The assumptions below are the highest-signal, most decision-relevant claims from across all eleven organizations covered in this briefing, condensed for executive readability. Full context, nuance, and sourcing for each appears in the sections that follow.

1. Customer: Agent-mediated commerce becomes a primary channel. Accenture and Amazon both assume AI agents will increasingly compare, choose, purchase, and manage returns on a customer's behalf, with a meaningful share of shoppers ready to switch brands based on their AI assistant's recommendation.

2. Customer: Brands will be evaluated by agents, not just by people. BCG assumes AI agents will judge brands on observable performance rather than marketing claims, and that companies which fail to adapt risk becoming background utilities inside agent-controlled marketplaces.

3. Customer: AI will begin acting as a personal agent for high-stakes decisions. Google's Sundar Pichai assumes AI will soon guide decisions such as investment choices and medical treatment options directly on a person's behalf.

4. Customer: The interaction model itself will be disrupted. Gartner assumes agentic AI will trigger the first real challenge to mainstream productivity and interaction tools in 35 years, shifting value toward agent-mediated experiences.

5. Workforce: Human-in-the-loop oversight will decline sharply. Gartner assumes human-in-the-loop checkpoints in IT operations will fall to 40% by 2028, down from 95% in 2025, as agentic autonomy expands.

6. Workforce: Every employee becomes an “agent boss.” Microsoft assumes workers will shift from doing tasks themselves to building, delegating to, and managing AI agents, with 82% of leaders expecting digital labor to expand workforce capacity within 12 to 18 months.

7. Workforce: Corporate headcount will shrink as agents absorb routine work. Amazon assumes its own corporate workforce will contract as agents take on more rules-bound, high-volume tasks, and has already cut tens of thousands of corporate roles as a result.

8. Workforce: Organizations, not individuals, determine whether AI actually changes output. Microsoft's 2026 data assumes the deciding factor is whether a company has restructured itself around agents — echoed by Amazon's shift to small, flat, agent-native teams.

9. Workforce: Impact will be uneven — some roles deskilled, others upskilled. Anthropic's usage data assumes AI absorbs the most demanding parts of some jobs (deskilling) while freeing others for higher-value work (upskilling), with productivity growth of roughly 1.8 percentage points.

10. Workforce: The most valuable human skill shifts from execution to systems thinking. Nvidia assumes that once most software is written agentically, the premium human skill becomes designing and reasoning about complex agent-run systems rather than writing code line by line.

11. Operating Models: Over 40% of agentic AI projects will be canceled by 2027. Gartner's headline Strategic Planning Assumption attributes this to escalating costs, unclear business value, inadequate risk controls, and legacy systems that cannot support real-time agentic execution.

12. Operating Models: 2027 is the inflection year from generative to agentic AI. Deloitte assumes organizations that built real agentic infrastructure in 2025–2026 will enter 2027 with a compounding advantage over those still running pilots.

13. Operating Models: Process design shifts from tasks to outcomes. BCG assumes agentic enterprise operations will be redesigned end-to-end around outcomes rather than tasks, with early transformations already showing up to 80% straight-through processing and 60% cost-out.

14. Operating Models: Compute demand will be driven by agents, not humans. Nvidia assumes roughly $1 trillion in AI compute demand will materialize through 2027, with the semiconductor industry needing to grow 10x over the next decade to serve a world of 100 billion AI agents.

15. Operating Models: Adoption will lag ambition. McKinsey and AWS-commissioned IDC research both assume a persistent gap between experimentation and production — McKinsey finds 62% of organizations experimenting but fewer than 25% at scale, while a competing IDC estimate has 65% expecting full deployment by 2027, underscoring how unsettled this forecast still is.