Enterprise AI Struggles to Deliver Value Without Proper Context Layer — Cloud and enterprise technology

Despite widespread adoption, fewer than one in five organizations are extracting transformative value from AI due to a lack of enterprise context. Experts argue that without an architectural layer connecting business systems to AI models, sophisticated tools often produce confused or incorrect outcomes.

Across modern businesses, numerous AI platforms, systems, and tools operate simultaneously, ranging from customer platforms and conversational assistants to workflow automation. While these individual components are sophisticated and responsive, their collective output often proves wrong, confused, or disconnected from the actual business reality. This disconnect is identified as a primary reason why many organizations struggle to realize returns on their AI investments.

Data from an MIT Technology Review Insights poll highlights this disparity. Although every surveyed organization uses AI to some degree, only 16% report extracting transformative value from their deployments. The majority of respondents, representing 40%, observe moderate gains in specific tasks or teams, while 21% note meaningful gains. Conversely, 23% of organizations are seeing minimal returns or none at all.

Current usage patterns indicate that most companies remain in the pilot phase or use AI primarily as an assistive tool. According to the survey, 38% of respondents are piloting AI use cases, and 35% utilize tools like co-pilots or chat assistants to support employees. Only 12% have deployed AI agents in production, and a mere 3% report using AI to handle most work within key processes.

Prukalpa Sankar, co-founder of Atlan, emphasizes the significant gap between model capabilities and their actual utility within organizations. He notes that while artificial general intelligence exists for real-life intents, it remains largely ineffective in most corporate environments due to a missing context layer. This layer serves as architectural infrastructure situated between business systems, such as data records, and AI platforms. Its function is to translate enterprise intelligence into governed, machine-usable context, enabling knowledge to compound.

Sankar illustrates the necessity of this infrastructure by comparing AI without a context layer to hiring the world's smartest intern, blindfolding them, and immediately placing them into an organization. Without understanding the specific business environment, even advanced models cannot deliver accurate or useful actions. This content was produced by Insights, the custom content arm of MIT Technology Review, in association with Atlan.

Sources: MIT Technology Review.