How Outlier Media built AI into a text platform designed for trust
In July 2023, the American Journalism Project launched its Product & AI Studio to explore the smart application of generative AI and other technology within local news. This Studio, made possible by OpenAI with additional support from the Patrick J. McGovern Foundation, is sharing regular updates of its learnings, including this update from Outlier Media.
In 2016, Outlier Media was created with the sole objective of giving Detroit residents the information they needed quickly and directly. Social media, mobile apps and newsletters were on the rise, making the news landscape more dynamic and offering digital audiences more choices than ever
But there were still acute information needs — particularly for people living in lower-income communities. Sarah Alvarez, the founder of Outlier Media, saw an opportunity to fill those gaps in non-traditional ways. She created the TXT OUTLIER service as a way to deliver valuable and actionable information via text message to meet Detroiters’ daily challenges, such as “eviction, navigating property tax foreclosure or avoiding utility shut-offs.” The platform also allows users to talk directly to a reporter.
Outlier started with a staff of two and has grown to a full-fledged newsroom of 16 with a $2.8 million budget. In 2021, when Outlier first joined the AJP portfolio, it primarily served Detroit residents through a third-party SMS system, along with occasional long-form accountability and investigative stories published by media partners. Outlier strengthened its ability to understand and directly serve the needs of more Detroiters by expanding its editorial capacity and adopting innovative approaches to community engagement. This work included growing through acquisitions and the development of new newsletter products, integrating the Detroit Documenters program to enhance civic engagement and local government reporting, and rebuilding its website to function as a comprehensive news and information hub.
Community response verified that TXT OUTLIER filled a need. And as the demand grew, Outlier’s reporters sometimes struggled to keep up with and be responsive to the growing numbers of queries.
In September, Outlier Media announced the relaunch and expansion of TXT OUTLIER — a more intuitive system that’s still using content written by journalists, but that also leverages artificial intelligence for select functions, including facilitating property lookups in its database, assisting reporters with research and summarizing conversation trends for the newsroom. This new SMS system allowed Outlier to address more individual questions and share timely information with a broader audience while fostering long-term, trust-based relationships with the community.
With its improved platform, Outlier can now send weekly messages to over 15,000 Detroiters chosen from its list of more than 700,000 contacts. The publication can segment its audience to ensure more regular communication with the most engaged users, while also reaching those it hasn’t connected with recently. This segmented approach is key to building lasting relationships with the community and keeping Detroiters informed, Alvarez said.
The custom-built system was developed in partnership with engineering and design workshop Public Data Works and included AI integrations, making the platform faster, more efficient, and better-equipped to handle the information needs of Detroiters. Outlier Media has transformed how it connects with its audience, offering an even more personalized and responsive service. Alvarez and Kate Abbey-Lambertz, Outlier Media’s product director, recently shared their technology transformation journey with the AJP Product and AI Studio grantee community.
Streamlined and scalable communication
One of the most impactful changes has been the upgrade to TXT OUTLIER’s backend communication systems. Anyone can text DETROIT to 67485* and select from a menu of options, or they can ask a question or enter any Detroit address to get back a short report with property information. Responses to common inquiries are pulled from a database. There is also an option to speak to a reporter.
With the old system, messages would occasionally get lost and there was no ability to segment or target from the massive database of phone numbers they collected. If a new housing law was taking effect, for example, there was no possibility of targeting users who had submitted housing questions.
The old system also didn’t allow Outlier to get visibility into user habits. They often couldn’t tell if users were returning to the platform, had questions about multiple topics or mostly stuck to one issue.
“We [also] needed help with internal workflows,” Alvarez explained. “We also really wanted some help with the analytics. We have a very big list. But we know very little about any of those people because of the systems we have been using in the past.”
The new system integrates powerful tools including Missive, Twilio, OpenAI’s GPT-4.o mini and LlamaIndex hosted on Amazon Web Services.
These new integrations don’t just make it easier for Outlier’s team to respond to individual inquiries; they give them options to scale and better track and manage user conversations over time. For example, there were little to no user analytics available on the old system, making it hard to convince sponsors they were reaching real people. With the new system, “Our development and sponsorship teams love the analytic summaries, and we just did our first sponsored text message,” noted Abbey-Lambertz.
AI-powered responses and a more personalized approach
By integrating AI through OpenAI’s API and LlamaIndex, TXT OUTLIER now handles user inquiries in ways that were previously out of reach. “The AI integrations we’re using… look through the entire history of back and forth between the system and a texter, between a reporter and a texter, and tell us what’s happening,” said Alvarez. AI currently assists with property lookups, summarizing conversations, and analytics reporting, helping Outlier’s team stay up-to-date on each user’s needs and respond more thoughtfully.
“We’ve been able to engineer prompts that let us track changes in these conversations and keep summaries accurate and up-to-date,” Alvarez added about the analytics.
TXT OUTLIER’s framework allows Outlier to continue building functionality by expanding content and data sets so users can ask about neighborhood-specific news or city services and receive quicker, more precise responses.
As the team develops more advanced segmentation capabilities, they plan to send specific information to users based on their unique interests, such as housing news or community activities.
Strengthening community trust
Building and maintaining trust is essential to Outlier’s mission; the outlet has shown that by responding to the information needs of community members, it’s possible to build trust in communities — even with groups who have had a longstanding distrust of the media.
The updates to TXT OUTLIER, including the integration of AI, have strengthened this commitment, Alvarez explained. Even small design choices, like those differentiating Outlier’s own responses from publicly-sourced responses, are geared toward clarity and trustworthiness, she said.
The hope is that people will donate to support the service if they find it valuable, Alvarez said. But for the moment they are being very thoughtful about re-engagement, broadcasting messages and building that trust.
Looking ahead
Looking forward, Outlier’s next steps include rolling out features to enable more granular targeting and information sharing.
“We want to eventually deliver specific information based on users’ interests,” explained Abbey-Lambertz, “like letting users sign up for housing or event updates.”
With these updates, TXT OUTLIER is poised to improve support of Detroit residents, providing crucial information and direct communication in a time when reliable local news is more important than ever. Whether it’s staying on top of housing issues, connecting with local events, or turning to a trusted source in a time of need, the Outlier team hopes these improvements position them as a stronger, more responsive ally in Detroiters’ daily lives.
Interested in learning more about the tech? Outlier’s shared it all on GitHub.
—
*(Message frequency will vary. Message and data rates may apply. Reply “help” for assistance or “stop” to cancel. Please read Outlier’s terms of service and privacy policy.)
