ASU's OpenAI launch was a promise; its October update was a count

The classroom note

When Arizona State University announced, on 18 January 2024, that it had become the first university to collaborate directly with OpenAI, the announcement itself described an intention rather than a finished programme. ASU said it would bring ChatGPT Enterprise into three areas, enhancing student success, opening new research avenues, and streamlining administrative processes, and that starting in February 2024 it would invite faculty and staff to submit proposals for how to use it. For a professor reading the news that week, there was nothing yet to sign up for beyond a promise that a process would open.

What the evidence says

ASU's own later reporting gives the process a shape the launch announcement did not. By 8 October 2024, the university's news service reported that the first round of what it now called the AI Innovation Challenge had drawn over 175 proposals, of which 105 were accepted, granting over 800 ChatGPT Enterprise licenses across the ASU community; a second round of proposals had also been submitted and reviewed by that point, though without matching published totals for acceptances or licences. The same article names specific funded work rather than general categories: one project used ChatGPT to help build personalised sociology lecture scenarios tied to student career interests, and a second-round project used it to revise learning outcomes and career-based scenarios across two sociology courses. This is self-reported university data, not an independent audit, and it counts proposals funded rather than measuring what those projects changed for students.

The implementation question

A competitive proposal process is a way of rationing a fixed number of enterprise licences rather than a way of deciding whether the tool works: 105 accepted out of 175 submitted describes selection pressure, not evaluated outcomes. The licences themselves came through ASU's existing enterprise agreement with OpenAI rather than a separate line-item purchase by each department, which lowers the individual barrier to entry for faculty but concentrates the underlying cost and data-governance decisions with the university's central technology office. Neither ASU document discloses what the enterprise agreement itself costs the university, only what faculty and staff receive once inside it.

What holds and what fails

The challenge structure holds as a record of interest and distribution: it shows real demand, with acceptance rates and licence counts a reader can check against later years. It does not, on its own, hold as evidence that funded projects improved learning or research output, since the university's reporting to date describes what was funded and built, not what changed for students afterward. Reading a proposal count as a success metric would be an editorial overreach the sources themselves do not make; the honest reading is that ASU has documented uptake, not yet impact.

• What became of the specific funded projects after the round in which they were announced, and is there a public follow-up?
• How many of the 800-plus licences are still active, and for what proportion of the original proposals?
• What does ASU's enterprise agreement with OpenAI cost the university, separate from what an individual licence-holder pays?

An announcement of collaboration and a later count of accepted proposals are both real records, but only the second describes anything that actually happened inside a classroom.