Research ethics depends on careful academic judgment. Before that judgment can be exercised, someone has to assemble the submission, find the supporting documents, identify gaps and prepare the questions that need an answer. That preparation is a practical opportunity for AI to give academics time back.
At UBI Business School, we have connected our Research Ethics Committee (REC) workflow to DeeJay, our AI colleague in Microsoft Teams. Staff can find cases, check their progress and carry out authorised steps through a conversation, with separate human confirmation for decisions and letter issuance.
The productivity model we are building towards is straightforward: an AI agent prepares the first cut, academics review and refine it, and authorised people give final approval.
Give academics a useful first cut
The next phase adds an AI-assisted first review of the submitted documents. It starts with the academic team’s required-document checklist and de-identified examples of acceptable and incomplete submissions. Those materials establish what the agent should check and what a useful review should look like.
The intended output is a concise review brief: what the project proposes, which required materials are present, what appears ready for academic review and where clarification or further documents may be needed. DeeJay would prepare an initial opinion against the agreed criteria, with references to the relevant material so a reviewer can check the basis for it.
Where a submission needs work, the agent can draft follow-up questions and a request for additional information. The academic can then examine, edit and approve that draft rather than reconstructing the case and writing every question from a blank page.
This first-cut document review is the next implementation phase; the verified Teams workflow already handles case coordination and human-confirmed actions. The academic checklist and reference examples are the starting point for enabling and evaluating the additional review capability.
Keep academic judgment and final approval with people
A first cut gives the academic a prepared starting point. The tutor still reads the relevant documents, checks the agent’s interpretation and makes the academic decision. The responsible Quality team retains oversight of the process and final issuance.
Reviewers need to be able to correct an initial opinion, add questions, identify concerns that the checklist misses and decide that further review is necessary. The agent’s output remains a draft until a person has assessed it.
In the connected workflow, DeeJay presents a proposed action for explicit confirmation by the staff member. The record captures the decision and who made it. Checks on the current case status and processed confirmations help prevent a repeated request from becoming a duplicate operation.
Connect preparation, review and the final letter
Students enter through the REC submission form, providing their project details, selecting a tutor and uploading the supporting documents. The workflow organises those files into a case folder and creates the register entry.
Staff can ask DeeJay to show pending REC cases, open a particular case or prepare a tutor decision with comments. After the appropriate review and a separate human confirmation to issue, the workflow generates the REC reference and PDF letter, emails it to the student with the tutor copied, and updates the register.
The Teams path has been tested with a fresh synthetic submission through tutor confirmation, separate issuance confirmation and receipt of the PDF. A repeated confirmation was refused, and only one matching approval email was received. The workflow remains a test prototype for the academic team to evaluate before institutional rollout.
Measure productivity in the work academics actually do
The potential gain comes from reducing repeated searching, checking, summarising and drafting across many submissions. Academics can start closer to the point where their expertise matters: considering the research, questioning the assumptions and making a sound decision.
We will assess the first-cut review through staff handling time, corrections to the agent’s initial assessment, the usefulness of its follow-up questions and the quality of the completed record. Faster preparation is valuable when it also supports a thorough, accountable review.
This work is shaping IJK’s AI implementation consulting arm. We begin with a real workflow, define what the agent should prepare, and make the human review and final approval explicit. For academic teams, that creates a practical path to greater productivity while keeping responsibility for research ethics with the people entrusted to exercise it.