Overview: AI in Community Support

NeighborLink: An AI Community Resource Agent

This is a two-part series created by urban planning and urban technology students at University of Michigan for the Urban AI course. Part I is an overview of a topic of interest, and Part II is one or more replicable code tutorial(s).

(Part I) Overview: AI in Community Support
(Part II) Tutorial: NeighborLink - An AI Community Resource Agent

Context

As AI becomes more common in community public services, many communities are starting to use it in everyday support work. Common applications include organizing resident feedback, providing chat-based Q&A, and offering translation. These tools have made it easier for people to access information and find what they need.

Case Studies

We have found two example applications of this type of technology used in other contexts. Our first case study is of a company called Witlingo, a system that allows organizations to develop conversational AI systems for their specific needs and information from said organization. An example processing flow of this application goes through User Interaction → AI Interpretation → Resource Retrieval → Response Delivery.

Case Study #1: Witlingo Community Resource Agent

https://witlingo.com/

Witlingo develops conversational AI systems that allow organizations to communicate with and support large communities through automated chat and voice interfaces. These AI agents are designed to answer questions, guide users to relevant information, and connect people to resources without requiring direct human assistance for every interaction. Organizations often implement these agents on websites, messaging platforms, or voice assistants so community members can easily access information when they need it.

One common application of Witlingo’s technology is within nonprofit or membership-based organizations that manage events, volunteer opportunities, and community programs. In these environments, members frequently have similar questions about schedules, participation requirements, or registration processes. Rather than requiring staff members to respond individually to each request, the AI agent can automatically interpret incoming questions and provide accurate responses drawn from the organization’s information systems.

For example, a community member might ask a question such as “How do I sign up for the volunteer event this weekend?” The AI system analyzes the text of the message using natural language processing to determine the user’s intent. In this case, the system recognizes that the user is asking about event registration. The agent then searches the organization’s event listings or database and returns a response containing relevant details, such as event information and a registration link. This interaction allows the user to receive immediate assistance while reducing the administrative workload placed on staff. Over time, repeated interactions help organizations identify patterns in the types of questions community members ask most often. This information can be used to improve resource visibility, refine program communication, and better understand community needs.

Community Participation and Engagement

Beyond answering basic questions, AI community agents like those developed by Witlingo can support greater community participation. By lowering the barrier to accessing information, these systems make it easier for individuals to learn about opportunities to participate in events, programs, or local initiatives. For instance, if a user asks about upcoming volunteer opportunities, the AI agent can provide a list of available events, explain how to participate, and guide the user through the sign-up process. This type of interaction helps ensure that information about participation opportunities is available at any time, rather than only during office hours or through manual communication.

AI agents can also encourage engagement by directing users to surveys, feedback forms, or discussion platforms where community members can share their perspectives. In this way, the system acts as an intermediary between the organization and its community, helping facilitate ongoing communication and participation. By simplifying access to information and participation channels, conversational AI tools can help organizations create more responsive and accessible community environments.

---

Example Processing Flow Logic

User Interaction → AI Interpretation → Resource Retrieval → Response Delivery

1. User Input
   A community member submits a question through a chat interface or voice assistant.

2. Intent Identification
   Natural language processing analyzes the message to determine the purpose of the request (for example, event information, volunteer opportunities, or resource access).

3. Information Retrieval
   The system searches the organization’s knowledge base, program listings, or resource database for relevant information.

4. Response Generation
   The AI agent provides the user with the most relevant information, such as links, instructions, or event details.

5. Escalation if Needed
   If the request cannot be resolved automatically, the system directs the user to a human administrator or additional support channel.

Case Study #2: Chicago’s “Array of Things” | Community-Centered Urban AI

Instead of simply collecting data for city agencies, Chicago’s Array of Things (AoT) was designed so that residents, community groups, and researchers could also access the data and build solutions themselves. The project installed sensor nodes on streetlights across Chicago neighborhoods, collecting real-time environmental and activity data such as air quality and pollution, pedestrian and vehicle movement, temperature and heat island effects, noise levels, and flood risk indicators. AI models were then used to process these large streams of sensor data and identify patterns that humans could not easily detect in real time. Importantly, the system did not record faces or identities, which became a key design decision after community concerns about surveillance.

This approach produced several community-oriented outcomes. In transportation, researchers used AI models to analyze pedestrian and vehicle patterns at intersections, which helped identify dangerous crossings, streets with high near-miss collision patterns, and areas where signal timing needed adjustment. This gave the city evidence it could use to redesign intersections or modify traffic signals in order to reduce accidents. In environmental justice, AoT data helped address long-standing pollution burdens in neighborhoods on Chicago’s South and West sides. Community organizations were able to map air pollution hotspots, advocate for tree planting and green infrastructure, and push for changes to truck routes and industrial zoning. In this way, environmental discussions were strengthened by data-backed advocacy rather than relying only on anecdotal complaints.

AoT also supported heat resilience for vulnerable communities. Because urban heat islands disproportionately affect lower-income neighborhoods with less tree canopy, AI analysis of temperature data revealed block-level heat patterns that helped planners identify extreme heat pockets, prioritize cooling centers, and target tree planting programs. For older residents and people without air conditioning, this data directly informed public health responses. At the same time, because the data was publicly accessible, universities and civic tech groups were able to build new tools on top of it, including apps that visualize neighborhood environmental conditions, community dashboards tracking pollution and temperature, and urban planning research on walkability and transit. In effect, the city created an open urban data platform that local communities could actively build on.

A major reason this case worked better than many other smart city projects was that it was not purely top-down. AoT combined public data access, community advisory boards reviewing deployments, privacy-first sensor design, and local partnerships with universities and civic groups. As a result, the system functioned more as a form of civic infrastructure than as a surveillance tool or a corporate technology experiment. Overall, the benefit of this case is that it shows how urban AI can support safer streets, environmental justice, heat resilience, and local innovation at the same time, especially when data systems are designed to be open, privacy-conscious, and genuinely useful to communities rather than only to government agencies.

Issues

But most of these applications still focus on information rather than action. They help people understand services, but rarely help communities organize and coordinate their own resources.

This matters because everyday community life often depends on small, informal forms of support. Someone may want to borrow a tool, need short-term childcare, look for temporary help, or recruit volunteers for a local event. At the same time, other people in the community may have spare time, useful skills, or unused items, but no clear way to offer them.

In many cases, the problem is not that the right resources do not exist. The problem is that they are hard to identify, connect, and coordinate. As a result, small needs go unmet, useful resources stay unused, and communities become less able to support themselves over time.

This leads to the question we want to explore: can AI do more than provide information? Can it help communities better identify, connect, and coordinate the resources they already have?

Our Concept: an AI Community Resource Agent

Based on this problem, we propose an AI community resource agent that uses natural language processing to help communities identify and match underused community resources more efficiently. Its role is not simply to provide information, but to turn vague needs into a practical matching process.

For specified technical methods, please check our tutorial.

Motivation

Imagine a neighborhood where residents use a group chat to share second-hand items, ask for help, or offer resources. At first glance, this seems like a simple and effective form of community support. But in practice, these posts are easy to miss: people have to see them at exactly the right moment, and after a day or two the information quickly gets buried in newer messages. A similar problem exists on platforms like Facebook, where useful resources may already have been posted, but when someone actually needs them, they still have to search through old posts while the algorithm tends to prioritize recent content. In this sense, the challenge is often not that resources do not exist, but that they are difficult to retrieve at the right time. An AI agent could simplify this process by helping people surface, organize, and match relevant information more efficiently, rather than relying on manual scrolling and chance discovery.

Users

The main users of this platform include residents, local businesses, and community organizers. These groups are not fixed as either providers or seekers. In different situations, they can all act as both. Residents may offer time, skills, or items, while also asking for help such as childcare or borrowing tools. Local businesses may need temporary help, but they may also provide space, materials, or support. Community organizers may coordinate events and mobilize resources, while also relying on existing community support networks.

Why AI Matters Here: From Matching to Community Coordination and Diagnosis

AI’s role in this system is not only to clarify vague needs and match them with suitable resources. It also helps quantify informal community activities.

Through such an application, interactions around community resources can gradually form a data record. This makes it possible to diagnose how resources circulate within a community and identify common needs that recur among residents, local businesses, and community organizers.

At the same time, when users successfully match resources through the platform, these acts of mutual help are also recorded. Over time, this creates a way to observe patterns of self-organized activity within the community. Make it easier for community organizers to understand where local self-organization is active and where it is not, helping them identify areas that may require additional support or intervention.

Risks and Governance

Of course, this kind of AI community support system also comes with risks that cannot be ignored. These issues cannot be solved by technology alone. They also require deeper discussion around rules, governance, and regulation.

False information can weaken platform trust

AI is not good at determining whether user-submitted information is true. If people provide false information about available resources or needs, or use the platform for inappropriate or illegal purposes, the system’s matching results can quickly become unreliable. Without stronger verification rules and platform governance, trust in the platform may be weakened.

AI cannot replace human trust

Even if the platform successfully matches people with resources, trust-related problems do not disappear. Disputes, no-shows, or unclear responsibilities may still happen once people interact offline. Human intervention or clear institutional rules are still needed to protect both sides.

Privacy risks remain unavoidable

To make resource matching work well, the platform would likely need access to sensitive information, such as users’ locations and personal schedules. Although map-based features can reduce exposure by using blurred geographic data, the risk of data leakage or privacy misuse still remains. This makes privacy protection a central issue rather than a minor technical detail.