A top official within the Department of Veterans Affairs (VA) Office of the Chief Technology Officer (OCTO) said today that the VA is taking a more “cautious” approach when it comes to veteran-facing AI and is instead leveraging the tool internally to empower employees helping veterans.
“We’re being more cautious on the directly veteran-facing applications of AI,” VA OCTO Data and AI Health Product Lead Kaeli Yuen said during Washington Technology and Nextgov/FCW’s 2024 Federal Health Summit in Reston, Va., today. “The way that we are impacting veterans with AI at this point is by empowering the employees that are answering their questions.”
“We don’t have, to my knowledge, direct generative AI chatbots directly talking to veterans or anything like that,” she continued, adding, “We’re moving much more cautiously there until we can understand the risks better.”
Yuen added that despite not offering AI tech directly to vets, “everything we’re doing is ultimately veteran-focused, and even things we do for employee productivity, it’s so we can more quickly serve veterans.”
Yuen highlighted that VA health clinics have been leveraging AI tools for decades – particularly when it comes to risk prediction for suicide – but with the rise of generative AI in the past few years, the department has been able to accomplish even more.
“Now with generative AI being in the place where it is, a lot more doors have opened – largely in the realm of administrative burdens, clinical documentation burdens, and so forth,” she explained. “VA is now focusing on things like AI scribes, processing of community care documents from outside hospitals, and things like that that take up a lot of clinicians’ time and are not directly related to patient care.”
“This is tackling the problem of physicians being overwhelmed by documentation,” she continued, adding, “The way it works is you turn on the recording in the visit, and then the idea is that instead of typing to take notes, your physician is now 100 percent focusing on you instead of the computer, and then the AI tool will generate a draft note, which the physician then reviews and edits as appropriate before signing.”
When it comes to AI governance within the VA – which Yuen said is a large part of her job – the AI lead said that, at this current moment, the majority of the VA’s efforts are focused on getting a “baseline understanding of what AI is in use at VA today, and what the baselines are for some of the standards we’re interested in.”
“We’re focused on getting an inventory,” Yuen said. “The next step will be to assess all of those use cases in the inventory and get a baseline understanding of how they meet certain standards. And at this point, what I mean by standards is … do you have a way to monitor changes? Do you have a way to ensure that diverse data is used in the training? And so forth.”
“But exactly what the standards are for those things are a little bit still to be defined, so it’s more about transparency, getting a baseline, and taking the next steps from there,” Yuen said.
