Tag Archives: ET

UT Sage and Social Learning

AI, ,

Most leadership development in a technology organization looks the same. A workshop, a cohort, a facilitator, a shared reading, ninety minutes on the calendar. I have run these and I have sat in them. The thing that almost never happens in that room is someone saying out loud that they do not understand the reading. Senior people especially will not do it. The cost of admitting confusion in front of your peers is higher than the cost of nodding along, so people nod along, and the conversation stays at the altitude where everyone is safe.

Over the past few weeks I have been designing a set of self-paced learning labs for Enterprise Technology. Each one pairs a foundational text with a translation into our work and a set of practice moves. The first is Wenger on communities of practice. The second is Roy Pea on distributed intelligence, which asks how intelligence travels through a system rather than sitting in any one head. The third is David Weinberger’s Small Pieces Loosely Joined, which opens on the difference between the Hoover Dam, where every dependency was planned in advance, and the Web, where almost nothing was and whose designers chose it that way. All of these readings and associated insights have been key ingredients when I have taught my Disruptive Innovations class that my good friend and colleague, Scott McDonald, and I co-designed back in my Penn State days.

Screenshot of the ET LEarning Labs showing three labs.

The interesting part to me is that I have finally figured out a compelling use for AI within the flow of instruction that seems to work. Each lab also ships with its own UT Sage Tutor. That is the piece I want to share, because it is doing something I did not fully anticipate when I started.

The Office of Academic Technology and Enterprise Technology partnered to build Sage at UT for course tutoring. An instructor uploads their materials, Sage builds a tutor scoped to that course, it lives inside Canvas, and it teaches Socratically rather than just handing over answers. It has been available campus wide since last fall and it works really well. What I did here was point that same capability at something else entirely, which is leadership development for our own staff.

Screenshot of ET Learning Lab showing the Wenger example

The tutors in these labs are trained on at least two things. The first is the source text, so the tutor can genuinely discuss Wenger or Pea rather than gesture at them. The second is a large document I assembled by scrubbing our own website, everything Enterprise Technology says publicly about who we are, how we are organized, what we believe, and what we run.

That second corpus is what makes it work in the context of ET. A tutor trained only on the reading can tell you what a community of practice is. A tutor that also knows our org chart, our governance bodies, our service portfolio, and our stated values can help you work out whether the thing you have been calling a community of practice at UT actually is one.

So each lab runs two scaffolds at once. There is a public one, the text and the structure and the practice moves, which everyone in the lab shares and can argue about together. And there is a private one, the tutor, where you can be exactly as confused as you actually are. No cohort, no facilitator, no assessment, nobody watching. You can ask the question you would never ask in the room. The private scaffold is not a retreat from the social one. It is what makes the social one possible, because the confusion gets worked out first, and what people carry into the group conversation is an argument instead of a nod.

Screenshot of Sage showing an active learner chat.

I have shared this with about a dozen people so far, and everyone is blown away by the conversations they can have with the Sage Tutor. What comes back to me is not “I read the Wenger text.” It is a specific argument about whether one of our teams is a community of practice or just a reporting line. If Wenger tells us learning is social, then that is the conversation I was hoping for.

The labs are built in combination with Claude Code and are published to GitHub Pages inside our own enterprise GitHub organization, gated so that only the people we have granted access can read them. Within each learning scaffold there are one button copy starter prompts that a learner taps and is taken to Sage to start a conversation. The number of conversation screenshots I got last week was very cool.

Screenshot of the learning scaffold showing the Sage prompts.

After building the first two tutors I noticed I was repeating myself, so I built a front end plugin for Claude that walks a designer or instructor through the steps of building one. It asks for the real learning objectives before anything else, pushes for outcomes you could actually observe a person demonstrating, and keeps pulling the designer back toward grounded pedagogy before a single line goes into a prompt.

Most of us build a tutor by describing the personality we want and hoping instruction falls out of it. This works the other way around, and it turns out the instructional design is what makes the tutor good, not the prompt craft. I was surprised how easy it was, and more surprised by how repeatable.

When people ask me how AI is disrupting education, I now have a real-world example of how to move AI into the learning process rather than around it. I have known for a couple of years that using AI responsibly actually makes me understand the problems I am trying to solve in a deeper way. Why? Because learning is a social endeavor, and Sage is the closest thing I have found to a colleague who has read everything and still wants to hear what you think.

If you are a UT Austin campus member and want to try this out or talk more about these ideas, drop me a comment or hit me on Teams.

Organizational AI Enablement

AI, SLT, Thoughts, ,

Last March I attended my first Kung FU AI advisory board meeting. I walked away feeling way out in front, we had contracts in place with Microsoft, OpenAI, and Anthropic for campus to safely consume the best of the emerging AI market. It was a classic “if you provide it, they will come” approach. Fast forward to March of 2026 and I walked away from the same annual meeting with a different feeling: that we have let the early momentum slide.

What I listened to as I sat and looked around the room were CTO, CEO, COO type people not talking about procurement wins or what they were thinking about AI, but instead about how they spent much of the year activating and enabling their IT organizations to be AI-first in their decision making. The intentional adoption of AI to drive impact was more important than early wins and “low hanging fruit.”

I left that meeting feeling a different kind of urgency, one that was aimed directly at enabling the teams that make up Enterprise Technology. One of the first things I did was schedule an all day retreat with the ET senior leadership team. Mario and I sat down and laid out the sketch of how we wanted to approach the SLT and built a framework that was meant to challenge each of us. The challenge was to come to a place where we would take a very intentional turn and focus on internal adoption, diffusion, and enablement.

That is a hard turn for an organization that spends nearly all of its time looking out across campus. We have rightly focused on campus adoption, diffusion, and enablement while often times leaving ourselves behind. What I learned from the advisory board meeting was that the only way to grow outside of ET is to first grow inside ET.

With that in mind we walked into our retreat and declared that that day was Day Zero of ET’s intentional AI adoption and enablement practice. We each made commitments. We also addressed the outcomes of our first ET AI Use Survey, where we learned that our staff were reluctant to adopt AI for reasons we hadn’t considered: environmental, ethical, and uneven tooling concerns dominated the results. It was an opportunity for us as the leadership team to square our organizational expectations and commit to focusing significant effort inside ET to enable success for all of campus.

Today we sent a strong message to the ET staff that they have the support and commitment from their leaders to bring us all into the world of AI to improve our own work, support campus, and to challenge ourselves to rise to this moment. My favorite part of the memo reads:

ET’s core values are not a backdrop to this work. They are the reason for it. We exist to advance the university’s digital ecosystem so the Longhorn community can learn, discover, and succeed. That mission requires us to develop our people, engage our campus partners, build an organization that is agile enough to act on new opportunities, measure our progress honestly, and enhance the technology environment for teaching, learning, and research. AI is the most consequential technology shift of our time. Sitting it out is not an option that is consistent with who we say we are. 

In the coming weeks and months there will be lots of ongoing conversations, training, and events designed to enable our teams. We have AI Enablement workshops starting next week that will bring all of our teams up to a baseline with both OpenAI and Claude. We have made the SLT commitments available to every member of Enterprise Technology and we will deliver. We have another AI in a Day event planned that will allow our staff to hear directly from the audiences we serve across campus to spark ideas and new partnerships. And we will culminate this calendar year with an event designed to bring our AI champions together in an ET Build event that will focus on deepening our skills and using them to solve our own problems of practice.

All of this is designed to get us to a point where we can begin to work with our campus partners in new and exciting ways. This means understanding what the offices of the CFO, COO, Advancement, Student Affairs, Legal Affairs, and all the others struggle with — challenges that can be augmented by AI, agents, and workflows. The focus inward is intentional and we believe it will be the unlock we need to continue our AI leadership in higher education.

Future state of Enterprise Technology Agentic pods.

As our teams grow their skills in AI, we will pair with offices across campus to create “agentic pods” to move our enablement work to them. The method itself isn’t new; it is simply an evolution of the tooling and practices we are all familiar with. When the UT.AI Studio fully launches, we will be able to augment our teams with our AI Student Builders and further expand impact. I am grateful to be at a place that demands to be on the forefront of discovery and we get to be part of that journey in a very meaningful way.

Do Our Agents Need Identities?

AI, Thoughts,

OpenAI released a 13-page policy document today called Industrial Policy for the Intelligence Age: Ideas to Keep People First. It’s ambitious. It proposes robot taxes, a national public wealth fund seeded partly by AI companies, automatic safety net triggers tied to displacement metrics, containment playbooks for rogue AI systems, and pilots of a 32-hour workweek framed as an “efficiency dividend.” Sam Altman told Axios that the scale of what’s coming is comparable to the Progressive Era and the New Deal.

I read the whole thing. And I want to engage with it seriously, because the ideas matter. But I also want to say something that I think is missing from the conversation, something that becomes visible only if you’re operating at the institutional layer where these impacts actually land.

Universities sit at the intersection of almost everything this document talks about. We are workforce development engines, research enterprises, employers of tens of thousands, and the training ground for the next generation of workers whose careers will be shaped by whatever policy regime emerges. We hold sensitive data, manage federal compliance obligations, and operate complex enterprise systems that keep all of it running. If transformative AI is coming, and I believe it is even if the timeline is debatable, the university is where the policy meets the pavement.

He proposes distributing AI-enabled research infrastructure broadly across universities, community colleges, hospitals, and regional hubs. Good. It talks about portable benefits that follow individuals across jobs and industries. Good. It calls for modernizing the tax base away from payroll and toward capital gains as automated labor displaces human labor. That’s a real conversation worth having. And it proposes that workers should have a formal voice in how AI is deployed in their workplaces, something I believe in deeply.

But here’s where I want to push further, because I think there’s a conversation that we aren’t having yet.

We are already in the early days of deploying AI agents that perform real institutional work. Not chatbots answering FAQs. Agents that process transactions, triage requests, route approvals, generate reports, monitor systems, and make decisions within defined parameters. The trajectory is clear: these agents are going to take on more responsibility, operate with more autonomy, and become embedded in workflows that currently depend on human staff. So here’s my question: if an agent is doing the work of a full-time employee, shouldn’t it be governed like one?

I don’t mean this as a thought experiment. I mean it operationally. At UT Austin, every employee has an Enterprise ID, an EID. That EID is the key to everything: system access, role-based permissions, org chart placement, budget allocation, position control, performance accountability. Our Workday HCM instance manages the lifecycle of every employee from hire to retire. Now imagine an AI agent that manages reimbursement exception processing, or monitors infrastructure and initiates remediation workflows, or handles first-pass review of procurement requests against policy. That agent consumes resources. It has a cost. It operates within a reporting structure. It needs access controls. It needs to be auditable. And someone, a human, needs to be accountable for what it does.

As of today, none of us have an institutional framework for this. Agents float in a governance gap. They aren’t in Workday. They don’t have position numbers. They aren’t reflected in our staffing models or our budget structures. They aren’t covered by the HR lifecycle processes that ensure every human worker has clear accountability, supervision, and a paper trail. And yet they are increasingly doing work that, if a human were doing it, would absolutely require all of those things.

This isn’t a technology problem, it is a human capital problem. OpenAI’s document talks about shifting the tax base from payroll to capital gains as automated labor grows. That’s a macro policy question. But at the institutional level, the equivalent question is: how do we account for an agent’s labor in our workforce planning? If an agent handles the equivalent of two FTEs worth of procurement review, does that show up in our staffing model? How does it affect position requests? Budget justifications? If we’re reporting headcount to the Board of Regents or to federal agencies, do we need a parallel accounting for agent capacity? And what about accountability? When a human employee makes an error in a compliance-sensitive process, there’s a clear chain. When an agent makes that same error, who owns it? The developer who built it? The product owner who scoped it? The CIO whose organization deployed it? We need to accelerate toward conversations to these questions.

Do we need something like a UT Agent Registry, a formal institutional record for every AI agent that performs work on behalf of the university? A governed registry that captures what the agent does, what systems it accesses, what authority it has, who supervises it, and how its performance is measured and audited. The equivalent of an EID. A position description. A reporting line and a professional development budget. This might sound like bureaucracy. It’s not. It’s the same governance discipline we apply to every other resource that operates on behalf of the institution. We don’t let humans access sensitive systems without identity management, role-based access, and a clear accountability chain. We shouldn’t let agents do it either.

At the UT System level I am responsible for exposing what AI tools we have in our environment and what types of work they do, but nothing at the level I am working through.

OpenAI’s policy document is forward-looking in many ways, but it still frames the AI transition primarily as something that happens to workers and to economies, with governments and companies managing the fallout. What it doesn’t reckon with is that institutions like universities are going to be running hybrid workforces, humans and agents, long before the national policy framework catches up. We will be making these decisions in our ERP systems, in our identity platforms, in our governance structures, whether or not Washington has figured out the robot tax question.

At UT Austin, we’ve been working toward this, but the writing on the wall is becoming more and more clear with each experiment and through each observable outcome. UT.AI is our common AI environment where data protection, privacy, accessibility, security, and institutional identity are the foundation. The self-healing campus concept I wrote about last month is built on the premise that agentic AI will enable domain experts to build personal, ephemeral interfaces to institutional data. But that vision only works if the foundation is right, and part of getting the foundation right is being honest about the fact that agents are becoming part of our workforce and we need to govern them accordingly.

I don’t have all the answers. But I think the conversation needs to start with a simple recognition: the line between a tool and a worker is blurring, and our institutional frameworks haven’t caught up. OpenAI may be right that we need a new industrial policy. But we also need a new human capital policy, one that accounts for the non-human actors that are increasingly doing institutional work alongside our people.

Practicing Like We Play

Thoughts, ,

On game day, Darrell K Royal Stadium becomes the heart of campus. More than one hundred thousand fans show up to cheer for the Longhorns, filling the stands in burnt orange. I look forward to Saturdays in the fall for so many reasons, but the best is that I get to feel like I am truly part of something far bigger and more meaningful as I take in the scenes from the stands. For the fans, we all want the day to feel seamless. Tickets scan, Wi-Fi connects, replays play, and the stadium feels secure. That simplicity is not an accident; it’s the result of months of preparation and a game day of real-time effort from teams in Enterprise Technology and our partners all across campus.

Just as the football team prepares with practices, film study, and repetition, we practice the way we intend to perform. Perfection is always the goal. The Networking team tunes wireless coverage across the stadium and ensures every vendor, ticketing station, and media outlet can connect. The Cable and Construction team checks and runs the fiber and cabling that carry instant replay, coach-to-sideline communication, and live broadcasts across the nation. The Warehouse team stages and delivers every piece of gear, from radios and cables to generators and water, so that when the call comes, it’s ready. The Electronic Physical Security Systems team sets schedules, monitors cameras, and ensures that safety is woven into the game day experience from the start.

Evening at DKR, Austin, TX

When kickoff arrives, those teams are moving together as one. Networking watches over every access point and switch. EPSS monitors security and supports the Emergency Operations Center. Cable and Construction crews are ready for rapid response. The Warehouse team keeps supplies moving to where they’re needed most. It’s live, it’s fast, and just like the players on the field, execution must be perfect.

Game day is proof of a larger truth at UT Austin, technology touches nearly every aspect of campus life. From classrooms to research labs, from student housing to DKR Stadium, our work shapes the experience of our community. Like all the teams across this campus, we accept that responsibility with seriousness and pride. Our role is to prepare, to execute, and to remain in the background so that students, faculty, staff, alumni, and fans can focus on what matters most. That is the measure of success. When UT takes the field, and the stadium hums with energy, the technology simply works. To me, this is what it means to lead at a world-class university. Quietly, reliably, and with the same pursuit of perfection we expect from the Longhorns on the field. That is the standard we set for ourselves and that is just some of the work that makes me proud to serve this community. Hook ’em!

Exposing the Missing Pieces in Our Content

AI, Thoughts, ,

Part of our campus AI journey is to design and deploy AI agents that can utilize key information from exisiting websites across campus. These agents may replace the sites, reducing technical bloat and information drift. While doing so, an unexpected benefit has emerged, one that speaks volumes about the evolving relationship between technology and content strategy on a highly decentralized campus.

When we first set out to build these agents, we did what most teams do, we pointed them at our sites or their underlying data, ingested the knowledge, tested retrieval, and began crafting conversations. But something interesting happened when we put these agents to work. People have started asking for things we couldn’t give them.

In short, the agents began surfacing questions we hadn’t anticipated; questions students, faculty, staff, and prospective Longhorns are likely asking every day. And, just as importantly, they showed us where our data and content fell short.

They have become mirrors, reflecting the structure, and the fragmentation, of our institutional knowledge. The things they cannot answer point directly to gaps in the content architecture: outdated FAQs, scattered documentation, siloed policy pages, and even buried gems of information lost in PDF archives or legacy web systems. It’s not that the information doesn’t exist. It’s that it’s too hard to find, inconsistently written, or lacks the context necessary to form a coherent response. We are sure the agent isn’t making mistakes per say, it tells us what it can’t say, and that’s been incredibly valuable.

One of the more revealing moments for me came when we began evaluating how the agent performed with the A–Z directory. This is a resource that has long served as the backbone for finding services and offices across the university. But once we put the agent to work with this data, the limitations of that system became painfully clear. What we had assumed was structured, complete, and reliable turned out to be limited, outdated, and in some cases, misleading.

UT Spark AI interface showing the A-Z agent.

This has been a bit of a wake-up call. It is so tempting to take a “lift and shift” approach, move what we have on the web into the AI agent and assume it will just work. But that does not hold up. The agent exposes what the web often hides. It forces precision. It requires context. And it absolutely demands trust in the data that fuels it.

We are now integrating these insights into a more systematic approach. Each time a query breaks down, we want to trace it back. What we need to be asking centers on: Should this information exist? If so, where should it live? Can we make it easier to find, easier to understand, and easier for the agent to serve up confidently?

This work is not just about making our AI better. It’s about making our websites more accessible, our documentation more useful, and our services more responsive. Every gap we close improves the experience not just for the agent, but for the human trying to find their way. I didn’t expect this kind of feedback loop to emerge so quickly, but I’m glad it has. It reminds us to slow down, look closely, and be intentional, not just with how we build agents, but how we steward the information we share across this institution.