AI Disrupts Pricing Foundations: Bidding Farewell to 'Man-days', Navigating Towards a New Era of 'Model + Data + Process

Traditional SaaS business logic is rooted in the linear growth of user scale, meaning 'the more people use it, the more revenue is generated.' However, the rapid development of AI technology is disrupting this balance. As observed by Jake Saper, the increased efficiency of AI software directly leads to a reduced demand from customers for seat numbers, creating the anomalous situation where 'the more successful the sales, the fewer units available for sale.' This is not merely a contradiction between efficiency and revenue, but a fundamental issue with the underlying unit of measurement—when intelligent Agents can complete tasks independently, continuing to account for value based on 'people / seats / hours' is akin to using an outdated proxy variable to measure new productive forces.

This transformation is far from simple efficiency optimization; it is a business paradigm revolution. There are four core criteria for judging whether it is revolutionary: the billing anchor shifts from input to output; scarce elements transform from human labor to models, data, and processes; marginal delivery costs shift from increasing with scale to decreasing; and suppliers bear the risk of delivery outcomes, not merely tool availability. Currently, practices such as layering AI within seats, additional Copilot licenses, or still quoting 'per man-day' after AI-driven efficiency gains, are all improvements within the old paradigm. The true revolution lies in Agents becoming the primary delivery entity, with suppliers charging based on 'successful outcomes' and assuming corresponding risks, and scarce elements shifting to 'model + data + process.' However, it must be emphasized that this revolution is ongoing and has not yet reached its conclusion.

Although the torrent of paradigm revolution is surging, we are still in a transitional state where 'old anchors have not yet died, and new anchors have not yet taken shape.' A survey by Bain & Company in August 2026 of SaaS vendors [Verified] reveals that up to 35% of companies merely bundle AI features into higher-priced tiers, 65% adopt a 'seat + usage' hybrid billing model, while companies that have fully transitioned to pure outcome or usage-based pricing remain at zero. This data warns us not to mistake a trend for an established fact. However, evidence of billing anchors loosening and shifting downwards is already clear. For example, Intercom's Fin AI Agent has switched to outcome-based billing [Verified], charging $0.99 for each problem solved and $9.99 for qualified lead identification, with no charge for failures or transfers to human agents; Salesforce Agentforce's billing unit has also further descended from 'conversations' to the 'actions' level [Verified].

During this transitional period, AI has not only altered billing methods but also spawned unprecedented business models. Cases of per-employee revenue breaking historical limits in the software industry are constantly emerging [Verified (media/third-party estimates)]: According to Epoch AI's estimations, Anthropic, OpenAI, and NVIDIA have per-employee revenues of up to $9 million, $5.6 million, and $5.1 million, respectively. More astonishingly, companies composed of very few individuals, such as Anysphere and Midjourney, can achieve per-employee productivity of even $15 million to $20 million. This systematically proves that revenue scale can now be decoupled from employee scale. Furthermore, data has transformed from a previously free input into a chargeable asset: Reddit and Google reached a $60 million content licensing agreement [Verified], and Cloudflare also launched a 'Pay Per Crawl' marketplace [Verified], allowing content providers to charge for AI crawling, signaling a new direction for data monetization.

If the old paradigm becomes obsolete, where will the value of the new paradigm come from? We infer [Inference] that value capture will concentrate on three separately measurable inflection points. First is the 'model' itself, which can serve as an external capability billed by token/action, or be internalized as a delivery engine. The core of value capture lies in the scarcity of the model's capabilities. Second is 'data.' Proprietary data, customer feedback, and evaluation sets are not only the cornerstone of training but also a valuable asset. For emerging businesses, high-quality domain-specific data and acceptance feedback are themselves core assets, rather than mere process byproducts. Finally, there is 'process.' Orchestrating domain-specific SOPs into reliable Agent workflows constitutes a new form of 'complementary asset' as defined by Teece (1986). When model capabilities tend towards commoditization, data and processes become crucial for filling reliability gaps and building competitive barriers.

Citing Coase's (1937) theory of the firm, when external services can be obtained via AI Agents at a lower cost than internal coordination within the enterprise, firm boundaries will systematically shrink, and the 'one-person company' will thus transform from an anomaly into a viable mainstream organizational structure. And according to Kuhn's theory of paradigms, when the pricing language of an old framework fails due to accumulated anomalies, a true paradigm shift occurs, which is precisely the definition of a revolution. However, it is important to be wary that replacing human labor is not without friction. Klarna claimed in 2024 that its AI assistant handled the workload of 700 customer service agents, but restarted recruitment in 2025 due to declining service quality [Verified]; after Salesforce cut its support team, executives admitted that the model began omitting instructions after more than 8 commands [Verified]. These cases reveal [Inference] that human labor, mistakenly deemed replaceable, is actually a complex entity encompassing 'execution + verification + ultimate responsibility.' The old paradigm's failure is not due to a lack of efficiency, but because its unit of measurement cannot match this complex and precise composition of value.

Faced with the sweeping paradigm shift, enterprises must act immediately to adjust their value capture and pricing strategies. First, they should actively shift pricing anchors towards 'output.' Even if pure outcome-based billing cannot be immediately implemented, the capability to measure 'actions/outcomes' should be established, as this is the foundation for building new anchors. Second, domain-specific data and acceptance feedback must be managed as core assets. They are no longer mere operational byproducts but rather the source that drives the value cycle and enables compound growth: usage generates feedback, feedback drives evaluation and improvement, which in turn leads to more usage.

Third, processes should be built as the enterprise's moat. In the context of increasingly widespread models, what truly differentiates companies is their ability to orchestrate models into reliable, verifiable, and billable outcomes. Fourth, do not blindly 'de-humanize'; critical verification and fallback mechanisms must be retained. The lessons from Klarna and Salesforce are clear [Verified]: Rushing to cut human labor may come at the cost of service quality and accountability. Finally, be wary of the cognitive bias of equating 'trends' with 'facts.' The current 65% hybrid billing model reminds us that the market is still in a transitional phase [Verified], and pure outcome-based systems still require room for error tolerance; a steady approach is key to success.