“AI pulls back the sheet on your data access” Data Permissioning Q&A with CIO Dalan Winbush
Nasuni CIO Dalan Winbush shares his four-phase approach to over-permissioning and explains how he’d use DryvIQ to reduce months of governance work down to minutes.
November 10, 2026 | Dalan Winbush
We caught up with Nasuni CIO Dalan Winbush, fresh from delivering his sessions on data governance at Workato’s Conference in Las Vegas, for an honest discussion about data access and over-permissioning. In this Q&A, he unpacks why addressing over-permissioning is not a one-off fix, what a bunch of keys has in common with an exposed AI agent, and how he’d use DryvIQ to reduce hours of governance work down to minutes.
Q: Many enterprises don’t know the extent of their permissioning problem until they give AI access to their file data. How does AI raise the stakes around permissions and data access?
AI pulls back the sheet on your data access, exposing over-permissioning at scale. Whatever you give it access to, it will aggregate that data. And if you don’t understand who has access to what, that’s an operational and security risk. The threat vector expands to boundaries you aren’t familiar with.
Last month, I was on a panel with a number of CIOs who are scared to even open up their unstructured data to AI because of fear that their access control lists (ACLs) are over-permissioned. That fear keeps them stuck at the bottom layer of the pyramid of agentic adoption, figuring out how to properly manage their data. It’s cannibalizing their ability to get value from what AI brings to the table.
Q: When an organization realizes they have an over-permissioning problem, how do they get a handle on how deep it goes? Where’s the starting line?
I’d recommend a four-phase approach: understand, manage, protect, activate.
- You can’t fix what you don’t know, so start by understanding the health of your data. Inventory what you have, where it lives, and who can open it. That’s how you find out how deep your over-permissioning problem goes.
- Then comes data management. When files are scattered across servers and clouds, each with its own ACL, over-permissioning becomes inevitable. Good data management brings permissions back in line with policies.
- Next, take action to safeguard your data. That action can be manual or automated. This is where you shut the doors. Every person – and agent – gets access only to what they need for their job.
- The final stage is activating AI apps on top of what you’ve already governed. AI should only ever show people what they’re already allowed to see. Once you’re confident of that, then you can open your data up without fear.
But the four phases aren’t sequential. You have to build a continuous cycle of improvement. Once you have reconciled your permissions, you need to manage them consistently over the long term so that you don’t return to an insecure, over-permissioned state.
Q: Earlier this year, Nasuni acquired DryvIQ, adding AI-powered content intelligence and automated governance capabilities to Nasuni’s File Data Platform. Where in that four-phase cycle have you found DryvIQ most valuable?
DryvIQ is one of the best products I’ve used, and I’m not saying that out of bias. Before the acquisition, Michael Normington, our Director of IT, spent hours manually cleaning up permissions to understand whether we could open up our environment to AI. Those are expensive hours. If we had DryvIQ four months ago, we could have accomplished that within a couple of minutes, and without the risk of human error. And not just reconciling the permissions, but proactively managing them going forward. It shows what’s in your file data and helps build resilience into your protection layer.
Q: Does tighter data governance mean more friction for data users who are accessing data for their day-to-day work?
Not if it’s done right. Tighter governance doesn’t cause friction because data users only access the files they need. A person who’s over-permissioned isn’t necessarily aware of the over-permissioning themselves unless AI exposes it by returning information that they shouldn’t be seeing. That’s the part that scares me.
The other day, I found a set of keys in my house and I had no idea what they had access to. And it so happened that one of the keys was to my parents’ rental house. It’s identical to my key, so I didn’t even know that I had it. But it meant that those keys were just lying on my table all along.
Now picture an agent with access to data it shouldn’t have. If its credentials get exposed, then that agent becomes a bunch of keys on the table. And you can bet that somebody’s going to try every single key until they find a way to get inside.
Q: Once governance is in order, how should CIOs think about building permission checks into their AI deployment process going forward? Is it possible to become proactive when AI is moving so fast?
You need preventative measures in place to ensure that you don’t go back to the old behaviors because unstructured data that’s over-permissioned is useless. That’s a 50-year-old problem. And in 2026, it’s holding organizations back from realizing the value of AI.
AI is changing our environment too fast for most data governance and security postures to keep up with. Here’s why that’s dangerous: Let’s say your car gets a flat and you put a donut on. The donut comes with a warning saying you can’t go faster than 50 miles per hour. Even though you still have three perfectly good tires, driving at 100 mph could end in disaster. To create an equilibrium around your vehicle, you need everything to move at the same pace. Your car is only as fast as its weakest tire.
That donut is data governance lagging behind the velocity of AI development. It both exposes the organization to risk and holds the business back. If your AI strategy isn’t advancing, you can’t differentiate yourself in your market and your competition will operate more efficiently or get to market faster.
Q: What practical advice do you have for CIOs who are struggling to move at the pace of that change?
AI moves faster than humans can, so use AI to help you govern. CIOs – including myself – with petabytes of data, could have 100 people with hands on keyboards going file-by-file to figure out permissions, and still not catch threats in real time. But products like DryvIQ give you the ability to govern permissions at scale across more than 40 repositories.
It doesn’t stop at visibility. You can have an agent sitting on top of DryvIQ working around the clock to monitor all directories, identify anomalies, and take action. If they find that someone has access to data beyond their remit, they can independently reach out to the domain owner to query it. They continuously govern what you’ve already put in place — in real time. That’s the next paradigm and that’s where governance is headed.
CIO Corner is the executive lens on what’s next in IT, delivered by Nasuni CIO, Dalan Winbush. With a firsthand perspective from the frontlines of IT leadership, Dalan unpacks what it really takes to modernize infrastructure, harness AI, and lead through complexity. This series tackles the hard questions CIOs face today, such as scalability, resilience, velocity, and value. It’s a candid look at how cloud and AI are reshaping enterprise IT — from someone who’s doing it, not just talking about it.
Related resources
October 14, 2026 | Dalan Winbush
“We’re building agentic AI upside down” Data governance Q&A with CIO Dalan Winbush
Nasuni CIO on why “What’s your AI strategy?” is the worst question, what terrifies him about AI data governance, and how CIOs can make every AI dollar…
Read more
September 16, 2026 | Dalan Winbush
Which Data Management Trends Should CIOs Have on Their Radar? A Q&A with CIO Dalan Winbush
Nasuni CIO Dalan Winbush shares how he keeps up with unstructured data management trends and his predictions for the future of the data infrastructure…
Read more
The Enterprise Guide to Data Governance and Compliance
96% govern their data. Only 31% enforce it automatically for AI agents. What enterprises need to know about data governance and compliance in the AI era.
Learn more
