New January reports from PwC, Anthropic, and OpenAI reveal a critical pivot: while AI usage is scaling, financial returns remain elusive for the majority. Here is the new measurement playbook for crossing the gap.
The era of unmeasured experimentation ended this month. A cluster of reports released between January 15 and January 22, 2026—from PwC, Anthropic, OpenAI, and Google—signals a shift from capability to accounting. The verdict is clear, and it is uncomfortable: adoption is scaling, but value is stalling.
For boardrooms, the implication is immediate. The metric of 2025 was “users.” The metric of 2026 is “auditable outcomes.”
The ROI Reality Check
The disconnect is glaring: investment surges, but returns lag. PwC’s 2026 CEO Survey provides the definitive checkpoint: 56% of CEOs report neither increased revenue nor decreased costs from AI in the last 12 months. Only 12% report achieving both.
This figure indicts “pilot sprawl.” It suggests that while access to tools has democratized, the transformation required to monetize them has not. The divide is structural rather than accidental. CEOs who report financial returns are two to three times more likely to have embedded AI extensively across decision-making and demand generation. They haven’t just bought licenses; they have rewired operations.
The lesson is stark: AI spend does not become ROI simply because usage goes up. Value capture requires workflow redesign, not just license distribution.
The New Unit of Analysis: “Primitives”
If “number of users” is a vanity metric, what replaces it? Anthropic’s January 15 report offers a superior framework: “economic primitives”.
Rather than tracking generic login activity, this approach measures task complexity, autonomy, and success rates. It distinguishes between a user asking a chatbot to summarize an email—low complexity, low autonomy—and a user delegating a multi-step coding workflow.
This granularity is essential for ROI modeling because different tasks have vastly different economic footprints. Anthropic’s data indicates that a software development request averages 3.3 hours of human-equivalent work, while personal management tasks average just 1.8 hours.
Boards must ask CIOs to report on this “work type mix.” High adoption of low-value primitives is a cost center; targeted adoption of high-complexity primitives is a productivity engine.
Breadth vs. Depth: The Capability Gap
OpenAI’s January 21 analysis reinforces this “depth” thesis. Their data reveals a massive “capability overhang”—the gap between what the models can do and how they are actually used.
Two data points define the opportunity. First, the Power User Dividend: typical power users rely on advanced “thinking capabilities” seven times more than average users. Second, the National Divergence: across 70+ countries, there is a 3× gap in the usage intensity of these advanced capabilities.
For multinationals, this creates a new competitiveness variable. Talent markets are no longer just about digital literacy; they are about agentic fluency. A team in a high-depth region will outperform a team in a low-depth region, even if they use the same software.
The Control Plane: Instrumentation as Governance
Management requires measurement. Google’s January 20 update to Workspace is strategically telling: it exposes Gemini usage metrics—active events, unique users—directly in admin dashboards.
This moves AI from a “shadow IT” phenomenon to an instrumented line item. Dashboards create the audit trail necessary for finance to track consumption against departmental P&L. If you can’t measure the usage spike, you can’t validate the efficiency claim.
What This Means For Leaders
- Separate Usage from Value: Demand reporting that distinguishes “breadth” (logins) from “depth” (complex workflows).
- Target the “Primitives”: Audit your AI pilots for complexity and autonomy. If it’s just summarization, it’s not transformative.
- Instrument the Workflow: Use admin dashboards to map consumption to specific teams, then correlate with output KPIs.
- Fund the Redesign: Recognizing that ROI correlates with “embedded” AI, allocate budget for process re-engineering, not just software.
- Monitor the 12%: Study the minority of competitors reporting both cost and revenue gains—they are your actual threat.
Watchlist: Next 90 Days
- KPI Standardization: Expect CFOs to demand standardized “AI P&L” packs rather than narrative updates.
- Vendor Lock-in via Metrics: Vendors will race to define the “standard” measurement of productivity.
- Regulatory Interest: As measurement improves, regulators will request data on “autonomy” and “safety” primitives.
- The Agent Pivot: Investment will shift toward agentic workflows that promise higher autonomy—and higher risk.
The bottom line: The honeymoon phase of “AI magic” is over. We have entered the phase of AI accounting. Competitive advantage will now accrue to the organizations that stop celebrating pilots and start auditing outcomes.