Evidence-based perspectives on AI enablement, attestation, and the organizational challenges facing senior leaders today.
The research is unambiguous. McKinsey's 2025 State of AI survey — drawing on nearly 2,000 respondents across 105 countries — finds that while 88% of organizations now use AI in at least one business function, nearly two-thirds remain stuck in experimentation or piloting mode. Only 39% report any enterprise-level EBIT impact.
This is not a technology problem. The models exist. The tools are available. The APIs are open. The bottleneck, consistently and across industries, is organizational — the gap between deploying a tool and embedding AI into how work is actually done.
McKinsey identifies workflow redesign as the single strongest predictor of enterprise-level AI value. High-performing organizations are not simply adding AI to existing processes. They are rebuilding those processes with AI as a structural component — redefining task ownership, decision points, and human oversight in parallel with capability deployment.
Bain's 2025 research reinforces this. Organizations achieving 25–30% productivity gains from AI — compared with the 10–15% most report — share a common characteristic: they treated enablement as a product, with centralized governance, role-specific training, and measurement infrastructure built before deployment, not after.
The implication for CEOs is direct: if your AI initiatives are not producing enterprise-level results, the diagnosis is almost certainly not the AI. It is the organization surrounding it. The solution is structured enablement, systematic workflow redesign, and the measurement infrastructure to verify that change has occurred.
88%
of organizations now use AI in at least one business function
McKinsey State of AI, 2025
39%
report measurable EBIT impact at the enterprise level
McKinsey State of AI, 2025
25–30%
productivity gains in organizations that combine AI with workflow redesign
Bain Technology Report, 2025
10–15%
typical gains without structured enablement and workflow redesign
Bain Technology Report, 2025
The phrase "human in the loop" has become the default answer to every board question about AI oversight. We hear it in earnings calls, board presentations, and regulatory responses. It has become a verbal shorthand for responsible AI — and that is precisely the problem.
Bain's 2025 AI Enterprise research draws an important distinction between "human in the loop" — where humans supervise every AI output before it is acted upon — and "human on the loop" — where humans manage AI execution at a system level, reviewing exceptions and deviations rather than every output. Both are legitimate oversight models for different risk profiles. Neither is a governance strategy by itself.
A governance strategy requires documented policies defining which oversight model applies to which AI use case and why. It requires monitoring systems that can detect when AI behavior deviates from expectations. It requires escalation protocols, performance metrics, and — critically — the organizational capability to actually review what the monitoring surfaces.
Most organizations have none of this. They have a policy document, a reference to "human oversight," and a technology team that is too busy to build the monitoring infrastructure needed to make that oversight real.
The organizations that will lead on AI accountability are not those with the most sophisticated AI. They are those with the clearest answer to a simple question: if something goes wrong, who knew what, when, and what did they do about it?
Human In The Loop
Every AI output reviewed before action. Appropriate for high-stakes, low-volume decisions. Requires significant human capacity.
Human On The Loop
AI executes within defined parameters; humans review exceptions. Appropriate for high-volume, lower-stakes processes. Requires robust monitoring.
The AttestiFi position
Both models require the same foundation: documented policy, behavioral monitoring, and the organizational capacity to act on what the monitoring reveals. That foundation is what AttestiFi builds.
AI is not yet a CFO problem in most organizations. It lives in the CDO's office, the CIO's budget, and the CEO's strategic narrative. Finance is a consumer of AI outputs but rarely a principal in AI governance. That is changing — and the change is coming faster than most finance leaders are prepared for.
Three developments are converging. First, AI is moving into material financial processes: revenue forecasting, credit assessment, procurement, and financial close. When AI makes or informs decisions that appear in financial statements, the CFO has a direct accountability stake in that AI's behavior. Second, investor and board questions about AI are increasingly financial in nature: What is the return on our AI investment? What is the exposure if an AI system behaves unexpectedly in a regulated process? Third, the emerging framework for AI disclosure in financial reporting places the CFO at the center of any attestation to AI's role in material processes.
The CFO's AI agenda in 2025 has three items. First: know what AI is running in processes you attest to in financial reports. Second: have a monitoring capability that can tell you if something changes. Third: be able to answer, in plain language, how your organization would respond if an AI system in a material process produced an erroneous output. Most CFOs cannot yet answer all three. The ones who can will be significantly better positioned when the questions escalate.
01 — What AI is running in processes I attest to?
Complete inventory of AI systems in financial and regulated processes.
02 — Do I have monitoring that detects changes?
Behavioral surveillance with alerting when outputs deviate from parameters.
03 — What is my response protocol if AI fails?
Documented escalation, remediation, and disclosure procedures.
Research referenced: McKinsey & Company, "The State of AI in 2025: Agents, Innovation, and Transformation," November 2025; Bain & Company, "The AI Enterprise: Code Red," February 2026; Bain & Company, "Want More Out of Your AI Investments? Think People First," February 2026; Bain & Company Technology Report 2025. All statistics cited are drawn directly from published primary research. AttestiFi perspectives represent original analysis and editorial commentary, not assertions of fact from third-party sources.
We publish original analysis and curated research summaries for executive audiences. No noise — only what is actionable at the leadership level.
Request a Briefing