What Texas Is Telling Us About AI — and What UT Should Hear

The Anthropic Economic Index tracks how people actually use Claude — not how companies intend to deploy it, not how institutions plan to integrate it, but what real people reach for AI to do in their daily lives. The January 2026 dataset covers 15,600 conversations from Texas. Texas ranks 24th out of 50 states, with a usage index of 0.76x — meaning Texans are using Claude at about three-quarters of the rate the national model would predict given the state's workforce composition and population.

The rank and the index are interesting. But the pattern of use is where it gets meaningful.

The top use in Texas is academic

The most common thing Texans are using Claude for — by a meaningful margin — is completing academic assignments and creating educational materials (4.7%). The second most common is completing humanities and social science academic assignments across multiple disciplines (3.6%). Together, those two uses account for more than one in twelve conversations. Educational Instruction and Library tasks represent 11.6% of all Claude use in the state, the second-largest job category behind Computer and Mathematical work at 19.9%.

This is not the AI story we usually tell about Texas — the energy sector, the tech corridor, the rapid enterprise adoption in Austin and Dallas. The dominant signal in this data is educational. Texans are reaching for AI first when they need to learn something, teach something, or do academic work.

What the flagship owes the state

The University of Texas at Austin is the flagship public university in one of the largest states in the country. We serve more than 50,000 students on this campus and support faculty whose research, teaching, and public engagement shape how Texans understand the world. If the data tells us that the number one use of conversational AI in Texas is academic work — completing assignments, creating instructional materials, learning across disciplines — then this institution has a specific responsibility. Not just to provide AI tools, but to help our community understand them, use them well, and shape what they become.

That is not a passive role. It requires investment, experimentation, and honest engagement with hard questions. When a student uses Claude to complete a humanities assignment, is that a shortcut or a scaffold? When a faculty member uses it to develop instructional materials, is that efficiency or atrophy? The data does not answer those questions. It just tells us the questions are live — right now, in this state, at scale.

What Enterprise Technology is building toward

Enterprise Technology's AI work is built on the belief that UT needs to be inside this conversation, not observing it. That means institutional infrastructure: UT Spark, our enterprise ChatGPT instance built on a contract that prevents OpenAI from training on UT data; Anthropic Claude Enterprise, available to faculty and researchers for complex analytical work; and the UT.AI Studio, a coordinated framework for AI tools, governance, and responsible use.

It also means something harder: building human understanding alongside the tools. That is the animating purpose behind the ET Faculty Fellows program, which we are launching this fall. The program will fund three UT Austin faculty members each year to work alongside ET staff on real research questions — not to evaluate AI tools, but to investigate what AI in teaching and learning actually looks like in practice, and to share what they find with the broader UT community. The inaugural cohort will focus on generative AI in course design, AI literacy and critical thinking, learning analytics, accessible digital learning, research-integrated teaching, and human-AI collaboration across disciplines.

We are also investing in training at scale. Through UT Learn, LinkedIn Learning, and the workshop series running through spring and summer, ET is building the kind of AI literacy that goes beyond tool familiarity — critical evaluation, responsible use, and the judgment to know when AI helps and when it does not.

The honest part

Texas is at 0.76x. That means we are, in some aggregate sense, behind where the model expects us to be. Part of that gap is likely the educational use pattern itself — students and faculty tend to be cautious early adopters, more attuned to questions of integrity and appropriate use than general consumers. That caution is not a deficiency. It is the right instinct in a context where the norms are still being written.

But caution without engagement cedes the field. The institutions that shape how AI is used in higher education will be the ones that stayed in the room — that built tools, ran experiments, funded fellows, and kept asking hard questions. That is the work Enterprise Technology is trying to do. The data from the Anthropic Economic Index is one more reason to keep doing it.