Included with every capability
The agent-ready foundation under every capability: context, business rules and trusted data in one governed model.
See Data StudioPower BI, conversational analysis and an open API, all reading the one model.
See AnalyticsThis page
A digital workforce you can audit, working on your own governed model. Included, not an add-on.
AI Studio uses AI to explain your technology cost. The AI Economics capability measures what AI itself costs and returns.
Why it matters
Most AI in finance software works from an export and gives you a plausible answer about money, so someone in finance still has to check it. That is the work the AI was meant to save.
AI Studio works the other way around. The engine calculates every figure from the governed model, and the workforce reads those figures, prepares the work and explains it. Workers stand on the context Data Studio has already built: what each application is, who consumes it and what it costs. A data platform on its own can give an AI your rows. The model is what tells it what those rows mean.
So the work can be checked. The numbers came from the engine. The brief it followed is a saved method, not a prompt someone typed once.
No language model ever produces a number on this platform.
Meet the workforce
Every team builds its own workers. These are examples from a FinOps practice, written the way they were briefed.
Compares the newest settled day with a like-for-like day and names the services that moved.
ExampleVirtual machines down 30% on the day, licenses to zero alongside them
Watches every reservation and savings plan for low use and coming expiry, so renewals are deliberate.
ExampleOne reservation 12.5% used, ending in five weeks: let it lapse
Finds non-production environments still running at weekends and sizes the wind-down.
ExampleDevelopment and QA running at full weekday cost on Saturdays
Patrols ownership tags for duplicates, blanks and case splits, and names the resources to fix.
ExampleTwelve clients split by capital letters, so their cost lands in two places
Turns the week’s movements, savings and risks into one page for the leadership team.
ExampleSpend on track, two commitments to decide, one new workload without an owner
The workforce does the preparation. Your people make the calls. Workers prepare, check, measure, explain and report. They never post a charge, approve anything or change the model.
No worker produces a number. Every figure in their work comes from the model’s own calculation, the same one behind your reports and chargebacks.
What AI Studio does
The work
Choose who does it, give it a brief, set when it runs and who hears back. A daily spend check at six, a commitment watch every Monday, a variance pack before the month-end meeting. The work is ready before anyone asks for it.
Release 6.2A skill is a saved brief in your own words. Publish it across the organization and an analysis someone worked out once becomes a method every worker runs the same way.
A single brief can draw on more than one Specialist. Ask for the variance on a cost center and it reads actuals from one and the budget from another, then returns the difference with drill-down.
Each scheduled run is saved as a conversation, so every piece of delivered work has a permanent home you can open, read and trace back.
Release 6.2Who is on the team
The Model Assistant finds where cost stops flowing, names the root cause and drafts the driver rule for you to apply. The FinOps Assistant covers cloud, commitments and unit cost. The Planning Assistant covers budgets, forecasts and scenarios.
Specialists go deeper than the Assistants, one for each standard object, answering at field level. Each is narrow on purpose, so every answer is small enough to check and says where it came from. Specialists you build for your own use cases arrive in Release 6.2.
AI Enrichment runs inside Data Studio as data arrives: it classifies ledger lines, normalizes vendor names and flags AI spend, before anyone opens a report.
The Assistants carry TBM and FinOps knowledge. Ask what a rolling forecast is and it explains the method. Ask what your rolling forecast is and it reads your model.
Built to be trusted with money
Every figure comes from the governed model. A data platform on its own can give an AI your rows. The model is what tells it what those rows mean.
Workers prepare, check, measure, explain and report. They don’t change the model, post a charge or approve anything. Your people make the calls.
Through the Model Context Protocol, the AI tools your teams are building can call the same Assistants, Specialists and briefs, on the same governed numbers. We enable the connection for you.
On requestThe workforce is only as complete as the model it reads. When something a job needs isn’t in the model, such as a contract end date that was never loaded, the worker says so and tells you what its answer is leaning on instead. Workers read the model; they don’t write to it.
Or ask in a sentence
A workforce you can audit
The numbers came from the engine, not from a language model. The brief a worker followed is a saved, published method, not a prompt someone typed once. And the work sits on the same governed model as everything else in MagicOrange.
AI Studio isn’t a separate pipeline that copies your data somewhere else to be analyzed, so the questions your security team asks about the platform are the questions to ask about the workforce too.
Analyst view
What it covers
