Jason Buffington .com

Watching the NVIDIA AI Factory devices at hashtag#DellTechnologiesWorld (and traveling with a device that couldn’t run my usual AI tools) got me asking a question most IT leaders haven’t found great answers for yet: how are you ensuring the resilience of the infrastructure that fuels your AI-enhanced productivity?

DPM will be exploring that during our survey next month, but let’s start with a simpler question for you: how does your org back up your AI?

video transcript

One of the most impactful visuals coming out of Dell Technologies World was watching Michael Dell and NVIDIA CEO Jensen Huang walk across the main stage together — showing the range of Agentic AI devices from the GB10 (about the size of a Kleenex box) to a massive rack that could fuel the AI needs of a large enterprise: same underlying technology, right-sized for you.

This blog isn’t about Dell — but watching those local AI factories got me thinking about the number one rule of IT resilience: if your business processes rely on IT systems, those systems have to be resilient; and that includes being backed up and recoverable.

And that got me revisiting something most IT leaders underestimate: how are you protecting your endpoints? From a cyber security perspective:

• Most cyber events start with an end user clicking something they shouldn’t.
• Once that endpoint is infected, whatever it connects to is next.

But cyber resilience is more than detection or prevention. Check out my earlier video on how a pan-Dell strategy has some compelling cyber resilience ramifications that start right at the endpoint.

So, then I started imagining one of those GB10s sitting next to my desktop — and wondering, how would I back it up? Because while it may be accessing data storage that’s backed up elsewhere, it’ll have agent customizations, instruction files, installed skills, and configuration files that I do not want to recreate. Like any other IT resource that business processes rely on, it needs to be backed up.

Beyond those purpose-built AI devices, I realized how much I rely on Claude when I was on the road with a laptop that couldn’t run my usual AI tools. And even if it could, my customizations and recurring tasks live on my desktop. What if those were lost — or deleted? Think about just recreating your scheduled workstreams, your custom dashboards, the MCP connections that enable most of your agents to actually work.

Most IT environments ignore endpoints unless they’re OT rather than IT — since most IT data lives within a SaaS service or a cloud-synced file share. But as our endpoints start running unique AI tools and configurations, we need to revisit how we ensure their resilience. And that starts with asking how we back up what’s on them.

That’s one of the areas DPM will be testing in our upcoming AI-Enhanced Resilience research. So let’s start with one key question: how are you ensuring the resilience of the AI tools and platforms your business relies on?  Leave your thoughts below.