By the numbers
- 95% of executives have faced at least one negative incident from enterprise AI use.
- 39% describe the damage from such incidents as “severe” or “extremely severe.”
- Less than 2% of companies meet best-practice standards for Responsible AI (RAI).
What’s happening: A new report from the Infosys Knowledge Institute warns that companies are embracing AI faster than they’re building the safeguards to deploy it responsibly, a gap that could have serious financial and reputational consequences, especially as “agentic AI” systems, which act autonomously, become more common.
Why it matters: AI models now play roles in credit decisioning, fraud detection, and consumer engagement. A single flawed algorithm can cause regulatory headaches, lost revenue, and reputational harm, whether due to bias, poor explainability, or security lapses. With regulators signaling stricter oversight, Responsible AI is shifting from a “nice to have” to a core operational requirement.
Key findings:
- Incidents are common: Most often privacy violations, systemic failures, and harmful predictions. While 77% of AI incidents result in direct financial loss, executives rank reputational damage as more threatening to long-term business health.
- Execution lags enthusiasm: 78% of leaders see RAI as a growth enabler, but 83% of companies implement it piecemeal. Only 2% qualify as “RAI leaders,” meeting high standards for trust, governance, and sustainability.
- Bigger teams, mixed results: Larger RAI teams handle more AI projects but tend to see lower success rates, partly because increased volume brings more risk exposure.
- Agentic AI raises the stakes: 86% of executives aware of the technology believe it will create new compliance and risk challenges, requiring tighter governance frameworks.
The playbook for safer, more effective AI: The report makes four recommendations:
- Learn from leaders who have navigated a wide range of AI incidents.
- Combine product and platform models to scale AI while keeping centralized oversight.
- Embed RAI guardrails directly into AI development platforms.
- Establish a proactive RAI office to monitor risks, enforce governance, and adapt to emerging threats.
The bottom line: AI can accelerate recoveries, personalize outreach, and reduce operational costs, but without embedded Responsible AI practices, those same tools could trigger costly mistakes. As agentic AI takes hold, the winners will be the ones that build trust, manage risk, and treat RAI as a growth driver, not just a compliance checkbox.
