Research conflict of interest review that reads the public record, asks the investigator first, and cites every item
Undisclosed ties found before just-in-time, and explained by the researcher first.
See one case, screen by screen ↓It is research conflict of interest and research security review for COI offices: agents read each senior and key person's papers, patents, talks and public profiles and tie every item to the disclosure, other support line, appointment or management plan that covers it. What nothing covers is "Not yet matched", never a violation, and the investigator is asked one neutral question before anything is recorded. The COI analyst, the Director, the research security officer and the signing official decide and sign.
AI in education: a buyer's guide to agentic AI for colleges, universities and school districts →
Updated
Research conflict of interest review today compares forms with forms. An investigator's annual disclosure, their NIH other support and their HR appointment record can agree perfectly while one of their own papers lists a second affiliation and thanks another institution for a visiting appointment, facility access and travel. That gap, between what was disclosed and what the public record shows, is what the federal foreign-support cases turned on, and since January 25, 2026 every NIH Common Form carries a personal certification of no participation in a malign foreign talent recruitment program.
Nobody has time to read the public record. A COI analyst cannot read 31 papers, 2 patents and 14 talks for every senior and key person in NIH just-in-time, so the check runs only after a tip or a sponsor letter, and a check triggered by a name or a nationality is unfair. Discrepancies are emailed back and forth while the PHS clocks run: 30 days to disclose a new interest, 60 days to report a financial conflict of interest, 120 days for a retrospective review. Research Conflict of Interest & Research Security Review gives every key person the same check, on documented facts, and lets the investigator explain before anything is recorded.
Estimated working days for one investigator whose public record shows something nothing on file covers, by hand and with the agents. The 'after' rows from the question to the recorded answer stay within the 2-day target median; the rest is our arithmetic, kept modest.
Seven agents read the public record, tether it to what was disclosed and ask the investigator; people decide and sign.
Dr. Tobias Okonjo-Reyes, Associate Professor of Biomedical Engineering at Halden State University, is in NIH just-in-time for his R01 application SP-27-0044, with other support due Friday, October 30, 2026. Here is his review, screen by screen, in the working solution.
Before any agent read a record, the Fairness guard removed a place-of-birth line on one public profile and a citizenship field in the HR feed. The Footprint agent then read his publications, patents, talks, public profiles, trips and HR appointment overnight. One conference site timed out, so the Talks source shows 13 of 14 read with a Retry button; a gap is never silent.
28 items tied at 0.95 or more and 3 not yet matched; handed to Mira Sandoval with one question drafted.
Today shows four numbers: 2,846 of 3,100 annual disclosures certified, 46 open reviews (27 hers), 1,046 key people checked for just-in-time this year and 3 clocks at risk. "Needs you" is led by Dr. Okonjo-Reyes: 3 items not yet matched, one question ready.
On the left, 8 records he disclosed or that are on file, from his 2026 annual disclosure to management plan MP-24-031. On the right, 9 of 31 public-record items. Six are tied to the record that covers them with the Matcher's confidence, 0.95 to 0.99; three are not.
Not yet matched · 1 question for the investigator
She opens the April 2025 paper in Hydrogel Science Letters: page 1, footnote 2 names a second affiliation at the Laboratory for Soft Materials, Varnholm University, and page 14 thanks it for a visiting appointment, rheometry facility access and travel support. She asks whether anything on file mentions Varnholm University; the cited answer is no, from his annual disclosure, R21 other support, HR record and the travel system.
The agent grouped the paper, the lab's web page and a March 2024 seminar into one question, CL-26-0193, each item cited, asking what the appointment involved, its dates and any pay, resources or travel support. Mira sent it for the Director's check at 9:05 a.m.; he approved it and it went out at 9:12 a.m., due Wednesday, October 28.
No "violation", "undisclosed" or "foreign"; every item cited; due before the October 30 deadline.
He attaches the appointment letter, the closing email and two travel reimbursements. The Clarification agent checks each point with a citation: the dates agree (0.96), no remuneration or equity (0.94), $930 and $1,380 of travel paid by a foreign university to disclose (0.97), and other support to correct for the 10-month overlap with his R21 (0.95).
"This was an unpaid visiting appointment from January 2023 to June 2024. … I have no continuing role."
Mira records E2, explained, records to update, with a note: unpaid appointment confirmed by the letter and closing email, no remuneration or equity, two trips to add, other support to correct, no FCOI. The Director confirms at 11:06 a.m., and the item goes to research security.
She sees the appointment's documented facts: unpaid, January 15, 2023 to June 30, 2024, ended, $2,310 in travel. Each of the six talent-program features reads "Not present", and a note confirms name, nationality, citizenship and birthplace were removed. She records S1: not a malign foreign talent recruitment program.
In his portal the Disclosure agent proposes the two trips, October 9–20, 2023 ($930) and March 4–15, 2024 ($1,380), with the travel rule explained and cited. He confirms both and certifies; his portal also answers whether this will be reported to NIH as a conflict: it will not.
The Support update agent added three cited lines to his Common Form other support: the unpaid visiting appointment, about 6 weeks of facility access and $2,310 in travel support. He certified at 2:40 p.m.; Karen signed and submitted at 3:05 p.m., 9 days before the deadline.
What is not changed: no FCOI report, no "violation" anywhere, plan MP-24-031 unchanged.
The same review, seen by the people who carry it: what their day looked like, and what it looks like now.
Seven agents read the public record, tether it to what was disclosed and ask the investigator; people decide and sign.
Guides annual and ad hoc disclosures in plain questions, explains each threshold in a sentence, and prefills from HR, travel and the investigator's own answers.
Reads a person's publications (affiliations and funding acknowledgments), patents, talks, public profiles and travel records.
Ties each found item to the disclosure, other support line, appointment, invention disclosure or management plan that covers it, with a confidence and a citation at both ends.
Removes name origin, nationality, citizenship, place of birth, ethnicity and language from every agent's input, logs what it removed, and blocks any statement that rests on them.
Groups unmatched items by topic, drafts one neutral question, then checks the investigator's answer against the sources and the rules.
Drafts the corrected Common Form other support and the disclosure update, line by line, with each source cited.
Drafts management plans from like plans, FCOI reports and annual reports, and keeps the research security certification binder current.
Decide, confirm and sign. The agents propose; a named person decides.
Ask about any investigator, item, clock or report, or say what to prepare. Every answer cites its source.
Why is the Varnholm affiliation not yet matched?
Nothing on file covers it. His April 2025 paper lists the Laboratory for Soft Materials, Varnholm University, as a second affiliation and thanks it for a visiting appointment, facility access and travel support. His annual disclosure, R21 other support and HR record list no outside appointment. It is grouped with the lab's web page and the March 2024 seminar into one question. It is not a finding: he explains first.
Which clocks are at risk this week?
Three: Dr. Okonjo-Reyes's just-in-time other support is due Oct 30; Dr. Beatriz Nwosu's review is on day 41 of 60; Dr. Imani Rutherford's annual disclosure is 21 days overdue after two reminders.
Draft the corrected other support for Dr. Okonjo-Reyes
The corrected lines come from his recorded answer: the unpaid visiting appointment, the in-kind facility access and $2,310 in travel support, with the R21 overlap. Unpaid appointments count. He certifies, then Karen Ostrowski signs.
Show the research security certification evidence
The binder has 26 items across the four program elements; 24 are ready. Training is current for 1,201 of 1,284 people. Two items are missing: one college's cybersecurity attestation and an export control roster. The Vice President for Research certifies when all 26 are ready.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds and clocks every review runs on are settings, versioned, and changed only with the Director's approval.
| Setting | Default | Choose from |
|---|---|---|
| Significant financial interest threshold | $5,000 | PHS rule; the institution may set it lower |
| Sponsored or reimbursed travel | Every trip | Except travel paid by US universities and governments |
| Just-in-time footprint | Every senior/key person | Cannot be limited to some people |
| Matcher confidence | Ties on its own at 0.90 and above | 0.70–0.89 asks the analyst; below 0.70 stays not yet matched |
| Reply to a question | 5 business days | Then 2 reminders, then the department chair |
| FCOI report to the agency · retrospective review | 60 days · 120 days | Clock warnings 5 days and 1 day before; escalations to the Director |
| Fairness guard | Locked on | Name origin, nationality, citizenship, place of birth, ethnicity and language removed |
| Review records | 3 years from the final expenditure report | 42 CFR 50.604(i); auditors see initials only |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
"demo" = seen in the working solution, on its sample data · "target" = design goal, measured in the live solution · "estimated" = our arithmetic, kept modest (the days split) · "typical" = published figures: US Department of Justice, October 2, 2023 ($1.9 million, foreign support omitted for 12 faculty from proposals to five federal agencies); HHS Office of Inspector General, 2019 ($5.5 million, foreign grants not disclosed on NIH applications). Rules named as context: 42 CFR Part 50 Subpart F; NIH NOT-OD-21-073 and the Common Forms, NOT-OD-26-018; CHIPS and Science Act §§10631–10638; NSPM-33 and the OSTP research security program guidelines of July 9, 2024; FERPA, 34 CFR Part 99. Halden State University, Varnholm University, the companies, journal and society, and the people named on this page are fictional — characters and sample data in the working solution.
It is research conflict of interest and research security review for COI offices: agents read each senior and key person's papers, patents, talks and public profiles and tie every item to the disclosure, other support line, appointment or management plan that covers it. What nothing covers is "Not yet matched", never a violation, and the investigator is asked one neutral question before anything is recorded. The COI analyst, the Director, the research security officer and the signing official decide and sign.
Under the PHS rule, 42 CFR Part 50 Subpart F, an investigator discloses significant financial interests above $5,000, and the institution decides whether an interest is related to the research and could bias it. A related interest that is a financial conflict of interest gets a management plan, approved by the COI committee, and is reported to the PHS awarding component within 60 days. In the solution the analyst and Director work the review, the committee votes and the Chair signs the plan.
NIH's other support instructions cover all resources in support of an investigator's research, including in-kind resources and foreign appointments, whether paid or unpaid. In the sample review, an unpaid visiting appointment that a 2025 paper revealed was added to the corrected R21 other support with its facility access and travel support, certified by the investigator and signed by the authorized organization representative.
Every senior and key person in NIH just-in-time gets the same footprint check, and the deadline is tracked to the day. In the sample review, the request arrived October 8, the question went out October 21 at 9:12 a.m., and the corrected other support was submitted that afternoon, 9 days before the October 30 deadline.
Institutions with more than $50 million a year in federal R&D must certify a research security program covering cybersecurity, foreign travel security, research security training and export control training. The solution keeps the evidence in one binder, each item linked to its source: training within 12 months, travel briefings, foreign-appointment decisions and footprint checks. The research security officer prepares it and the Vice President for Research certifies.
No. A Fairness guard removes name origin, nationality, citizenship, place of birth, ethnicity and language before any agent reads a record, shows what it removed, and cannot be switched off. Every key person gets the same check on documented facts: appointments, payments, affiliations, acknowledgments, contracts, travel and dates.
Yes. Agents never label anything a violation, never contact an investigator without a question the Director has checked, and never submit anything to an agency. The analyst records each outcome with a reason code and the Director confirms it; the research security officer decides talent-program questions; the investigator certifies and the authorized organization representative signs.
The Agentic Solution Engine builds and deploys it from your requirements and documents, such as your COI policy, disclosure forms and a sample of past reviews, and it goes live once every quality gate has passed. We will walk you through it on your own key personnel first.
Two blind votes on every record, a person on each conflict, every number tied to its page.
1-min demoArchival Transcription & Finding AidsBacklogs described in weeks, not years, with every line approved by an archivist.
1-min demoResearch Integrity ScreeningDuplicated figures fixed before submission, not retracted after publication.
1-min demoWe'll run Research Conflict of Interest & Research Security Review on a sample of your own disclosures, other support and public records.