Research contract review that drafts the redline from your own playbook, cites every rung and waits for the right approver
Policy-based redlines the same day, and every deviation approved and logged.
See one case, screen by screen ↓Research contract review is the work a university's research contracts office does on every data use agreement, material transfer agreement, sponsored research agreement and confidentiality agreement: reading the other party's terms clause by clause against the university's negotiation playbook, drafting a redline and getting each deviation approved. Research Contract Review puts six agents in front of that work: they place every clause on the playbook's ladder with the policy line cited, draft the redline at the nearest acceptable rung, and route each deviation to the person who must approve it.
AI in education: a buyer's guide to agentic AI for colleges, universities and school districts →
Updated
Research contract review is the clause-by-clause work behind every new study, dataset or industry collaboration: a data use agreement to receive data, a material transfer agreement to receive or send research materials, a sponsored research agreement for industry-funded work, a confidentiality agreement to talk about it. Each arrives on the other party's paper. A contract negotiator rereads every clause against a playbook kept in a document or in memory, emails the IRB and the data security office, writes the redline in a word processor and waits days for the associate director or the general counsel's office to approve any deviation.
Meanwhile the data terms drift apart — the agreement says destroy the data in 30 days, the IRB protocol says keep it 7 years — and a public university's limits on indemnity, governing law and public records slip through or wait weeks for counsel. Fallbacks are accepted by email, so nobody can later say who approved which deviation, on what precedent. Research Contract Review puts six agents in front of the reading and drafting, so the negotiator spends her afternoon on the one clause that needs her.
Estimated split of a typical data use agreement's business days to a first redline, by hand and with the agents; the by-hand total sits at the low end of the published 4 to 6 weeks. In the working solution the sample agreement went from arrival to sent in 78 minutes. Our arithmetic, kept modest.
Six agents place every clause on the policy ladder, draft the redline, cross-check the IRB protocol and route approvals; people decide.
DUA-26-1187 arrives at Halden State University on Tuesday, October 20, from the State Department of Education: 11 pages, 14 clauses, on the department's own form. Here is what happens to it, screen by screen, in the working solution.
The department shares student-level records for the 2017 to 2024 high school graduating cohorts, about 412,000 students, for the Dual Enrollment and College Completion Study under FERPA's studies exception. The Reader agent identifies a data use agreement at 0.99, links IRB protocol 2026-0418, data security plan DSP-26-077 and award HSU-SP-26-0388, and starts the first-redline clock, due November 3.
23 agreements arrived today, 104 this week, about 5,200 a year.
Each clause is a row on the four-rung ladder of Halden's Research Agreements Playbook, RC-7 version 12: Preferred, Fallback 1, Fallback 2, Walk-away. The sponsor's language is marked on its rung, the drafted redline's rung is ringed, and the row says who decides. The full redline is drafted 40 minutes after the agreement arrived.
8 Preferred · 3 Fallback 1 · 2 Fallback 2 · 1 Walk-away
Clause 9 asks Halden to indemnify, defend and hold harmless the Department. Contracting Policy RC-4 §3.2 says Halden, as an agency of the state, may not indemnify another party, so the Clause agent places it on Walk-away at 0.97. The Drafter wrote the redline at Fallback 2 — Halden is responsible for claims arising from its breach of Sections 5 and 12, to the extent the State Tort Claims Act permits, without waiving sovereign immunity — and Priya uses it.
The Precedent agent: this department accepted Fallback 2 in 5 of 6 earlier agreements.
The Cross-Check agent compares 8 data terms across the agreement, IRB protocol 2026-0418 and data security plan DSP-26-077, every cell cited. Six match. Exhibit B names 4 authorized users where the protocol lists 5, so the redline adds graduate assistant Kai Nakamura. Retention conflicts: clause 6 destroys all data within 30 days of the project's end, while the protocol keeps the identifiable linked file 7 years.
The agent proposes two fixes and recommends amending the protocol: identifiable data destroyed within 30 days of the project's end, the de-identified file kept 7 years, because the department refused longer retention in 4 of 4 earlier agreements. Only the IRB analyst can confirm it; Owen files modification MOD-2026-0418-02 and the PI is told. The FERPA written-agreement test shows all four elements, each cited to its clause.
Theo's queue shows each fallback with the sponsor's rung, the redline rung, the drafted language and what this department accepted before: additions to authorized users by notice in 5 of 6, 30-day destruction in 6 of 6, a 45-day publication review in 5 of 6, 24-hour breach notice in 6 of 6. He approves all four for DUA-26-1187 in one step.
Indemnity and public records language reach counsel, not the associate director, and a walk-away row reaches her only after the negotiator has decided it. Renata approves Fallback 2 on clause 9 and Fallback 1 on clause 11, which names the State Public Records Act, wording this department accepted in 6 of 6 earlier agreements.
Send stays off until the checklist is complete: the walk-away clause decided, 6 of 6 deviations approved, the IRB protocol amended to match clause 6, and FERPA's elements 4 of 4. Priya sends the redline with a cover note that explains every change in plain words to Marisol Duarte, Data Governance Manager at the department.
Sent to Marisol Duarte at 3:10 p.m., drafted in 40 minutes. First-redline SLA met with 9 business days to spare.
The agreement's record holds the timeline from receipt to send, every version, and the linked IRB protocol, security plan and $640,000 award that starts January 1, 2027. When the department agrees, Dr. Achterberg signs through the e-signature system with the deviation summary in front of him: each of the six deviations, who approved it and when.
Every placement, redline, approval, amendment and send is on the audit trail, for agents and people alike, and so is a refused action: confirming the IRB protocol amendment belongs to Owen Takeda, and the attempt by anyone else is refused and logged. Kendra reads and exports; she can build a redacted export pack for internal audit or the state agency.
The same data use agreement, seen by the five people who carry it — what their day looked like, and what it looks like now.
Six agents place every clause on the policy ladder, draft the redline, cross-check the IRB protocol and route approvals; people decide.
Identifies the agreement type — data use, material transfer, sponsored research or confidentiality — with the parties, term and template, splits it into clauses, and links the IRB protocol, data security plan and award.
Places each clause on the playbook's ladder for its clause type and cites the policy line behind the rung.
Writes the redline at the nearest acceptable rung, with a plain-language note for each change and a cover note for the other party.
Compares the data terms with the IRB protocol and the data security plan, tests FERPA's written-agreement elements, and proposes the fix on whichever side is wrong.
Finds how the same counterparty settled each clause before, with each earlier agreement cited.
Routes approvals by rung and clause type, keeps the SLA clock, and prepares the signatory's one-page deviation summary and the e-signature package.
Decide, approve, send. The agents propose; a named person decides.
Ask about any agreement, clause or rule, in plain words. Every answer points to the clause, the protocol or the policy line it came from.
Why is clause 9 walk-away, and what did you draft?
Clause 9 asks Halden to "indemnify, defend and hold harmless the Department". Halden, as an agency of the state, may not indemnify another party under Contracting Policy RC-4 §3.2, which is the walk-away rung of the indemnification ladder. The Drafter wrote the redline at Fallback 2: Halden is responsible for claims from its own breach of Sections 5 and 12, as the State Tort Claims Act allows. This department accepted Fallback 2 in 5 of 6 earlier agreements.
Does this agreement match the IRB protocol?
Almost. Of 8 data terms, 6 match. Exhibit B names 4 authorized users and the protocol lists 5, so the redline adds Kai Nakamura. Clause 6 destroys the data in 30 days and protocol §8.3 keeps it 7 years; amending the protocol is proposed, and Owen Takeda decides. Security matches the Level 3 plan.
What does FERPA require in this agreement?
Because the department shares student records with Halden for a study done for it, the studies exception applies, and it needs a written agreement with four things: purpose, scope, duration and the data (clauses 1, 2 and 4); use only for that purpose (clauses 1 and 3); no identification of students outside the team (clauses 3 and 8); destruction when no longer needed, with the time period (clause 6). All four are present.
Why do state agencies push indemnity past walk-away?
Their own forms are written for vendors, who can indemnify; a public university cannot. This quarter 38 of 93 state-agency data use agreements arrived with it, and 34 closed at Fallback 2.
13 screens from the working solution, on its sample data. Pick one to see it large.
SLAs, approvers, thresholds and retention are settings, saved as a new version and logged. The ladders themselves change only with counsel's approval.
| Setting | Default | Choose from |
|---|---|---|
| First redline · data use agreements | 10 business days | Set per type: confidentiality 2, material transfer 5, sponsored research 15 |
| Approvals by the associate director and counsel | 1 business day | Escalate to the director 2 days past SLA; remind the sponsor after 5 business days without a reply |
| Who approves a deviation | Business-term fallbacks: associate director · legal terms: counsel | Preferred needs nobody; walk-away goes to the negotiator first |
| Two approvers for sponsored research over $1 million | On | On · Off — the associate director and counsel |
| Edited language always goes to counsel | Off | On · Off — when the negotiator writes beyond the rung |
| Send fully Preferred confidentiality agreements without a queue | On | On · Off — the negotiator still sends; no approver needed |
| Agent confidence thresholds | Reader 0.95 · Clause 0.85 · Cross-Check 0.90 · Precedent 0.80 | 0.80 · 0.85 · 0.90 · 0.95; below it, the row goes to a person |
| Keep executed agreements | 6 years | After the agreement ends, or 3 years after the award's final report if longer (2 CFR 200.334) |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
Sources: a university library's published data use agreement guidance (4 to 6 weeks for a typical agreement) · university research office pages (a two-week minimum from submission to signature; an 11.3-day average for incoming material transfer and data use agreements) · FERPA studies exception, 34 CFR 99.31(a)(6) · 2 CFR 200.334 (records retention) · "target" = design goal, measured in the live solution · "demo" = seen in the working solution, on its sample data · "typical" = published figures · "estimated" = our arithmetic from published inputs. Halden State University, the State Department of Education, the companies, agreements and people named on this page are fictional — characters and sample data in the working solution.
Research contract review is the work a university's research contracts office does on every data use agreement, material transfer agreement, sponsored research agreement and confidentiality agreement: reading the other party's terms clause by clause against the university's negotiation playbook, drafting a redline and getting each deviation approved. Research Contract Review puts six agents in front of that work: they place every clause on the playbook's ladder with the policy line cited, draft the redline at the nearest acceptable rung, and route each deviation to the person who must approve it.
A data use agreement is the contract under which a researcher receives data from another party, such as a state education agency sharing student records for a study. Under FERPA's studies exception, the written agreement must state the purpose, scope, duration and information disclosed, limit use to that purpose, bar identification of students by anyone outside the study, and require destruction with a time period. The Cross-Check agent tests those four elements and cites each to its clause.
Universities publish 4 to 6 weeks for a typical data use agreement, longer when terms must be negotiated. In the working solution, the sample agreement was fully redlined 40 minutes after it arrived and sent at 3:10 p.m. the same afternoon, with every deviation approved.
Yes. The playbook holds four-rung ladders for data use, material transfer, sponsored research and confidentiality agreements, and the Reader agent identifies which type arrived. A sponsor's right to approve publication is walk-away, because it would end the fundamental research exclusion for export control, and confidentiality agreements that sit fully on Preferred are ready to send in one step.
No. The agents read, place, draft, cross-check and route; they never approve, send, sign or change an IRB protocol. The Drafter uses only the playbook's approved rung language, and below each agent's confidence threshold the row is marked for a person.
Yes. The negotiator decides walk-away rows and sends; fallbacks on business terms go to the associate director, legal terms to counsel, and protocol amendments to the IRB analyst. The negotiator can never approve her own deviation, and Send stays off until every deviation is approved.
Yes. The ladders are your playbook's own rungs and language, cited on every placement. The director proposes a rung change and counsel approves it as a new version; SLAs, approval routing, thresholds and retention are settings, and each change is logged.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your playbook, your templates and a sample of executed agreements — and it goes live once every quality gate has passed. We will walk you through it on your own agreements first.
Five high-fit calls in minutes, each limit and cost share cited to the sponsor's own text.
1-min demoGrant Proposal ReviewEvery proposal reviewed inside the five-day window, every value tied to its clause.
1-min demoSubrecipient Risk Assessment & MonitoringRisk scored before signing and every invoice line checked before the university pays.
1-min demoWe'll run Research Contract Review on a sample of your own agreements and your negotiation playbook.