MIT and Atlan: Why Businesses Struggle to Extract AI Value

MIT Technology Review Insights Poll Finds 16 Percent of Enterprises Extracting Transformative Value From AI Deployments

Across businesses today, multiple AI platforms, systems, and tools operate simultaneously to assist with customer interactions, conversational tasks, and workflow automation. According to an MIT Technology Review Insights poll reported by technologyreview.com, fewer than one in five organizations—specifically 16 percent—are extracting transformative value from these deployments.

The Tech TL;DR:

• Only 16 percent of surveyed enterprises report extracting transformative value from their current artificial intelligence deployments.
• The largest group of respondents, at 40 percent, see moderate gains in specific tasks, while 23 percent report minimal or zero returns.
• Prukalpa Sankar, co-founder of Atlan, points to the missing context layer between business systems and AI models as the primary barrier to realization.

Deployment Realities and Enterprise Returns Across Sectors

The technological capabilities of modern models stand in stark contrast to their actual enterprise utility. While 40 percent of survey respondents observe moderate gains in distinct tasks or team operations, 21 percent report meaningful gains. Another 23 percent indicate they are seeing minimal returns or none at all from their investments, according to the technologyreview.com report.

Adoption stages vary widely across enterprises. Companies are predominantly still piloting AI use cases, a phase cited by 38 percent of respondents. Another 35 percent deploy AI primarily to assist employees through co-pilots or chat assistants. Production-level deployment remains rare: just 12 percent of enterprises maintain AI agents in operational workflows, and only 3 percent utilize AI to handle the majority of work in key processes.

The Structural Missing Link in Business Automation

The primary hurdle limiting return on investment involves bridging foundational business infrastructure with advanced machine learning systems. Prukalpa Sankar, co-founder of Atlan, addressed this disconnect in the technologyreview.com report.

“There’s a big gap between what the models are capable of and how the models are actually being useful inside organizations. For every real-life intent and purpose, artificial general intelligence is already here. It’s just not really useful in most companies today.”

Sankar identified the missing element as an enterprise context layer. This architectural infrastructure sits directly between core business systems—such as systems of record and underlying data repositories—and external AI platforms. The context layer translates proprietary business intelligence into governed, machine-usable inputs for AI models, allowing organizational knowledge to compound effectively.

Without this specialized translation layer, enterprise AI deployments struggle to produce accurate actions. Sankar characterized the current implementation failure mode directly in the technologyreview.com report:

“Without a context layer for AI, it’s almost like you’re hiring the smartest intern in the world, blindfolding them, and throwing them into your organization.”