The Massachusetts Federal Funds and Infrastructure Office (FFIO) estimates that Massachusetts communities are eligible for approximately $17.5 billion in federal grant funding, yet less than 30% of that is successfully accessed by many local governments. 

These funds are intended to address urgent needs such as repairing roads, replacing lead pipes, expanding broadband, and strengthening climate resilience. When this funding goes untapped, communities miss critical opportunities to invest in their infrastructure and well-being. 

When I joined AI for Impact, I thought I would be doing what engineers are trained to do: define a problem, build a solution, and improve it over time. 

In some ways, that is what happened. But over the past year, working on GrantWell taught me that building technology for the public sector requires something more fundamental than the design, architecture, or model selection. It requires listening carefully enough to understand the real problem in the first place.

Understanding the Barriers

As AI for Impact Fellows, we initially approached this as a writing challenge. I thought we were building an AI tool that would help local governments, especially under-resourced municipalities, write grant applications more efficiently.

The assumption seemed straightforward: if municipalities lacked staff capacity, generative AI could help them produce compelling grant narratives more quickly.

The assumption seemed straightforward: if municipalities lacked staff capacity, generative AI could help them produce compelling grant narratives more quickly. Our assumptions lasted until we started talking to municipalities.

Last spring, my teammate Jai Surya and I attended several of FFIO’s statewide roadshows, where municipal leaders, local officials, and community organizations gathered to discuss the challenges of pursuing funding opportunities. 

From Haverhill to Barnstable, I heard the same frustration again and again. Writing was only part of the challenge. Municipal officials struggled to understand the requirements buried in lengthy, technical funding calls. Even identifying relevant opportunities could be difficult.

Source: FFIO September 2025 City Listening Tour. Data prepared by the Edward J. Collins, Jr. Center for Public Management in partnership with the Healey-Driscoll Administration’s Federal Funds & Infrastructure Office.

At one session, a local official described spending hours reading a federal grant announcement only to realize near the end that their town was not eligible. That example stayed with me. It captured something I had not fully understood when we began: for many smaller communities, the biggest burden comes before the writing even starts. Without dedicated grant writers, many municipalities simply do not have the time or capacity to sort through dozens of complex opportunities and decide which ones are even worth pursuing.

In practice, much of the effort goes into simply determining whether an opportunity is relevant, feasible, and worth the time investment. For many smaller municipalities, that barrier alone can stop an application before it even begins. As a result, funding opportunities are often passed over, not because communities do not need them, but because they lack the capacity to evaluate them in time.

A different design question

Listening to communities shifted our design question. Instead of using AI solely to draft grant narratives, we began examining how technology could reduce friction across the entire grant discovery and application process.

That shift in thinking changed the product. Our first version had focused mainly on drafting support through a chatbot. But after hearing from municipalities, it became clear that drafting could not be the starting point. We needed to help users find opportunities, understand requirements, and locate critical details before they ever wrote.

GrantWell evolved through that feedback. Repeated conversations with municipalities, state partners, and technical assistance providers, along with testing with people who regularly work on grant applications, directly shaped the interface, summaries, chatbot experience, and drafting workflows.

The platform improved when we stopped designing around assumptions and started designing around how communities actually navigate grant processes.

The platform improved when we stopped designing around assumptions and started designing around how communities actually navigate grant processes.

GrantWell is the result. 

Grantwell helps grant writers find, understand, and act on funding opportunities more easily.

 Grant applicants in Massachusetts can use the free platform to:

• Find relevant funding opportunities

• Understand complex grant requirements

• Draft structured responses efficiently

Today, GrantWell combines several capabilities that address the challenges communities described:

1. The platform aggregates federal and state funding opportunities into a structured, searchable database so users do not need to navigate multiple portals.

2. It generates concise NOFO summaries that highlight key information such as eligibility criteria, required documents, deadlines, and evaluation factors.

3. A document-grounded chatbot allows users to ask questions about a specific opportunity and receive answers tied directly to the source document, helping teams verify details without scanning long PDFs.

4. The system includes a guided drafting workflow that walks users through a series of structured questions aligned to the selected grant and produces a draft grant narrative that staff can review and refine.

From “find a grant” to “fund a need.”

One insight from early testing was that many communities do not start with a grant in mind. They start with a need.

A town might be trying to upgrade wastewater infrastructure, expand rural broadband, or launch a workforce development program. Connecting those priorities to the right federal opportunities often requires time and specialized knowledge.

GrantWell is evolving toward a needs-first model. Users will be able to describe a project in plain language, and the system will surface relevant funding opportunities from the grants database.

GrantWell is evolving toward a needs-first model. Users will be able to describe a project in plain language, and the system will surface relevant funding opportunities from the grants database. Instead of beginning with a grant announcement, communities can begin with the problem they are trying to solve.

Responsible AI and secure infrastructure

Security and the responsible use of AI were central considerations in the system's design.

Another thing I learned early on was that usefulness alone was not enough. Many people were understandably skeptical about using AI in government workflows, especially when privacy and data handling were involved. In conversations about the platform, I often found myself explaining not just what the system did, but how it worked and where the data went.

GrantWell is deployed in the participating state's cloud environment, and model interactions are routed through AWS Bedrock. This architecture ensures that prompts and responses are not used to retrain external models and that project data remains within state infrastructure.

 In public-sector AI, trust is inseparable from the product experience. It is part of the product.

That technical design became an important part of trust. Communities can explore funding opportunities and draft responses while maintaining control over their data and meeting government security expectations. For me, this reinforced the idea that, in public-sector AI, trust is inseparable from the product experience. It is part of the product.

A broader lesson for public sector AI

Federal funding programs are meant to serve communities nationwide, yet access to those resources is uneven. Larger municipalities often have dedicated grant staff who can track opportunities and interpret complex requirements, while smaller communities frequently do not.

GrantWell is currently being piloted in Massachusetts and continues to evolve based on user feedback. 

Over time, platforms like this may provide a blueprint for other states seeking to expand equitable access to funding opportunities while maintaining transparency, accountability, and human oversight.

Related Materials

To learn more about GrantWell and the work behind it:

• Watch the InnovateUS workshop, "Introducing Grantwell: Using AI for Municipal Technical Assistance."
• Read the Massachusetts Governor's Office Press Release
• Explore the screencast demo.
• Try the tool: https://grantwell.virtualassistance.mass.gov/