Creating AI value with enterprise context

This piece comes from MIT Technology Review Insights, the publication's custom content arm, not its newsroom. That matters for how much weight it gets. But the poll numbers are worth knowing: every organization surveyed uses AI in some way, yet only 16% report transformative value. 40% see moderate gains in specific tasks or teams, 21% meaningful gains, and 23% minimal or no return.

Adoption is also shallower than the hype suggests. Only 12% have AI agents in production, and just 3% say AI does most of the work in key processes.

The numbers line up with what we hear across the industry. The models are capable. Getting them to understand a specific business, its data, its rules, its history, is the hard part. A co-founder of Atlan, a data company, puts it as the gap between what models can do and how useful they are inside organizations, and compares an AI with no business context to a brilliant intern dropped in blindfolded.

We think that framing is basically right. A lot of disappointing AI deployments are really data and governance problems wearing an AI label. A shared, governed layer that tells agents what the business actually means by a customer, a contract, or a margin could make answers more accurate and cheaper to produce.

But we'd keep our guard up. This is sponsored content, and the person diagnosing the problem builds products in the category being recommended. The poll's methodology isn't laid out in the excerpt we reviewed. And a 'company brain' that pipes everything into AI systems raises its own risks: access control mistakes, stale or wrong context spreading faster, and more sensitive data exposed to more tools.

Our read: the value gap is real and context is a reasonable place to look. Treat the vendor framing as one hypothesis, not the answer.

A context layer, as described here, sits between a company's systems of record and its AI tools, turning internal knowledge into governed, machine-usable context for models and agents.