How Boston Globe Media is putting AI to work in its newsrooms

A podcast recap: Building for Others with Shira Center and Sidhi Dhanda from Boston Globe Media

In this episode of Building for Others, we sit down with Shira Center (VP, Innovation and Strategic Initiatives) and Sidhi Dhanda (AI intern) to explore four AI use cases they built on elvex that are starting to change how a 150-year-old media company operates.

Four use cases that actually work

School board monitoring at scale

One of the most compelling examples comes from Sidhi, who built a system to monitor school board meetings across nearly 100 Massachusetts districts. The workflow is straightforward: meeting transcripts are uploaded to Google Drive, connected to an elvex assistant, and journalists can query the material directly in Slack.

School committee meetings are often lengthy and routine—and many won’t yield a story. But being able to scan across a large number of towns makes it easier to spot themes and early signals that would be hard to catch otherwise. Sidhi used the tool to help identify towns discussing AI literacy for an editorial—an example of how broader monitoring can support reporting and commentary.

Natural language archive search

The Globe’s second use case addresses a familiar challenge for media organizations: finding relevant material across decades of content and multiple systems. Their solution allows reporters to search roughly 25 years of archives in natural language, again through Slack.

Examples include: “What was the first public comment Michelle Wu made when running for city council?” or “What was the largest snowfall in the last 25 years, and how does it compare to recent storms?” For Boston-focused questions, this can be a faster starting point than conventional search—especially when reporters are looking to quickly orient themselves or validate facts before deeper reporting.

Headline generation with style

A third tool supports opinion headline drafting while aligning with the Globe’s style guidelines. Writers paste an article into Slack and receive multiple headline options, built to follow internal rules on capitalization, structure, and tone—while explicitly avoiding clickbait.

In practice, staff often treat the output as raw material: mixing and matching elements from different suggestions to get to a final headline faster, without outsourcing editorial judgment.

Alt text that people use

The final use case is less flashy, but widely used: an alt text generator that speeds up photo description writing. Drop an image into Slack, and it returns properly formatted alt text aligned with accessibility guidance (and useful for SEO).

It doesn’t change the substance of journalism, but it reduces friction on a task that has to get done—and that’s exactly why it’s been adopted.

Why these tools actually get adopted

Across the four examples, the throughline is implementation: the tools live in Slack, where Globe teams already work. That reduces friction—no separate logins, minimal workflow change, and less reliance on formal training to get started. As Shira puts it: “The best way to introduce these tools is to build on other tools, and meet people where they already are. It’s habit stacking, plain and simple.”

AI’s role in media: a four-part framework

Shira frames AI’s potential value in four areas:

• Newsroom tools: speeding up tasks like document review and basic synthesis.
• Reader experience: using data to better understand and serve subscribers.
• Upskilling: helping staff build skills that will matter as tools evolve.
• Business considerations: navigating issues like licensing, competition, and workflow efficiency.

A consistent emphasis is governance—moving quickly where it makes sense, while protecting the organization’s core asset: trust. “A human has put this content out there and has reviewed it,” Shira notes. “That is still incredibly important to us.”

Advice for builders: the learning curve is flat

The episode's most quotable insight comes from Shira: "The learning curve is flat." Even if you feel behind in AI adoption, you can catch up quickly because the fundamentals remain constant, and the ability to learn is faster than ever before. 

Sidhi's advice focuses on practical steps: Lower barriers to entry by building where people already work. Experiment broadly by trying different use cases and areas. Iterate relentlessly and fail fast to find what works. Show peers what's possible, because demonstrations beat training sessions.

Taken together, the approach isn’t about having the most advanced models. It’s about picking specific problems, integrating tightly into daily workflows, and letting usefulness—not mandates—drive adoption.

The bigger picture

Local media continues to face real constraints: fewer reporters, tighter budgets, and business-model shifts. The Globe’s AI work doesn’t solve those challenges outright, but it points to a pragmatic way to make limited capacity go further in a few areas—monitoring more sources, searching institutional archives more quickly, and reducing time spent on routine production tasks.

The goal isn’t replacing journalism; it’s making certain parts of the work easier so reporters and editors can spend more time on the parts that require judgment.