+ THREE INSIGHTS FOR THE WEEK |
1. Giving artificial intelligence access to your organization’s data isn’t the same as giving it the context to understand that data. That’s why researchers at the MIT Center for Information Systems Research argue for investing in a “semantic layer” to get real value from AI.
A semantic layer sits between an organization’s data and the people or machines using it, to explain what the data means, how different elements relate to one another, and which rules govern its use. Without the business context that gives data meaning, AI systems can produce answers that are technically plausible but also incomplete, misleading, or wrong.
The researchers recommend three steps for building a semantic layer:
|
-
Start with priority data assets. Rather than attempting to describe and organize every piece of enterprise data at once, leaders should identify the data needed for their highest-priority AI initiatives.
- Govern the semantic layer itself, with a clear owner accountable for its quality and results.
-
Use AI to help create and maintain metadata. AI can support tasks such as cleaning and classifying data, recommending access controls, and identifying connections among data assets.
|
|
|
|
2. A profound shift is reshaping employment culture in the U.S. — a change that MIT Sloan School of Management professor emeritus Paul Osterman says is happening at the expense of a large swath of the U.S. workforce: Increasingly, employers are bypassing long-term, mutually beneficial relationships with employees in favor of short-term contract work.
Drawing on original survey data from more than 6,000 civilian, nonagricultural working adults, Osterman found that the strategy already accounts for a large fraction of the job market and, he predicts, will become increasingly widespread.
In an excerpt from his new book, “Disposable Workers: The Transformation of Employment,” Osterman details the benefits that have accrued to standard employees since the end of World War II — and explores why gig workers, contractors, and others are left out of that bargain.
|
|
|
|
3. When Silicon Valley Bank failed in March 2023, it was the 16th-largest bank in the United States and held $208 billion in assets. Studies after the fact typically focused on the managerial and supervisory problems in the year leading up to SVB’s failure.
A study from the MIT Golub Center for Finance and Policy takes a longer historical view, with conclusions that are just as damning. “For its entire 15-year life as a regional bank, SVB held the same risky bet,” writes GCFP researcher Niccolo Comati, who co-wrote the study with Eric Rosengren. “The risks were visible the whole time, yet supervisors reacted only once losses had materialized.”
Specifically, the authors found:
|
- Between 2009 and 2023, SVB’s assets grew nearly twentyfold while its business model remained mostly unchanged.
- The bank’s deposits came almost entirely from startups, were largely uninsured, and paid little or no interest.
-
Those deposits were invested in long-dated securities highly sensitive to interest rates from the outset, a sensitivity the bank increasingly tried to keep out of its financial reporting.
- Regulators and supervisors had repeated occasions to intervene but acted only once the losses had materialized, a situation the authors characterize as “an endemic problem with how bank supervision is conducted in the United States.”
|
|
|
|
Can AI’s climate benefits outweigh its costs?
|
The dilemma: The rising energy demands of AI and data centers threaten corporate net zero commitments and worsen climate change — but AI can also help develop climate solutions.
|
|
|
|
Will artificial intelligence ruin the planet or save it? MIT Sloan professor John Sterman and Koru Labs president Jennifer Turliuk, a 2025 graduate of the MIT Sloan Fellows program, explore the question in a new paper that offers a framework for evaluating the climate impact of AI.
Some of their findings:
|
- AI and data centers cause roughly 0.5% of global CO₂ emissions today — a small but growing share that worsens climate change.
- AI might help develop technologies to reduce emissions, but even promising innovations will take years to move from the lab to real-world scale.
-
The bigger issue, the researchers say, is indirect: A modest AI-driven boost to economic growth could raise projected warming far more than AI’s own emissions — what’s known as an “indirect rebound effect.”
|
“Unless we implement policies and actions that rapidly lower emissions across the economy as a whole, any economic boost from AI will generate a lot more emissions, and the climate will get notably worse,” Sterman said.
The researchers have embedded the framework into En-ROADS, an interactive climate simulator available to the general public. The new features allow En-ROADS users to model for themselves the direct and indirect impacts of AI on energy use, greenhouse gas emissions, and climate change.
|
|
|
|
minimum viable governance
|
|
|
|
|
One Main Street 9th Floor | Cambridge, MA 02139 US
|
|
|
|
|
