Agencies Tackle Data Challenges in Preparation for AI

Federal leaders teased more artificial intelligence development plans on the way as agencies increasingly look to the technology to help their operations while grappling with associated challenges around data management.

The State Department plans to release its new AI and data strategy early next year as the agency pushes forth its digital diplomacy and AI adoption plan globally.

“AI is becoming an underpinning of every multilateral negotiation and engagement across the world,” said State Chief Data and AI Officer Matthew Graviss at an industry event earlier this month.

At State, diplomats are using AI and available open-source models to translate and summarize daily news alerts and prepare congressional reports respectively. The open-source model acts as a research assistant to build reports about the agency’s 270 global missions and would save employees time when completing several reports.

“We’re talking about savings at 60,000 hours across the globe by automating a lot of that synthesizing,” said Graviss.

The department is looking to AI to support its workforce and create workload efficiencies. The department launched in August an unclassified generative AI chatbot that supports 10,000 users, with potentially more on the way.

“We are data people first. Prompts and responses are data, which means you can analyze, and you can know who’s using it and how,” said Graviss. “We’re learning [how employees are using AI], packaging that up into training and communications programs, and sending it off for everybody else to start adopting.”

The Air Force Research Laboratory also released a chatbot it calls Non-classified Internet Protocol Generative Pre-training Transformer (NIPRGPT) to “alleviate toil” and allow employees to focus on the mission, said AI Lead Amanda Bullock.

“If I’m doing an administrative task, that is taking away my efforts to help the warfighter, that is toil,” said Bullock at the event. “We really want to leverage AI and data to tackle that … and not spend the time doing things that I don’t think anyone in the federal government loves doing.”

One of the biggest focuses for agencies in developing AI tools has to do with data storage and managing the data lifecycle.

“The planning factor is a big thing we’re focused on. The storage costs are going to eat us alive if we don’t make better informed decisions about when we can keep data,” said Intelligence Community CDO Lori Wade at the event.

Wade advised the IC to automate data management processes in a June 2024 directive and said she would soon release a data reference architecture as part of that strategy. The release was shared with private sector partners prior to the public release.

“It’s a CDO and a CIO document and architecture … that will show how we’re going to automate all of this and how we bring down data silos. It’s based on data mesh principles across a distributed ecosystem,” said Wade.

Similarly, Customs and Border Protection (CBP) is prioritizing data tagging and inventorying — which includes noting whether data is on prem, in the cloud or “underneath someone’s desk,” said CBP CDO Michelle Lebowski. “If we don’t know exactly where our data is now, then we’re going to be missing things and potentially some very important things.”