In September 2025, BCG published “The Widening AI Value Gap,” a study covering 1,250 senior executives and AI decision-makers across 9 industries and more than 25 sub-industries. The report notes that, of the AI-driven bottom-line value reported by respondents, Agentic AI’s share is expected to rise from about 17% in 2025 to 29% in 2028. The denominator for the 29% is the expected allocation of AI-driven bottom-line value among predictive, generative, and agentic AI; it is not total company value, a share of revenue, or realized results.
This figure is a survey expectation. It does not mean every company will achieve it, nor does it support a causal conclusion that “AI agents can solve hallucinations.” The same BCG report describes agents as presenting both opportunities and risks, and lists hallucinations, bias, data misuse, and adverse customer experiences as issues requiring governance. Agents can plan, call tools, and adjust steps, but they can also amplify errors across multi-step processes or external systems.
When verifying agents, companies should test capability and controls separately: first define the ground truth for the task, the data sources, and who is responsible for updates; then measure factual accuracy, citation traceability, tool-call success rate, erroneous action rate, and human takeover rate. For actions involving writes to ERP/CRM, external communications, payments, quotations, or permission changes, also apply least privilege, human approval, complete audit trails, stop conditions, and recoverable mechanisms.
BCG also notes that agents depend on human orchestration and oversight, and that value comes from end-to-end process redesign, not from adding another model or automated check. From the perspective of “The CIO for Your Business Operations,” Unima recommends using four gates to decide whether to scale: value baseline, data and evaluation, permissions and accountability, and monitoring and exit. This is an implementation recommendation compiled from first-party research; it does not represent specific client results, zero hallucinations, or a return-on-investment commitment.
First-party sources: Boston Consulting Group, “The Widening AI Value Gap: Build for the Future 2025,” 2025-09-30, https://www.bcg.com/publications/2025/are-you-generating-value-from-ai-the-widening-gap; full report: https://www.bcg.com/assets/2025/the-widening-ai-value-gap.pdf. This article was verified on 2026-08-25; the 29% is limited to the source-defined expected allocation of AI-driven bottom-line value in 2028.