Across our pipeline and portfolio, we’re seeing B2B pricing shift from “pay for access” to “pay for work done.” Seats and basic usage tiers made sense when software was mainly a better tool for humans. But as AI agents and workflows start doing a meaningful share of the job, pricing is catching up with that reality and shifting towards credits, tokens and outcome-based constructs.
While the usage based pricing started becoming more commonplace a few years back, even this pricing model is proving inadequate in the current scenario. Billing only on API calls or tokens makes invoices hard to predict for customers and unit economics hard to manage for vendors. A single misconfigured workflow can consume huge volumes of tokens; a small group of power users can drive a disproportionate share of costs. Outcome pricing goes further—charging per successful hire, per audit passed, per playbook fully deployed, per ticket resolved—so the invoice lines up with a business result, not just activity in a dashboard.
| Software Generation | Pricing Complexity | Prevalent Pricing Models |
| Software Era | Simpler pricing models, one person could manage pricing | Perpetual Pricing, Seats |
| Access Era | Pricing changes need engineering, GTM & Product Support | Consumption, Subscription, Seats |
| Value Era | Pricing and value impact every function. | Consumption, Subscription, Seats, Agents, workflows, Outcomes, Actions, Hybrid, Commit based, Credit, Token, Inputs, Future Unknowns |
Source: Metronome | Whitepaper: The Monetization Operating Model Pentathlon Ventures Quarterly Newsletter: July-September 2025
This is tightly linked to the productisation of services. Buyers increasingly do not want to assemble tools, systems integrators and training to get to an outcome; they want a bundled offer where the vendor owns the result. A good example is SAP or large CRM implementations. Historically, this has been a pure services business run by global integrators—multi-year projects staffed by TCS, Infosys and similar players. The billing here was on the basis of effort (manhours). We are now seeing product companies build templates, automation, remote expert pods to productise the services. The degree of productisation varies by use case, but every step in that direction expands TAM: instead of only selling licenses, these startups now compete forthe same multi-million-dollartransformation budgets that used to be reserved forthe big services firms.
The same pattern shows up in hiring-as-a-service, GTM implementation and compliance. A hiring product that delivers onboarded reps, a RevOps platform that shows up as configured Salesforce with live playbooks, a compliance stack that keeps a company audit-ready—each of these sells against an internal team, agency, BPO or consulting spend, not just “software tools” budget. That is why we think the move towards outcomes and productised services structurally increases the opportunity size, even if the pricing model becomes more complex.
The focus on outcome (and not just access or usage) changes the way the offerings are delivered and often has an impact on the unit economics. Gross margins are under pressure as AI infra and humans-in-the-loop become part of delivery—benchmarks illustrate the spread: For AI native companies, research suggest the following gross margins: Anthropic (~60% GM as of July), Perplexity (~60% as of 2024), OpenAI (~50% projected for 2025), StackBlitz (~40% as of July), Lovable (~35% as of May), and Replit (~23% as of July). The nuance is that many AI-native companies are also scaling meaningful revenue with remarkably small teams, so lower headline GM can still coexist with healthier EBITDA if indirect costs (sales, implementation, support) compress faster. Old proxies for scale—like needing ~50 people in India to reach ~$1m ARR—are breaking; it’s increasingly plausible to see Series A companies with fewerthan ten employees, even in enterprise categories, and customer success is a visible example where one CSM can manage more high-ACV accounts because workflows, in-product guidance, and AI assistance absorb repetitive work.
Net-net, we may be entering a regime of lower gross margins but structurally lower indirect costs. EBITDA can look healthier than the headline GM suggests because AI compresses both the cost of delivery and the cost to sell, implement and support. That raises a new question for early-stage investors: if companies become profitable earlier, how should we think about “acceptable” burn and the timing of profitability, especially in venture-backed businesses that raise funding for mostly either building tech or distribution. If startups achieve profitability sooner, even with lower Gross Margins, they will ultimately require less venture capital in the long run and late stage rounds in B2B software tech could become less common. Will secondaries in future rounds of financing remain as viable of an exit option for us in such businesses?
Bottom line
For us as a fund, this reinforces that “SaaS vs services” is no longer a clean binary. We expect to evaluate both: classic SaaS where “good” still means strong NRR, disciplined CAC payback, and 70–80% GM, and outcome/credit-driven businesses where “good” means improving contribution margin per outcome, a clear roadmap to automate delivery over time, and contracts that migrate from project-based work to multi-year commitments. Across both, pricing plumbing—metering, billing, and value communication—is increasingly where strategy shows up.
