Adjutant AI Domain Workspace
Business questions, answered with panels you can defend
Anyone on the business side asks in plain language. The workspace shows its plan before anything runs, executes under the asker’s own Splunk roles and returns a verified panel with every claim labelled. When the data cannot answer, it says so and names the missing feed.
Splunkbase app 9691. Splunk Enterprise and Splunk Cloud, 9.4 through 10.4.
At first glance it looks like a dashboard. It is not.
A dashboard answers the questions somebody predicted. The Domain Workspace answers the question you actually have, and shows why the answer holds.
- The AI never writes search language and never produces a number. Splunk computes everything, from reviewed searches.
- You see the plan first: which reviewed search, which parameters and what it costs to run.
- Your roles decide what you see. The workspace only promises what the viewer is allowed to see.
- No confident guesses. When the data cannot answer, it names the missing feed instead of showing an empty panel.
How an answer is made
Ask
A business user asks in plain language, in their own terms.
See the plan
The workspace picks a reviewed search and shows the parameters and the cost before running.
Run under your roles
Splunk executes the search with the asker’s own permissions.
Get a verified panel
Every claim is labelled as fact, calculation or interpretation.
Where it stands today
Released before the code on purpose, so the people who will use it can shape it. Here is exactly what exists and what comes next.
Now: clickable preview
A guided three-minute tour with fictitious data for a bicycle retailer’s finance domain. It runs no searches, stores no data and calls nothing external.
Next: narrated walkthrough
A video that walks through the concept end to end, including how answers are verified.
Then: the running workspace
The working Domain Workspace, running on your own Splunk under your own roles.
If you run a SOC, you already know this
The first preview domain is finance, because fraud makes a vivid demo. The engine is domain-generic by design: a security domain runs on the same runtime.
- The domain model is your use case library made executable, with entities and relations instead of a wiki page.
- The question bank is the little sibling of your detection catalogue, for investigation questions, proven nightly.
- The interactive report is your triage queue, with stable work items instead of a spreadsheet export.
- Per-viewer honesty is your RBAC, finally reflected in what the assistant is willing to promise.
Tell us which claim you do not believe
That is the most useful feedback you can send. Criticism, questions about your own domain and design-partner interest all go straight to the developer.