Lighthouse AdvisorySLED AI Adoption Intelligence

Campus Operations

Prices and assumptions edition. Open records show UT Austin pays about $1.1 million a year for ChatGPT Edu and committed $2 million over two years for Claude Edu, with heavy use billed to departments and about 18,000 accounts created. A Goshen College column put a $48-per-license price on its BoodleBox rollout and objected to required course use. The University of Twente launched a low-risk, knowledge-base-only IT assistant in Teams. Pathify's Alabama A&M story reaches 970 staff hours from an assumed deflection rate. Two patterns: license prices going public, and help desk launches whose savings are still assumed rather than measured.

Evidence records
4
Cross-source patterns
2
Evidence classes
3 independent reporting1 vendor claim
Outcomes
3 emerging1 cautionary
Source freshness
3 new this fortnight1 recent
Research completed
2026-10-31

Choose a role to see its takeaway beside every record in the ledger.

Patterns across the evidence

Synthesis · Lighthouse Advisory interpretation

2 patterns, each supported by at least two sources
  1. Campus AI license prices are reaching the public through records requests and student reporting

    The Daily Texan published UT Austin's contract figures, about $1.1 million a year for ChatGPT Edu and $2 million over two years for Claude Edu. A Goshen College column put a $48-per-license price on its BoodleBox rollout. Interpretation: per-seat prices are becoming comparable across campuses, and adoption and use rules will be judged against them.

    Operating questionCould we explain each AI license price against active users and the rules for required use if it were published tomorrow?

    Supporting evidenceThe University of Texas at AustinGoshen College

  2. Help desk AI assistants launch with narrow scope, while savings claims still rest on assumptions

    Twente launched UT Assist limited to knowledge-base answers, rated low-risk, with personal cases left to staff, and published no targets. Pathify's Alabama A&M story reaches 970 hours by assuming 50% deflection and five minutes per conversation. Interpretation: launch controls are documented, but measured deflection and time saved are not yet reported.

    Operating questionBefore launch, have we defined how we will measure deflection and handling time instead of assuming them?

    Supporting evidenceUniversity of TwenteAlabama A&M University

Evidence ledger

Full record · every source keeps its link and limitations

4 records
  1. Independent reportingEmergingNew this fortnight

    Open records show UT Austin committed about $4.2 million over two years for campus-wide ChatGPT Edu and Claude Edu, with heavier use billed to departments

    Contracts obtained by The Daily Texan show UT Austin pays about $1.1 million a year for a two-year ChatGPT Edu agreement covering 75,000 users and 24.5 million credits, and committed $2 million over two years for Claude Edu for 70,000 users, plus a $73,000 premium agreement. About 18,000 people have created accounts on both since June. Heavier ChatGPT use is billed to departments in capped tiers, and Claude usage limits are set by Anthropic rather than visible to UT.

    The University of Texas at AustinTexas, United States29 October 2026 (The Daily Texan)

    Why it matters, evidence and limitations
    Why it matters
    Rare public price detail for campus-wide licenses from two AI vendors, useful as a benchmark for renewals. Adoption of about 18,000 accounts against 70,000 to 75,000 licensed users is an early utilization signal, not an outcome measure.
    Evidence and measured results
    Contract figures from open records; adoption figure from a UT official; cross-institution price comparison is partial.
    Limitations and uncertainty
    Student reporting with a published correction; per-user comparisons are approximate; no measured benefit reported.
  2. Independent reportingEmergingNew this fortnight

    University of Twente launches UT Assist, a Teams-based IT help assistant grounded in its service desk knowledge base and rated low-risk

    The University of Twente launched UT Assist on 30 October, a chatbot in the Microsoft Teams sidebar that answers simple IT questions around the clock by searching the ICT Service Desk knowledge base and building step-by-step guides. Urgent, complex or personal issues still go to human staff. The assistant passed a security assessment and was classified as low-risk, and users are told not to share personal information or passwords.

    University of TwenteEnschede, Netherlands30 October 2026 (U-Today)

    Why it matters, evidence and limitations
    Why it matters
    A narrowly scoped, knowledge-base-only help desk assistant with a documented risk classification is a common first agent for campus IT. European context; transfer to U.S. campuses depends on local risk processes.
    Evidence and measured results
    Launch report from the university's independent news outlet; no performance data yet.
    Limitations and uncertainty
    No metrics, targets or vendor detail; outcome unknown at launch.
  3. Vendor claimEmergingRecent

    Pathify says Alabama A&M's AI agent saved 970 staff hours, a figure built on an assumed 50% deflection rate and five minutes per conversation

    Pathify reports that Alabama A&M's AI agent handled 23,272 conversations in a year. Assuming half were resolved without staff and each saved five minutes, the vendor calculates 970 staff hours reclaimed. The university also consolidated its portal, mobile app, intranet and chatbot onto Pathify, which the story says saves $25,000 a year. Interpretation: the hours figure is an estimate from assumptions, not a measurement of deflection or staff time.

    Alabama A&M UniversityAlabama, United States7 October 2026 (Pathify customer story, page metadata)

    Why it matters, evidence and limitations
    Why it matters
    Why now: campuses launching help assistants this week, such as Twente, will be offered savings figures built the same way. The source shows the arithmetic plainly, which makes it a useful test case for checking assumptions.
    Evidence and measured results
    Vendor story with a quoted university administrator; key inputs are explicit assumptions.
    Limitations and uncertainty
    Vendor marketing; assumption-based hours; HBCU-specific context; no independent check.
  4. Independent reportingCautionaryNew this fortnight

    Goshen College student column puts a $48-per-license price on its campus BoodleBox rollout and objects to coursework that requires it

    A Goshen College student opinion column states the college began a $38,784 campus-wide yearly subscription to BoodleBox, a multi-model AI platform, at $48 per license on 2 October. It cites a Service Desk email listing uses that include completing course assignments that require BoodleBox, and objects that optional access may become required. The writer's wider claims about AI harms are opinion and were not checked here.

    Goshen CollegeIndiana, United States29 October 2026 (The Record, opinion)

    Why it matters, evidence and limitations
    Why it matters
    Small colleges are buying aggregator AI platforms at low per-seat prices, and student press is publishing those prices and objecting when coursework requires the tool. Limited applicability: one small private college and an opinion source.
    Evidence and measured results
    Opinion column; only the contract price, dates and Service Desk email are used here.
    Limitations and uncertainty
    Opinion piece; figures attributed to the writer; no college response included.

How to read this edition

Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.

Independent reporting
Independent reporting with attributable sources but without a formal evaluation design.
Vendor claim
A supplier-provided assertion that has not been upgraded to independent evidence.