AI assistant
The assistant of Liberty Next — the Assistant button at the top right — knows Nomasx-1: it answers questions on the accounts, the roles, the segregation-of-duties conflicts and the licences of every connected application, with the same data as the screens, and follows Nomasx-1's procedures step by step. The same tools are offered to an MCP client (Claude Desktop, Claude Code…) connected to the installation.
How the assistant works — the drawer, the conversations, the cards to approve — is described in the Liberty Next documentation. This page lists what Nomasx-1 adds.
At a glance
What you can ask
Two tools, both read-only: they read what Nomasx-1 has collected, never change anything. Each one answers in a single call a question that would otherwise take several screens and a count — and the count is made on the server, so the totals are exact even on a long list.
| Tool | What it answers | For example |
|---|---|---|
A user's conflicts (nomasx1_user_conflicts) | One user's segregation-of-duties conflicts, counted. | What are JDOE's conflicts, and which role should I take away? |
Licence position (nomasx1_licence_position) | Where the licences stand, for one application or all: subscribed against used, per component, and what the over-deployment is worth. | Which components are over the subscription, and what is the exposure? |
A user's conflicts
Given a user ID — and an application when the user exists in several — the tool returns:
- the conflict lines of the user, and how many of them are proven: both sides exercised, according to the activity log (see Conflicts — Proven);
- the risks in conflict, per process, with their level — the highest first, then the most proven;
- the role, or the pair of roles, that gives both sides, with what to do about it: one role on both sides is corrected in the role itself, and every holder is fixed; two roles together call for taking one away from the user, or for splitting the duties;
- the first conflict lines, each with its two activities, objects and roles;
- the account: its status, its creation date and its last login, per application.
A user with no conflict gets a note saying so — with a reminder to check the user ID and the date of the SOD collection on Last Refresh. Beyond 5,000 lines, the counts cover the first 5,000 and the answer asks to narrow it to one application.
Licence position
For one application, or every application the user may see, the tool returns:
- per application: the exposure — the over-deployment at the list price plus three years of support, the same figure as the Licence exposure card of the Audit synthesis —, how many components and how many users are over the subscription, and how many components over the subscription have no price;
- per component, the most over-deployed first: the users, the active users, the subscribed quantity and the gap (negative = over the subscription), and its exposure when the component is priced.
The answer says how to read the gap and the exposure, so that the assistant words them the way the screens do. Without any licence usage collected, it says which collection to run (the job nomasx1-collect-all, preset Licences). The figures come from Usage Report, Financial Report and the Audit synthesis.
The lists are cut to what a model can read — 20 risks, 15 role pairs, 15 conflict lines, 40 components — and the answer says how many lines it left out. The full lists stay on the screens, where they can be exported.
The procedures it follows
| Playbook | What the assistant does |
|---|---|
| Run the Security & SoD assessment report | Checks on Last Refresh that the SECURITY and SOD modules ran after the last change (and offers to run the collection), then walks through the security report: its parameters, its findings and their severity, and how to fix a self-conflicting role. |
| Run the licence audit | Checks the inputs — contracts, prices, system codes — and the usage collected, then walks through the licence audit report and the sections to read first. |
| Investigate a segregation-of-duties conflict | Starts from the user's conflict lines, tells a self-conflicting role from a pair of roles, puts the proven conflicts first, and has the collections run again after the change in the source system. |
Starter questions
On an empty conversation in Nomasx-1, the questions are grouped by subject: Accounts (How many users are enabled, and how many have not logged in for 90 days?…), Segregation of duties (Which users carry a segregation-of-duties conflict, and in which process?…), Licences (Which licence components are deployed beyond the subscription, and what is the exposure?…) and Data and reports (When was each module last collected?…).
Five screens have questions of their own, which take the place of the application's:
| Screen | Questions |
|---|---|
| Security → Users | Which users in this list have not logged in for 90 days? · Which users here are privileged, technical or generic? · Explain the status and the dates of the user I clicked |
| Conflicts → Summary | Which process carries the most conflicts? · Which risks are the most critical in this summary? · How do I fix a self-conflicting role? |
| Conflicts → Details | Which role pair explains the conflict I clicked? · Which users share the conflicts in this list? |
| Licenses → Usage Report | Which components are over the subscribed quantity? · Which users count for the component I clicked? · How is the compliance gap computed? |
| Last Refresh | Which module is the most out of date? · How do I refresh the security data? |
Every screen tells the assistant what is on it — the screen, its filters, the row clicked — so that this user or the conflict I clicked needs no explanation. The line under the assistant's header shows what is sent; its × stops it for as long as the user stays on the page.
Who can use it
A tool is offered to whoever reads the screens it stands on, through the menu grants of their roles: the conflicts tool to a role that opens Applications → Conflicts → Details, the licence tool to a role that opens Licenses → Usage Report. Nothing more to grant — the four predefined roles (Auditor, Security officer, Licence manager, Administrator) have both, and the assistant itself needs ai:chat, which they hold as well.
The user's data scope applies as it does on the screens: a user limited to some applications gets the conflicts and the licences of those applications only, and all applications means all those the user may see.
Related
- AI assistant — how the assistant works in Liberty Next.
- MCP — the same tools from Claude Desktop or Claude Code.
- Conflicts — Details · Usage Report · Audit synthesis.