代理型人工智能：向自治系统的演变：第一部分

EMERGING TECH RESEARCHAgentic AI: The Evolution to Institutional Research Group Autonomous Systems: Part I Dimitri ZabelinSenior Research Analyst,AI and Cybersecuritydimitri.zabelin@pitchbook.com Oscar AllawaySenior Data Analystpbinstitutionalresearch@pitchbook.comPublished on April 9, 2026 From outputs to outcomes PitchBook is a Morningstar company providing the most comprehensive, mostaccurate, and hard-to-find data for professionals doing business in the private markets. Contents Key takeaways •VC investment in agentic AI reached an inflection point in 2025 with $24.2 billionraised as enterprise adoption shifted from experimentation to deployment. •Investment activity is concentrated in IT, where the cybersecurity, developertooling, and productivity verticals are the most promising due to their fasterdeployment and measurable ROI. •M&A remains the dominant exit path across agentic AI, but platform-scalecompanies are more likely to reach IPO scale. Part II of our analyst noteseries on agentic AI will bepublished later this month. Agentic AI is movingfrom experimentation todeployment, and capital isflowing accordingly. Agentic acceleration Agentic AI is moving from experimentation to deployment, and capital is flowingaccordingly. In 2025 alone, VC-backed agentic AI companies raised $24.2 billionacross 1,311 deals, representing just under 73% of their cumulative VC deal valuebetween 2015 and 2024. Enterprises are deploying AI at scale, and investors areunderwriting the category based on real adoption and performance. The shift is structural. Traditional software models, particularly software as a service(SaaS), are built around user interaction and seat-based pricing. Agentic systemsoperate differently. They execute end-to-end workflows, replacing human labor ratherthan augmenting it. Value is increasingly tied to outcomes, not usage. This changeshow software is priced, adopted, and ultimately valued. This note argues that agentic AI is not simply a new application layer; it is reshapingthe software stack itself. As agents evolve from assistive tools to autonomousoperators, control points are shifting toward orchestration layers, workflow ownership,and system-level integration. This bifurcates the market into platform-scalecompanies that can coordinate multiagent systems and application-layer companiesthat are more likely to be absorbed through M&A. From an investment perspective, this transition in agentic AI has clear implications.Capital is concentrating in IT-centric verticals where deployment is fastest and returnon investment (ROI) is measurable. Cybersecurity, developer tooling, and enterpriseproductivity are emerging as the most credible near-term opportunities, while moreregulated sectors or those based in the physical world lag despite long-term potential. The purpose of this note is to move beyond a simple mapping of the marketand provide a framework for capital allocation. It examines where value is beingcreated, how exit pathways are evolving, and which segments are most likelyto support scalable outcomes. The goal is straightforward: Identify where newcapital should be deployed as agentic AI transitions from early adoption to coreenterprise infrastructure. This note focuses on vertically integrated agentic AI companies operating withindefined end markets and excludes horizontal platforms such as OpenAI, Anthropic,and Glean. While these platforms underpin the development of agentic systems,including them would distort valuation benchmarks and limit comparability acrossvertical business models. For the purpose of this analysis, “agentic AI” refers tosystems that autonomously plan and execute multistep tasks to achieve definedobjectives, operating across tools and workflows with limited human intervention andcontinuous feedback. North American gravity North America dominates both deal volume and value for VC-backed agenticAI companies. 60.7% of all VC-backed agentic AI companies come from NorthAmerica, primarily the US; this is almost three times greater than Europe’s shareof 21.8% and several times greater than Asia’s 11.9%. When it comes to valuations,concentration at the top is even more pronounced: North America captures 95.6%of the combined post-money valuations of VC-backed agentic AI companies. NorthAmerica’s dominance reflects structural advantages across both the technology stackand capital ecosystem. The US leads in the development of frontier models, cloudinfrastructure, and semiconductor design—factors that anchor much of the agentic AIvalue chain in the country. This proximity to core enabling technologies lowers barriersto experimentation and accelerates product iteration. The US also benefits from being the deepest and most mature VC ecosystem,particularly at the growth stage, which supports rapid scaling once product-marketfit is established. Silicon Valley remains a central hub for AI talent, research, andcompany formation, reinforcing a concentration of in