision you make once; it’s a capability you build. Begin with one metric that customers understand and can predict, even if it imperfectly tracks value. The hybrid model, a base platform fee that gives the vendor predictable revenue and the customer predictable budgeting, plus a usage or outcome component that scales with value, has become the industry standard for AI products, now adopted by roughly four in ten AI vendors according to Bessemer’s pricing research. It’s the standard because it balances the legitimate needs of both sides. Pure outcome pricing sounds elegant but creates unpredictable bills that procurement departments hate, and it exposes the vendor to definitional disputes about what counts as an outcome. Define your value metric contractually and precisely, including what happens when the AI claims success and the customer disagrees.
Fourth, raise prices regularly and deliberately. Your product improves continuously; your price should reflect that. Grandfather your early customers if you need to preserve goodwill, but never let the fear of a difficult conversation lock your company into a price set when your product was half of what it’s today.
This connects directly to the second struggle: sales. Most technical founders treat sales as an unpleasant activity to be delegated as soon as possible, ideally to a professional salesperson hired with the seed round. This is almost always a mistake. The first ten or twenty sales must be made by the founders, not because founders are natural salespeople, but because early sales isn’t really selling; it’s learning. Every early sales conversation teaches you which problems resonate, which objections recur, what language customers use to describe their pain and what value they would actually pay for. A hired salesperson can execute a playbook; they can’t write one. Until a repeatable sales motion exists, with a known buyer profile, a tested value narrative and a price that survives contact with procurement, there’s nothing for a professional salesperson to execute, and hiring one simply burns capital while outsourcing the most important learning the company needs.
The good news for technical founders is that effective early-stage sales looks much more like engineering than like the stereotype they dread. It’s a discovery process: Form a hypothesis about the customer’s problem, test it in conversation, gather data, refine. The founders who sell well are rarely the most charismatic; they’re the ones who ask good questions, listen more than they talk and are genuinely curious about the customer’s operations. Qualify ruthlessly and disqualify fast: A polite no after two meetings is a far better outcome than a maybe after nine months. And sell the problem solved, never the technology. No buyer cares that your system uses a fine-tuned small language model with a retrieval layer. They care that invoices that took four days now take four minutes. The technology is your cost structure; the outcome is your product.
Pricing and sales are, in the end, the same discipline viewed from two angles. Both require the founder to deeply understand the value the product creates in the customer’s world and to communicate that value with confidence. The AI era has made this discipline more important, not less, because the pricing models are in flux and the companies that choose their value metric well will capture structurally more of the value they create than those that default to yesterday’s models. The technology may be new, but the underlying truth is old. To end with Warren Buffett: “Price is what you pay. Value is what you get.”
