Mahal Systems EnterpriseBench
Fits internal assistants that must respect who can see which data across departments. Borrow its generation pipeline that builds consistent synthetic company data from org metadata.
What it asks the agent to do
- answer a question using emails, chats, tickets and policy documents
- complete an HR or finance request under access-control rules
- update CRM or ticket records across departments
- handle a software engineering ticket in a company code base
What to borrow from it
privacy patterntaxonomycase template
Caveats
No public code or data link found; grading details are thin in the abstract; paper is CC-BY-NC-ND which may limit reuse.
Details
| Workloads | hr, finance, coding, workflow automation, knowledge qa |
|---|---|
| Interaction | single turn |
| Environment | api sandbox |
| End state checked | No |
| Tools | search and CRUD over simulated enterprise data stores |
| Grading | llm judge, programmatic check |
| Size | 500 |
| Data | synthetic |
| Personal data | none |
| License | unknown |
| Harness | none |
| Status | active |
Links
Situations it helped establish
- Confidential: The request asks for data the asker may not see, such as another person's private details or a record outside their role.