Only 16% of Companies Reap Real Benefits from AI

Organizations are using AI, but few are reaping transformative effects. A co-founder of Atlan points to the lack of a contextual layer to bridge business systems and models. Survey data primarily shows deployments in testing or limited to assistance, with very few agents in production.

Without Context, AI Remains Underutilized, Says Prukalpa Sankar

Prukalpa Sankar, co-founder of Atlan, describes a significant gap between the technical capabilities of models and their actual utility within organizations. She believes that, for most companies today, AI does not provide truly demonstrable use. For her, the missing piece is context, meaning how a system understands the business that employs it. This contextual layer corresponds to an infrastructure placed between an organization’s record and data systems and its AI platforms. Its role is to convert business intelligence into a governed and usable framework for agents or models, while also promoting knowledge capitalization. Sankar summarizes the issue with an analogy: without this foundation, AI resembles an extremely bright intern, but blindfolded, being thrust into the company. She also argues that for every concrete intention and objective, a form of general intelligence already exists, while acknowledging its limited usefulness in current business applications.

Measured Gains: 16% Report Truly Transformative Value

According to a survey, all organizations questioned are already using AI to some extent. However, only 16% believe they derive transformative value from it. The largest share, 40%, places benefits in specific tasks or teams, at a level deemed moderate. 21% of respondents mention significant gains, while 23% observe little to no returns. Overall, this paints a picture of a diffuse deployment, where clear benefits remain in the minority.

Deployments Primarily in Testing or Support, Few Agents in Production

The current priority remains exploration: 38% of companies indicate they are testing use cases. 35% use AI to assist their employees through co-pilots or chat interfaces. Only 12% report AI agents in production, and 3% claim that AI performs the majority of work in key processes. Within organizations, several categories of tools already coexist, from customer platforms to conversational assistants and process automation. Despite their sophistication, their outputs can prove erroneous or unsuitable for the business context, a situation many companies face that could be among the reasons for their struggles to achieve returns on investment.