Exclusive: KPMG and OpenAI bet the future of software is ‘headless’ — and the future of work is mostly talking

This reads as a budget-shift story, not a near-term earnings shock. The first capital dollars in agentic deployments should accrue to implementation-heavy firms and governance layers, because enterprises will pay for orchestration, permissions, auditability, and change management before they trust autonomous execution. That makes ACN and IBM-style services more relevant than pure model narratives, and it also pulls spend toward identity/security names such as PANW and OKTA as every agent action creates a new control point.

The more interesting multi-year risk is to seat-based SaaS with weak proprietary data. If the user interface becomes a conversational wrapper, the value migrates from app interaction to the orchestration layer, which can compress renewal quality and slow net new seat growth for CRM- and workflow-centric vendors. By contrast, systems-of-record franchises like ORCL and, to a lesser extent, NOW may be more resilient because the agent sits on top of their data and can actually deepen lock-in by becoming the queryable knowledge layer.

The contrarian take is that consensus may be overpricing immediate software disruption and underpricing verification overhead. Every faster workflow creates more exception handling, approvals, and liability transfer, which can expand the labor bill in the middle of the stack rather than eliminate it. The thesis is falsified if consulting firms fail to show AI-related bookings over the next 1-2 quarters, or if CRM/WORKDAY-type vendors demonstrate stable retention and pricing power despite agent rollouts.