Let’s talk AI systems designed from a workflow the organisation already runs, grounded in its own material, with people in the decision and a record of every answer. Custom AI engagements are contracted with Genyra L.L.C-FZ, the separate Dubai company that operates genyra.ai. Genyra.com describes the practice; the agreement names the company doing the work.
Discuss the workCollect the missing information before proceeding.
We start with the task: who performs it, what they read, what they decide and what they produce. From that come the inputs the system may use, the output it must produce and the point at which a person takes over. Only then is a model selected, and it is selected to fit the task rather than the other way round.
This is also where the contracting entity is confirmed. Custom AI engagements are contracted with Genyra L.L.C-FZ in Dubai, which operates genyra.ai; genyra.com describes the practice.
Legal matters, insurance claims, patient correspondence and financial files share a shape: many documents, one question, one accountable decision. A custom solution reads the documents, extracts what the task needs, drafts the summary or answer and attaches the source of every statement. The case stays a record with a history, not a chat.
Question · A. Reyes · 11:32
Which clauses limit assignment?
Three clauses limit assignment. The agreement may not be assigned without the other party’s written consent [1]; an assignment to an affiliate needs no consent while the assignor remains liable [2]; and the consent requirement applies to an assignment arising from a change of control [3].
Limits. Drawn only from the three documents above. Correspondence and any later amendment outside the matter file were not checked. Whether a change of control has occurred is for the reviewer.
| No. | At | Who | Record |
|---|---|---|---|
| 0415 | 2026-10-14 11:32:04 | A. Reyes | question · M-2041 |
| 0416 | 2026-10-14 11:32:05 | system | retrieved 3 · MSA-2041, SCH-3, AMD-2 |
| 0417 | 2026-10-14 11:32:09 | system | draft · model 2026.10-03 |
| 0418 | 2026-10-14 11:40:51 | A. Reyes | approved · docs MSA-2041, SCH-3, AMD-2 · model 2026.10-03 |
The system retrieves passages from the organisation’s own documents, indexed inside its boundary, and answers from those passages. Each answer names the document and the clause, page or entry it relies on. If the material does not contain an answer, the system says so instead of producing one.
A named reviewer sees the draft, the cited passages and a note on the limits of the answer, then approves or sends it back. The decision, the reviewer, the documents used and the model version are written to an audit log that the organisation keeps. Nothing goes out unreviewed unless the organisation has decided, in writing, that a class of output may.
Before anything is relied on, the system is run against cases the organisation has already decided, and its output is compared with what the people did. Disagreements are examined one by one. The evaluation set, the results and the changes made in response are kept, so the next change can be judged against the same examples.
The engagement states this plainly. Automated: reading, retrieval, extraction, drafting, consistency checks and routing. Not automated: the decision that binds the organisation, advice to a client or patient, and any action the organisation’s rules reserve to a person. The line is written into the design and enforced by the system, not left to habit.
From an unanswered question
to an engineered possibility.
