So long, SaaS: Why AI spells the end of per-seat software licenses - and what comes next (www.zdnet.com)

A McKinsey-led analysis argues AI — especially agent-to-agent architectures from companies like OpenAI and Anthropic — will disrupt traditional SaaS and per-seat licensing, ushering in a “post‑SaaS” era where software is sold by usage, outcome, or task rather than by user seat. Vendors expect material revenue and cost impacts (40% foresee >20% revenue lift; 11% >50%), and analysts predict seat counts could fall as much as 70% as AI agents automate workflows and act on users’ behalf. The industry is already shifting: about 63% of vendors say AI will fundamentally change their business model within three to five years, and leaders are moving to consumption- and outcome-based pricing versus token/compute-based schemes that many business buyers find hard to interpret. For the AI/ML community this raises technical and operational imperatives: product teams must design agentic systems that are reliable, observable, auditable and integrable across enterprise stacks; pricing and API models must support fine-grained usage measurement; and governance, transparency and human oversight become central as agents take on decision-making. There are real risks for customers—opaque outcome bundles, unverifiable metrics, and immature agent-to-agent stacks—so vendors should prioritize measurable SLAs, explainability, and robust testing, while buyers must demand performance data and trial single capabilities before committing to consumption-based contracts.