When it comes to implementing GenAI solutions in the federal government space, the most common refrain we hear from department leaders is, “This all sounds great, but where should I start to drive maximum impact as quickly as possible?”  

Though use cases for defense and intelligence abound, the real challenge is how to take a GenAI solution and mold it to your specific needs.

To help you draw inspiration from a real-life example, this Q&A focuses largely on how the Digital Transformation Office of the United States Air Force teamed up with Pryon to build DTO Wingman, a chatbot and report generator that serves both military personnel and civilians. You’ll also learn how federal organizations can get started with GenAI quickly and what kinds of content federal agencies tend to feed their RAG-based applications.

‍

QUESTION

How did the engagement with the DTO (Digital Transformation Office) begin?

Working with the Digital Transformation Office has been one of my most valued engagements at Pryon as the Head of Federal Solutions. This project marked an important milestone as Pryon's first implementation with a federal client and our initial collaboration with the Department of Defense (DoD). It is set to be our first client at ATO Level IL5.

The engagement commenced late last year when the Digital Transformation Office sought to leverage technology in a unique way. The goal was to enable users to access information directly from their website. Additionally, the project aims to incorporate publications from Air Force e-pubs, allowing users to efficiently query this content.

‍

QUESTION

What challenges came up while building the DTO Wingman?

‍

The first challenge we faced was that Pryon wasn't equipped to deploy directly to websites. Our platform connects data internally, allowing access to information without building chatbots or widgets, which the DTO didn't want to handle themselves. We decided to develop a chatbot that integrated with our system, marking our first chatbot built and hosted chatbot on the DTO website.

As the collection expanded, DTO realized the utility of adding more information, leading to one of Pryon's largest collections. Ensuring accurate and responsive information retrieval was crucial.

Another significant challenge was refining our generative technology, which was not fully mature. Initially, we relied on basic orchestration models. Our first chatbot deployment utilized a straightforward interface for Q&A using generative APIs with simple orchestration and attribution. This setup caused considerable latency due to numerous integrations across the full stack from user inquiries to our platform and back.

The system lacked robustness, often providing correct but insufficient answers, which left users dissatisfied. The absence of streaming capabilities further exacerbated latency issues, impacting user experience and perceived value. These challenges led to iterative improvements, resulting in the enhanced user experience now available on the website.

‍

QUESTION

Why did the DTO decide to go beyond just a chatbot?

Wingman has evolved from being just a chatbot to becoming a comprehensive ecosystem of tools and features. It's still connected to the same data source but has transformed to offer users a more tailored experience to answer their questions effectively.

Users can now interact in different languages and receive generative outputs that are both detailed and controlled in terms of creativity, leveraging the generative aspect while staying anchored to the foundational truth within Pryon.

We continue to expand features and capabilities. Recently, we introduced a report generator, allowing users to export content into Word documents or PowerPoints for sharing. An email formatter is now available, enabling users to craft emails based on specific publications. Additionally, we have integrated voice-to-text capabilities.

QUESTION

How do you see the deployment of DTO Wingman evolving? 

The fundamental technology remains focused on information retrieval, but the initial step is providing access to knowledge in a valuable format. But I think that's step one. Getting people access to the knowledge, but [also] giving it to the way that is valuable is the next piece. That's what we've really started to solve with Wingman and we're seeing [this use case] explode across the DoD.

We are eager to collaborate over the next year, leveraging our technology to push boundaries. This includes our first implementation in the DoD, entry into IL5, and initial chatbot instance. Our partnership with the Digital Transformation Office aims to expand technology use and deliver increased value to them and the broader DoD community.

QUESTION

What are some examples of GenAI deployments that government organizations can implement relatively quickly? 

Here are some use cases where GenAI solutions have provided immediate value:

1. Policy Analysis: Federal agencies and DoD organizations often deal with extensive policy documents, such as cybersecurity policies or risk management frameworks. GenAI can be swiftly deployed to analyze these documents, allowing users to pose questions and receive answers or summaries efficiently.

2. Technical Manuals: Agencies frequently possess technical manuals, like those for helicopter repairs or policy guidelines. GenAI can expedite the process of finding answers within these manuals, enhancing efficiency for entities like CBP or USCIS by integrating AI-powered chatbots for rapid information retrieval.

3. Research Applications: Organizations such as the FBI utilize GenAI for sifting through extensive case files and histories. This technology enables efficient summarization and pinpointing of crucial information within vast datasets, essentially locating a needle in a haystack.

‍

‍

READ MORE: Generative AI for Government: Practical Applications and Implementation Guide 

‍

QUESTION

How have you adapted Pryon RAG Suite to meet the needs of specific federal organizations?

As the head of federal solutions, I've had the unique opportunity to work with various federal organizations to leverage Pryon as a platform to solve numerous challenges. A primary trend across these organizations is the need to access their unstructured data more swiftly and effectively.

For the first time, technologies like generative AI and RAG systems are enabling us to extract information, draw insights, and interact with data conversationally. A recent application with the IRS involved managing a vast array of policy documents, both internal and external. The solution focused on customer support agents who needed to quickly access this repository of policy data to provide accurate, source-attributed answers to policy-related inquiries.

Another notable use case was with the Department of Air Force Digital Transformation Office (DAF DTO). They required a public-facing solution that allowed users to access a comprehensive collection of public Air Force publications, DoD instructions, and federal regulations. We implemented an AI-driven bot on their website, enabling users to swiftly obtain information from these diverse sources. Users can generate reports or long-form summaries, with each element of information directly attributed to its source document.

This deployment exemplifies Pryon's RAG technology, helping federal clients access critical information quickly to focus on strategic priorities. We're eager to support more clients and expand the range of use cases for this transformative technology.

‍

‍

Learn More

Eager for more information about implementing AI at the federal government level? Check out our guide to GenAI for Federal Government and watch more interviews with members of Pryon’s federal team for insights into how to overcome common objections, what the government can learn from the private sector, how to acquire funding, and more.