A public records request on a statutory clock. A retention schedule nobody can apply at scale. A policy question where the answer depends on which version was in force. Kotari turns your policies, records and archives into cited, permission-aware answers — and it treats every AI prompt and output as the record your retention schedule says it is. Your campus assistant was not built for any of that.
A working demonstration on a sample university archive. Switch the desk, tap a question — or type your own — and watch Kotari answer with its source.
Not another campus chat window. The work where being wrong costs money, time or a defensible position — and where the free assistants your institution already licenses were never aimed.
One governed pass across every custodial department's SharePoint, Drive and email. Kotari proposes the responsive set, flags likely exemptions with the statute cited, and drafts the exemption log. Your staff approve every redaction. It works alongside the request tracker you already run.
Content mapped to your schedules, redundant and obsolete material surfaced for disposition, legal holds honoured over schedules, and AI prompts and outputs captured as the records they are — so every assistant the campus licenses stops quietly creating unmanaged public records.
Answers cite the version in force on the date that matters, not the one on the website today. Every lookup is logged, and protected records can inform an answer without being displayed to someone whose access does not reach them.
Source-system permissions enforced at query time, an overshare report that tells you what AI could see before it sees it, and an API your existing assistants can call. You are not being asked to replace Copilot, or your own platform, or anything else.
Advisors and HR consultants answer the same repeat questions in seconds, with a citation they can forward. FERPA-scoped answer sets keep student records inside the roles entitled to them.
AI-drafted description and transcription with the archivist accepting every record, cited search across finding aids and digitised material, and donor and statutory restrictions enforced before retrieval rather than at the reading-room desk.
Most institutions now give staff a general assistant, often free with a licence they already hold. Those are good at drafting. They are not built for the jobs below, and buying another one would not help.
| The job | A general campus assistant | Kotari |
|---|---|---|
| Answer from records across departments | Sees what the person can see, in the tools it connects to | One governed pass across systems, scoped to the request |
| Cite the source | Sometimes, often approximately | Document, version and page, or no answer |
| Flag an exemption under statute | Not built for it | Proposed with the statute cited; a person decides |
| Keep AI interactions as records | A chat tool is not a recordkeeping repository | Captured, classified against the schedule, disclosable |
| Respect a legal hold | Has no concept of one | Hold overrides schedule, and conflicts are surfaced |
| Prove who saw what | Limited | Audit log built to survive a public records request |
We would rather you kept the assistant. Kotari is the governed layer underneath it — reachable by API, so the tool your staff already open can give a cited answer instead of a plausible one.
Every answer links to its source document, version and page. When the records do not contain the answer, Kotari says so and cites nothing, rather than producing a confident guess.
Source-system permissions are enforced at query time. No shadow copy widens access, and a protected record can support an answer without being shown to someone who may not see it.
School-official contract terms, no training on your data, student records segregated and role-scoped. Protected categories are classified and excluded by policy.
A full audit log of queries, answers and access — designed on the assumption that the log itself may one day be requested, and must be producible or withholdable lawfully.
Content is tied to your schedules, legal holds are honoured over them, and AI interactions are captured as records. Kotari proposes disposition; it never destroys.
Kotari proposes and staff decide — on every redaction, every release, every disposition, every archival description. That is a design constraint, not a setting.
Procurement, stated plainly. We are an early-stage company and will not claim certifications we do not hold. A completed HECVAT and a VPAT are in preparation and we will share current drafts on request; SOC 2 is planned, not held. US data residency, or deployment inside your own cloud tenant. SSO via InCommon or Entra ID. Ask us for the current status before your security review, not after.
Narrow on purpose, and small enough to sit inside delegated purchasing authority rather than a formal bid. You pick two custodial departments and one policy corpus. We measure against how you work today.
We time twenty recently closed requests: hours to collect, pages reviewed, days to close. Without this number, nothing afterwards is a result.
Read-only connectors to the systems those two departments already use, in your tenant if your security review prefers it. Nothing moves.
Your staff run live requests through it and check every citation. We build the gold-standard question set with you and score against it weekly.
You get the measured result and the raw data behind it, whatever it says. If the numbers do not justify going on, they do not, and you keep the findings.
These are the targets we agree to be measured against, not results we have achieved. Kotari is early, and a pilot that reports its real numbers is worth more to both of us than a case study that flatters us. For context on the size of the problem: the 236 Washington agencies that reported to JLARC for calendar year 2024 logged 483,861 public records requests, 1.7 million staff hours and $128 million in response costs — and only a quarter of higher-education institutions were above the threshold that requires reporting at all.
Keep them. They are assistants; this is the governed records layer underneath. It does public records, retention, exemption reasoning and audit — jobs they were not built for — and your assistants can call it by API.
No. Trackers do intake, deadlines and publication well. Kotari does the finding, the review and the exemption reasoning, then hands the result back to the tracker.
No. Your records are used to answer your institution's questions and nothing else.
Yes — your Azure or AWS tenant, if that is what your security review wants. Otherwise US-hosted.
Per institution and per module, never per seat. A fixed-fee pilot first, credited toward year one. Per-seat pricing would invite a comparison against free tools that neither of us should waste time on.
Directly, under delegated purchasing authority, at a first contract sized to stay below the formal-bid threshold. Consortium and state vehicles as they come available.
Bring one closed public records request and the policy corpus it touched. We will show you the same request answered with citations, and you can check every one of them against the file you already have.
Book a 90-day pilotTell us a little and we'll set up a 30-minute call.