{"resourceId":"isu-bengalbot-grant-locally-hosted-ai-canvas-writing-assistant-safeguards-20261022","versions":[{"version":"external-728eb453943c3777b8868fad6314cc1e1cba12fb8cb1c6fb37a4ea2677921df0","resource":{"id":"isu-bengalbot-grant-locally-hosted-ai-canvas-writing-assistant-safeguards-20261022","title":"Idaho State University wins a $300,000 grant to expand its locally hosted BengalBot AI, adding a Canvas writing assistant with Writing Center referrals, a bias audit and accessibility review before wider student use","organization":"Idaho State University","sector":"Higher education AI-supported writing help, AI curriculum and student administrative tools","geography":"Pocatello, Idaho, United States","publishedAt":"Idaho State University News, 22 October 2026","publicationDate":"2026-10-22","eventDate":null,"sourceName":"Idaho State University News","sourceLabel":"Institutional news item on a state-funded AI grant and its planned safeguards","sourceUrl":"https://www.isu.edu/news/2026-fall/isu-lands-grant-to-build-ai-literate-students-through-bengalbot-expansion.html","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["knowledge-work","infrastructure","governance-procurement","accessibility-workforce"],"finding":"On 22 October Idaho State University reported a $300,000, two-year Accelerate Idaho AI Innovation Award to expand BengalBot, its locally hosted AI system. The grant funds three efforts: embedding BengalBot in the Applied AI Certificate and new AI sciences bachelor's degree (50 to 75 students a year building and auditing AI tools); a Canvas writing assistant that gives feedback on structure and grammar rather than writing papers, with built-in referrals to the ISU Writing Center; and a Stage Hub to reduce administrative load on student stage managers. Per the grant application, data stays on ISU-controlled servers, the writing assistant gets an accessibility review and bias audit before expanding beyond its first group, an English professor reviews a sample of its feedback each semester, and no tool reaches students without passing an internal benchmark. About half the funding buys a server maintained by College of Technology students; code and curricula will be openly licensed, and a theatre pilot and limited English-class use are expected by fall 2027.","sledRelevance":"A concrete pre-launch control set for a student-facing writing assistant (human referral, faculty sampling, bias and accessibility review, benchmark gate) that other public institutions can adapt before scaling AI writing help.","evidence":"Institutional description of grant scope and planned safeguards. No measured outcomes; tools are not yet deployed to students.","architectureImplications":"On-premises, locally hosted model stack on a new university server, with Model Context Protocol connectors that hand precise math and logic tasks to better-suited tools, integrated into Canvas. Interpretation: this trades vendor dependence for in-house hardware, operations and security work.","governanceImplications":"Stage gates are defined before student use: internal benchmark, accessibility review, bias audit and semesterly faculty review of a feedback sample. Who sets benchmark thresholds is not described.","securityPrivacyImplications":"All data hosted locally on ISU-controlled servers with campus IT and network security oversight, per the application. Retention and logging rules for student writing are not described.","caveats":"Announcement of funded intent; no deployment, usage or learning outcomes. Safeguards come from a grant application. Single regional public university.","streamIds":["student-success"],"roles":{"sales":"Sales framing: ISU shows public institutions writing explicit pre-launch controls into AI writing-support plans. Use it to ask a campus how it will gate a student-facing writing assistant: referrals to human tutors, faculty sampling of feedback, bias and accessibility review, and a benchmark before release. A bounded engagement could design those gates. Do not promise writing improvement, retention gains or cost savings: the tools are not yet in student use and no outcome data exists.","engineering":"Engineering framing: this is a locally hosted model stack integrated into Canvas, with Model Context Protocol connectors routing math and logic to deterministic tools. Assess server sizing, model update cadence, Canvas integration, logging of student writing and who operates the hardware after the grant. Validate the internal benchmark itself: what tasks, thresholds and failure cases it covers, and whether referral triggers to the Writing Center fire reliably for drafts that need human help.","delivery":"Delivery framing: the value is in the gates, so delivery should make them operational. Define the first student cohort, the benchmark and pass threshold, the bias-audit method, the accessibility review standard and the faculty feedback-sampling rubric before launch. Track referral volumes to the Writing Center and sampled feedback quality each semester, and plan ownership of servers, models and support beyond the two-year grant."},"roleDetails":{"sales":{"customerProblem":"Institutions want AI writing help that supports rather than replaces student writing, with controls they can defend before students use it.","stakeholders":["Writing center director","Provost or teaching and learning office","Campus IT and security","Accessibility services","Faculty in composition or English"],"discoveryQuestions":["What benchmark must a student-facing tool pass before release?","When does the assistant refer students to a human tutor?","Who reviews AI feedback quality, and how often?","Who owns the system after grant funding ends?"],"valueHypothesis":"Interpretation: a gated, locally hosted writing assistant could extend writing support while keeping data on campus, if benchmark, referral and review gates work as designed.","potentialEngagement":"A bounded readiness engagement: define release benchmark, bias and accessibility review methods, referral rules and a one-semester evaluation plan for a first cohort.","unsupportedClaims":["Improved writing quality or grades","Reduced Writing Center load","Lower cost than commercial licences","Bias-free feedback"]},"engineering":{"fit":"Fits institutions with capacity to run on-premises AI infrastructure and an LMS such as Canvas.","architectureIntegration":"Locally hosted model server with Model Context Protocol tool connectors and a Canvas-embedded writing assistant.","prerequisites":["Server hardware and hosting","Model selection and update process","Canvas integration permissions","Internal benchmark suite","Writing Center referral workflow"],"constraints":["Not yet deployed to students","Hardware build planned for spring","Two-year grant funding","Single institution; transfer to another Idaho campus untested"],"security":"Keep student writing on university-controlled servers; define logging, retention and access rules for prompts and feedback.","proofOfValue":["Benchmark results against agreed thresholds","Bias audit across student groups","Accessibility review of the interface","Faculty-rated sample of feedback quality","Referral trigger accuracy tests"]},"delivery":{"work":"Build, gate and pilot a locally hosted Canvas writing assistant and related student tools.","dependencies":["Grant funding and hardware purchase","Campus IT and network security oversight","Writing Center participation","Faculty reviewer time"],"ownership":"Project co-principal investigators with campus IT; long-term owner to be named.","skillsAdoption":"Student developers and operators trained to build, audit and maintain AI tools; faculty trained to review AI feedback samples.","governanceCheckpoints":["Internal benchmark pass before student use","Bias audit and accessibility review before expansion","Semesterly faculty review of feedback samples","Post-grant ownership decision"],"acceptanceCriteria":["Proposed: benchmark thresholds documented and met before first student use","Proposed: accessibility review completed with issues resolved before expansion","Proposed: faculty review at least one feedback sample per semester with ratings recorded","Proposed: referral path to the Writing Center tested in scenario cases"],"risks":["Feedback that rewrites rather than coaches","Bias in feedback across student groups","Hardware or staffing gaps after the grant","Delays pushing pilots past fall 2027"]}},"retrievedAt":"2026-10-23T03:15:43Z","enrichedAt":"2026-10-23T03:19:48Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Accessibility review is a stated gate for the writing assistant. Students are paid to build and maintain the system, which develops AI auditing skills the university links to regional employers.","procurementImplications":"Grant-funded build rather than licence purchase; roughly half the money goes to hardware. Open licensing of code and curricula could let other Idaho institutions reuse the work.","operatingModelImplications":"A faculty-led, student-staffed build team with campus IT oversight; long-term ownership after the two-year grant is not described.","sourceVerification":{"openedUrl":"https://www.isu.edu/news/2026-fall/isu-lands-grant-to-build-ai-literate-students-through-bengalbot-expansion.html","referenceExcerpt":"The writing assistant will undergo an accessibility review and a bias audit before expanding beyond its first group of students","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}