新兴科技研究：透过棱镜：构建企业级AI的竞赛：用SaaS解锁劳动力市场

EMERGING TECH RESEARCHThrough the Looking Institutional Research Group Glass: The Race to BuildEnterprise AI Rudy TorrijosDirector, Industry Researchrudy.torrijos@pitchbook.com Derek HernandezSenior Research Analyst,Enterprise SaaS andInfrastructure SaaSderek.hernandez@pitchbook.com Unlocking the labor market with SaS pbinstitutionalresearch@pitchbook.com PitchBook is a Morningstar company providing the most comprehensive, mostaccurate, and hard-to-find data for professionals doing business in the private markets. Published on March 9, 2026 Contents This report is the second installment of the seriesSaaS is Dead, Long Live SaS.Read Part Ihere. constraintsExecutive summary intelligence1214The enterprise software industry is undergoing a profound structural shift, moretransformational than the move to the cloud 20 years ago. The seat-based subscriptionmodel on which software as a service (SaaS) has relied is dead. Service as software(SaS), defined by digital labor outcomes, is its replacement. This transition is bothtechnological and economic: It represents the merger of the over $900 billion globalenterprise software market with the over $50 trillion global knowledge worker labormarket.1, 2 The race to build enterprise AI is defined by two parallel tracks: the SaS vendor racingto capture the payroll budget via outcome-based pricing and the enterprise agentoperations (AgentOps) team racing to capture the intelligence via proprietary curation.Success will be measured by expanding revenue and margins on one side, and bycompounding value creation from the digital workforce on the other. This report provides a rigorous framework for identifying incumbents and startupscapable of migrating payroll budgets into software spend while maintaining stronglong-term viability. We define eight specific scoring frameworks stakeholders canuse to identify winning business models and durable moats across the SaaS and SaSlandscapes today. We also include market drivers, context, and a comprehensive glossary of essentialenterprise AI terms. Our core thesis is that generative AI (GenAI) has commoditizedcode production, eroding many of the defensive moats of the SaaS era. Existing software vendors must transition from selling access to tools that require humanoperators to selling outcomes that replace entire employee workflows. The primary barrier to this transition is what we call the “autonomy gap,” theoperational cost incurred when an autonomous agent fails and requires humanintervention. Until this gap is closed, the economics of labor replacement remainunattainable. Therefore, success in this new era depends on three simultaneousrequirements: (1) Vendors must close the autonomy gap and (2) adopt outcome-basedpricing to capture the value of labor, while (3) enterprises must adopt AgentOps togovern the risk of autonomy. The pivot from SaaS to SaS introduces a period of financial risk, especially forincumbents. As enterprises deploy digital workers, seat counts may reduce seat-basedrevenue before outcome-based revenue can scale to offset the loss. This is the SaSJ-Curve. To navigate this transition carefully, vendors must manage three distinctfinancial levers that determine their long-term viability: self-cannibalization, the cost ofproviding intelligence, and the true revenue potential of their customers. We believe we are at least 18 months ahead of when agentic systems will be readyto replace basic employee workloads given software development cycles, enterpriseoperational readiness, minimum large language model (LLM) expert intelligencelevel requirements, economically viable cost per token pricing, and the time-to-powerrequirements for next-generation AI datacenter computing capacity. Key takeaways •The autonomy gap is the gating factor:The defining challenge for enterprise IT ismanaging the cost of human intervention required to keep agents running. Successis measured by SaS’s ability to move quickly from requiring human validation formany AI decisions to only requiring human oversight in rare exceptions. •AgentOps is the new enterprise IT mandate:Deploying autonomous software requiresa fundamental retooling of the workforce. Enterprises must establish platform-centricinfrastructure and new roles such as “reliability architects” and “ris k governors” tomanage the life cycle, behavior, and skill atrophy of the digital workforc e. •Intelligence is elastic:Unlike legacy software, agentic performance scaleslinearly with compute cost. The physics of autonomy dictates that higher-fidelityreasoning consumes exponentially more compute. Intelligence is no longer afixed salary-based asset; it will be a variable cost. Vendors must balance modelexpert intelligence with inference cost-efficiency to maintain margins and prevententerprise pricing from exploding. •The market has bifurcated:This taxonomy distinguishes between incumbentsleveraging data moats and distribution to bundle AI and architecturally adaptiveAI