AI ROI Gap
Most Enterprises Report High Productivity But Few Measurable Financial Returns From AI
Anonymous
Key Points
1
California’s new AI-oversight push is creating opportunities for companies that provide algorithm governance, data protection, identity management and model-assurance services. The article highlights Commvault Systems, which protects and recovers AI data, logs and identity systems, and Okta, whose identity platform controls access by people, devices and AI agents.
2
Gartner forecasts that specialized generative-AI models will rise from about $1.1 billion of an estimated $14.2 billion in model spending in 2025 to more than half of enterprise model usage by 2028, increasing the importance of systems that can route work across multiple models.
3
McKinsey’s Kate Smaje said the 37% share of CFOs reporting an earnings uplift from AI “has not moved in the last 12 months or so,” underscoring that measurable financial impact has remained stagnant despite widespread productivity gains.
4
At the Jackson Hole gathering of central bankers and economists, financial innovation—and specifically AI’s potential effect on financial stability—was a major concern. The analysis argues that overly complex rules could give AI agents more opportunities to exploit gaps and that regulators may need AI tools of their own.
5
Indian fintech executives said AI systems that could spend customers’ money remain at the pilot stage, while the Unified Payments Interface processed 24.51 billion transactions worth Rs 29.82 lakh crore, highlighting the scale of the payment ecosystem in which such controls would operate.
Companies are spending billions on artificial intelligence, yet only 5% to 6% are seeing real financial returns McKinsey. While about 80% of workers say AI boosts their productivity, just 37% of CFOs report measurable earnings gains—and that number McKinsey has not moved in the last 12 months. The gap between hype and results is pushing enterprises to shift focus from flashy AI models to governance, security, and systems that can prove value.
As companies wrestle with the AI return problem, regulators and central bankers warn of new risks. AI agents could exploit gaps in financial rules, and overly complex regulations may make matters worse Jackson Hole Economic Policy Symposium. Meanwhile, enterprises are mixing frontier models, open-source models, and specialized tools—creating urgent demand for controls over who accesses what data and which model handles which task.
The Productivity Paradox: Gains That Don't Add Up
Workers love AI. About 80% of enterprise users say it makes them more productive. But that productivity McKinsey has not translated to profit. Only 37% of CFOs report measurable earnings gains from AI—and that share McKinsey has been stuck at 37% for a year. High costs per token and scattered, disconnected projects mean companies are spending without seeing bottom-line impact.
Benefits in customer service, contract review, and fraud detection are real but hard to quantify with traditional accounting. A Florida personal injury attorney TipRanks cut medical record review time by 75% using Anytime AI. Yet scaling such wins into company-wide earnings gains remains the central challenge. Enterprises are now demanding vendors prove outcomes, not just features.
From Foundation Models to Specialized Tools
Enterprises are moving away from one big AI model. Gartner forecasts that specialized models will jump from $1.1 billion of $14.2 billion in total model spending in 2025 to more than half of enterprise usage by 2028. Companies are now mixing frontier models, open-source options, and narrow tools built for specific jobs. That shift demands new infrastructure.
California's strict AI rules are accelerating this shift. Companies need stronger identity controls, data protection, and model auditing to manage multiple tools safely. Gartner analyst Arunasree Cheparthi noted that Gartner, "the biggest winners will be vendors that help enterprises manage where and how AI is used." Firms like Commvault Systems and Okta are capitalizing on demand for governance and identity platforms.
Financial Regulators Worry AI Could Break the Rules
At Jackson Hole in late August 2026, central bankers heard urgent warnings. Jackson Hole Economic Policy Symposium Princeton economist Markus Brunnermeier cautioned that Jackson Hole Economic Policy Symposium AI agents create asymmetric understanding in markets. Complex financial rules give algorithms room to find loopholes. Simpler rules and AI-powered supervisory tools may be needed to catch manipulation.
The risk is concrete. In India, automated spending agents—AI systems that move customer money—remain strictly in pilot programs. The Unified Payments Interface Live Research handles 24.51 billion transactions worth Rs 29.82 lakh crore monthly. At that scale, even small control failures could cause enormous damage. Regulators worldwide are racing to build safeguards before such systems go live.
What Enterprises Are Buying Now
Spending is shifting from raw model development to infrastructure that proves value. Enterprises are purchasing identity management tools, data recovery systems, algorithm auditing services, and platforms that route work across multiple specialized models. ConductorOne has outlined a governance-focused approach for adopting agentic AI over 90 days, emphasizing control and oversight from day one.
McKinsey and competing consulting firms are now tying fees to measurable earnings gains—shifting up to 25% of contracts to outcomes-based pricing. That move signals a sea change. Vendors can no longer sell capability; they must sell results. The companies winning are those that help CFOs close the gap between worker productivity and company profit.
Sentiment
Mostly NegativeNEGATIVENEUTRALPOSITIVE
Publishers
40
Articles
14
Reach
54
A
AnonymousAuthor
