{"resourceId":"utk-ut-verse-facilities-services-20261009","versions":[{"version":"external-ea1a6f089656cd385076b1afc69057445d3de418956724ab7fe3ce0e35509273","resource":{"id":"utk-ut-verse-facilities-services-20261009","title":"UT Knoxville Facilities Services helped shape UT Verse, an in-house AI platform on OpenAI models","organization":"University of Tennessee, Knoxville (Facilities Services and Office of Innovative Technologies)","sector":"Higher education facilities and IT","geography":"United States (Tennessee)","publishedAt":"2026-10-09","publicationDate":"2026-10-09","eventDate":null,"sourceName":"UTK Facilities Services","sourceLabel":"Institutional news story","sourceUrl":"https://facilities.utk.edu/2026/10/09/every-voice-every-department-how-facilities-services-helped-shape-ut-verse/","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["operating-model","accessibility-workforce","infrastructure","developers-agents"],"finding":"UTK built UT Verse in-house as an AI platform with an intelligence layer over OpenAI models, for up to 55,000+ users. Facilities Services tested early versions and pushed for translation in niche languages such as Swahili; a data bridge syncs OneDrive/SharePoint content and lets departments build their own assistants.","sledRelevance":"A rare facilities-workforce perspective on enterprise AI: frontline operations staff needs (translation, procedure lookup) shaped an in-house platform.","evidence":"Institutional story by a student communications assistant; no usage, accuracy or outcome data. Claims of unmatched accessibility integration are unsupported in the piece.","architectureImplications":"Build-on-API model: university-engineered platform layer over commercial models, with document sync connectors and department-built assistants.","governanceImplications":"Department-created assistants need review of source documents, access and ownership; not described.","securityPrivacyImplications":"Syncing SharePoint/OneDrive content requires permission-aware retrieval; the story does not explain controls or OpenAI data terms.","caveats":"Promotional institutional content; very small team; scaling described as hardest challenge; no metrics; accessibility superiority claim unverified.","streamIds":["campus-operations"],"roles":{"sales":"Frontline workforce needs are the opening: UTK Facilities asked for translation and procedure lookup for a diverse workforce. Conversations with campus operations leaders can explore multilingual access to SOPs and training material, but this story offers no results, so propose a bounded pilot. Do not imply UTK's accessibility claims are verified or that build-your-own platforms are cheap.","engineering":"Permission-aware retrieval is the key design concern: the data bridge syncs OneDrive and SharePoint into department assistants over OpenAI models. Engineers should check identity propagation, sync scope, model data terms and load handling, since scaling was the hardest challenge. Proof of value should test translation quality for niche languages with native speakers. Interpretation; the source has no technical metrics.","delivery":"Co-design with operational departments is the delivery pattern: Facilities tested early versions and fed back requirements. Delivery owners need a process for department-built assistants, document ownership and content refresh, plus support capacity, given a very small central team. Acceptance criteria are proposals. Applicability is limited where institutions lack engineering staff to build in-house."},"roleDetails":{"sales":{"customerProblem":"Facilities staff need procedures and communication in many languages, including niche ones.","stakeholders":["Facilities Services leadership","Central IT innovation office","Frontline supervisors","Accessibility office"],"discoveryQuestions":["Which languages do frontline staff need?","Where do SOPs and training materials live?","Who would own department assistants?"],"valueHypothesis":"Department assistants over existing documents may improve findability and translated communication, to be tested.","potentialEngagement":"A facilities-focused pilot of multilingual SOP assistants with native-speaker evaluation.","unsupportedClaims":["Accessibility leadership over other institutions","Productivity gains","Translation accuracy in niche languages"]},"engineering":{"fit":"Fits institutions with in-house engineering and Microsoft 365 document stores.","architectureIntegration":"Platform layer over OpenAI models with OneDrive/SharePoint sync connectors.","prerequisites":["Engineering capacity","Clean department document sets","Model provider agreement"],"constraints":["Small team capacity","Load and scaling limits","Dependence on one model provider"],"security":"Ensure synced content respects source permissions and model data-use terms.","proofOfValue":["Native-speaker translation review","Load test at expected concurrency","Permission-leak tests on synced docs"]},"delivery":{"work":"Onboard departments to build assistants over their procedures.","dependencies":["Central platform team","Department content owners","M365 administrators"],"ownership":"Office of Innovative Technologies with department content owners","skillsAdoption":"Workshops for department assistant builders; Frontline staff orientation","governanceCheckpoints":["Content review before assistant publication","Periodic access audit"],"acceptanceCriteria":["Proposed: translation rated acceptable by native speakers in sampled tasks","Proposed: staff report faster procedure lookup in survey"],"risks":["Key-person dependency","Stale synced content","Overstated accessibility claims"]}},"retrievedAt":"2026-10-11T21:56:30-05:00","enrichedAt":"2026-10-11T21:58:50-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Translation for multilingual facilities workers is a concrete workforce-inclusion use case; effectiveness not measured.","procurementImplications":"Build rather than buy; model-provider dependence on OpenAI remains.","operatingModelImplications":"Small central team (a few engineers) supports a campus-wide platform; departments co-design via testing and workshops.","sourceVerification":{"openedUrl":"https://facilities.utk.edu/2026/10/09/every-voice-every-department-how-facilities-services-helped-shape-ut-verse/","referenceExcerpt":"It uses OpenAI models, but UT Verse itself is an AI platform with an intelligence layer, built and engineered by UT","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}