))
Federal Tech Podcast: for innovators, entrepreneurs, and CEOs who want to increase reach and improve brand awareness
«
»
Ep. 279 Avoiding Agentic AI Pitfalls in Federal Digital Transformation
Manage episode 517637754 series 3610832
Вміст надано The Oakmont Group and John Gilroy. Весь вміст подкастів, включаючи епізоди, графіку та описи подкастів, завантажується та надається безпосередньо компанією The Oakmont Group and John Gilroy або його партнером по платформі подкастів. Якщо ви вважаєте, що хтось використовує ваш захищений авторським правом твір без вашого дозволу, ви можете виконати процедуру, описану тут https://uk.player.fm/legal.
A recent study from Carnegie Mellon University is titled "AI Agents Fail at Office Tasks Nearly 70% of the Time."
Federal agencies are adopting Agentic AI for the efficiency it can deliver. Unfortunately, many do not realize that Agentic AI is prone to operational risks, ranging from technical glitches to legal complications to accidental database deletion.
When Agentic AI causes problems at a federal agency, there can be lives at stake.
Today, we sat down with Travis Rosiek, Rubrik's Public Sector Chief Technology Officer. During the interview, he explores the federal challenges of implementing Agentic AI, building an Agentic AI inventory, and making Agentic AI visible, auditable, and reversible.
CHALLENGES
Everyone —from a systems administrator to an agency administrator —knows that data must be backed up. However, very few understand that Agentic AI is a collection of agents that can be attacked, just like a database.
Rubrik offers the capability to reassure users that Agentic AI can be reversed if malicious actors enter the picture.
STARTING POINT
Most cybersecurity professionals agree that one starts by understanding a system's apps, data, and connections. Five years ago, it was easy; getting a grasp on what Agentic AI connects to is a much more intangible concept.
During the interview, Travis Rosiek unpacks Rubrik's history and its unique ability to understand complex systems.
CAREFUL
In a rather shocking statement, Rosiek says one should approach introducing technology with the assumption that it will fail. This is not a pessimistic approach, but a nuanced understanding of how complexities in current systems can lead to unintended consequences.
Rosiek advises starting with the end goal in mind, planning for worst-case scenarios, and building trustworthy AI architectures to mitigate risks and ensure reliable operations.
281 епізодів
Поширити
