99% of Companies Eye Agentic AI, But Just 9-14% Put It to Work

Nearly 99 per cent of companies plan to deploy Agentic AI, but only 9-14 per cent have successfully moved such systems into production, highlighting a significant gap between AI experimentation and real-world business adoption, according to a report by Ness Digital Engineering.

The report describes this gap between proof-of-concept (POC) and production as “Death Valley”, where companies struggle to turn promising AI-agent experiments into reliable business applications. Concerns over the reliability of probabilistic generative-AI systems and the lack of visible improvements in users’ daily experiences are among the factors slowing adoption.

Agentic AI differs from conventional generative AI by allowing systems to plan and execute multi-step tasks, use tools and APIs, access authorised data and adjust actions based on results. Its potential extends beyond IT to sectors including banking, healthcare, retail, manufacturing, insurance, logistics, telecom and government services.

The report said traditional financial institutions are mainly exploring Agentic AI for back-office efficiency and IT development, while digital-native fintech companies are more likely to integrate it into customer-facing products and experiences.

The Agentic AI market in financial services is projected to reach USD 33.26 billion by 2030, increasing pressure on boards and investors to demonstrate measurable returns from AI investments.

Ness recommends that companies assess their business processes, technology infrastructure, data and API readiness, permissions and user-experience opportunities before scaling AI agents. It also suggests using a single interaction interface through which agents can gather information from multiple systems and execute actions according to business rules.

The report further recommends measuring AI projects by their overall impact on user experience, rather than simply counting automated tasks. Companies should also track LLM token consumption, AI subscriptions, cloud expenditure and other costs at the application level to evaluate return on investment.

As Agentic AI adoption expands, the report said the priority over the next four to five years will be moving beyond experiments and developing reliable systems capable of completing end-to-end business processes.