Partnership drives early implementation of generative AI

With growing interest in the transformative possibilities of generative AI since ChatGPT's public release, there is an urgent need for responsible, secure application of this technology in the federal space. Accenture Federal Services has implemented Azure OpenAI in diverse ways to interact with federal data and workflows, solve practical problems, deliver unique insights and create near-immediate value.

One of the more interesting outcomes of these early federal implementations is the unprecedented time-to-value ratio we’ve been able to achieve. Once a generative AI use case is clearly defined, and responsible AI guardrails and processes are established, an otherwise unwieldy challenge can be solved much faster.

Transforming massive global data holdings

One of our longest-running Azure OpenAI implementations is within our Next-gen Mission Analytics (NMA) capability, deployed in production to support ongoing information missions around the world. The use of Azure OpenAI has lowered the barrier of entry for analysts, sped up the filtering, summarization and report writing process, and transformed massive global data holdings into a navigable trove of analytical findings.

In support of Combatant Command (COCOM) missions, our teams are implementing the Azure OpenAI API and developing LLM-powered features to accelerate and augment analyst workflows. The models enable subject matter experts and new platform users to navigate global data holdings more easily, as well as to shape, summarize and glean insights from custom datasets. Our human-in-the-loop approach constrains model inputs and validates model outputs to reduce hallucinations and manage and mitigate the impact of any model hallucinations.

Optimizing natural language processing

We also integrated Azure OpenAI to implement, optimize and deliver an analyst-based generative AI chatbot. Using natural language prompts, analysts can rapidly generate complex multi-lingual, mission-specific Boolean queries. These queries automatically create datasets from extensive NMA open-source data feeds to expedite time-sensitive analysis.

Additionally, we further implemented Azure OpenAI to create a data service that clusters, summarizes and synthesizes approximately 10,000 articles for expert review in the span of 20 minutes. The result is an analyst-ready draft that saves significant analyst labor. This allows analysts to focus their limited time on high-value analysis and related tasks – making sense of summarized data instead of time-consuming, laborious efforts to create summaries.

For another federal client, we are helping analysts interact dynamically with data against a Databricks backend, using natural language interaction with Azure OpenAI models. We are optimizing and engineering model interactions to generate efficient, accurate outputs that align with our responsible AI approach.

Increasing efficiency with assistive coding

Assistive coding continues to be a top-of-mind need for federal agencies. Partnering with a federal client, we are using Microsoft generative AI technology to make software development more efficient, using GitHub Copilot for Code Assistance. While not a replacement for software developers, the capability is providing measurable efficiency improvements and reducing the learning curve for less experienced staff. Partnering with Microsoft to deliver this capability will provide access to new code-conversant GPT models over time and we anticipate direct performance uplift in development workflows.

Rapid prototyping for urgent, unexpected needs

In the spirit of exploration and innovation, we stood up an Azure OpenAI sandbox and hosted a company-wide generative AI hackathon. With safety and responsible AI guardrails in place, and using open data, teams without extensive AI expertise built prototype capabilities – several of which were mature enough to be taken directly to clients. While more complex, data-rich use cases require practitioners skilled in design, human factors, data science and data engineering, Azure OpenAI continues to provide a platform for anyone to convert ideas to reality.

These examples illustrate how our teams collaborate with federal clients to incorporate generative AI in novel, responsible ways. Now that the White House AI Executive Order has clarified the importance of responsible, trustworthy and secure exploration of AI for federal agencies, we are working with a growing number of clients seeking to leverage generative AI capabilities – and we continue to partner and innovate with Microsoft to solve the federal government’s most pressing challenges.

----------

Discover other areas generative AI will impact federal agencies in my perspective, “Generative AI for federal agencies: Five focus areas”.