Active IA: your company's enterprise AI hub
A single portal where managers and teams ask questions of the data in natural language, receive ready-made briefings and are notified when a KPI moves outside its normal range. Active IA queries the Power BI, databases and documents the company already has, respects each person's permissions and shows the source of every answer.
Architecture approved by the security, legal and AI ethics committees of one of the world's largest technology companies
Chat answers. Journal anticipates. Signal alerts.
Three modules on the same source, the same business vocabulary and the same access rules. None of them recalculates the official KPI.
Ask and get the answer right away
Ask the way you would ask an analyst: “why did margin drop in October?”. The answer gives the number, what explains the change and where the data came from.
Who it is for: people who need to decide now and do not want to open several reports or wait for someone to build the number.
Revenue closed below target, driven by three regions. South and Northeast lost volume with their two largest customers; Central-West had delayed invoicing on orders already approved.
| Region | Attainment | vs. prior month |
|---|---|---|
| South | 92% | −6% |
| Northeast | 84% | −12% |
| Central-West | 79% | −15% |
The business readout ready before the meeting
On the schedule each area chooses, a summary arrives with what changed in the KPIs, why it changed and what deserves attention. For example: every Monday at 7 a.m. for the executive team, every day for operations.
Who it is for: directors and managers who today depend on someone to prepare the meeting material.
Revenue for the week was 3.8% above the previous week, with stable margin. Growth came from repeat purchases by active customers; new customer acquisition slowed for the second week in a row.
- What changed: average order value rose 5% in the distributor channel.
- Watch item: average collection period went from 41 to 46 days.
- For the meeting: review the quarterly new customer target.
Alerts you before the problem reaches month-end close
When a KPI moves off target, the owner receives an alert with the number, the comparison with history and what to check first.
Who it is for: people accountable for a KPI who today only find the deviation at month-end close.
South Unit margin came in 3.6 points below target and outside its historical range. The drop began in the second week of the month.
- Evidence: freight cost per order rose 18% on inland routes.
- What to check: discounts above approval limits in three renewed contracts.
Illustrative screens with fictitious data.
One question, one answer with its source
A manager asks "Why did the South unit's margin drop this month?". Behind the scenes, the platform follows the path that today takes days.
Understands the question
Identifies the subject, the KPI and the period, using the company's vocabulary.
Chooses the source
The orchestrator calls the right agent: semantic model, database or document.
Queries with the permissions of whoever asked
The query runs under the user's identity. If someone cannot see a piece of data in the report, they cannot see it in the answer either.
Answers with the evidence
Number, comparison, main driver and the query attached, with a shortcut to the dashboard.
Active Agents API: one hub for agents from any platform
The company does not have to choose a single AI vendor. The Active Agents API connects to Active IA the agents and models that already exist or that come later.
- Microsoft Fabric and Copilot Studio agents. Agents built in the Microsoft ecosystem appear in the same portal, under the same governance.
- Databricks Genie. Questions about Unity Catalog data answered by the same hub, alongside the other sources.
- Azure OpenAI, OpenAI, Claude and Gemini. Each task uses the most suitable model, including the one the company already licenses, and switching does not require rebuilding the solution.
- Active BI agents. Agents by source, by area and by routine, built on the same architecture and added without a new deployment.
Observability and governance from the first agent
The list below was reviewed and approved by the security, legal and AI ethics committees of one of the world's largest technology companies, where the platform runs in production.
- Inherited permissions. Corporate sign-in with Microsoft Entra ID and Power BI RLS applied per user.
- Data stays in place. Sources remain where they are. The platform queries the data; it does not copy it.
- No training on your data. Models consumed through an enterprise API, under a contract that excludes using the data for training.
- Every answer logged. Question, source, query executed, model, time and cost, with an audit trail.
- Cost and quality by area. Dashboard with usage, accuracy and cost by agent and by area, reviewed every month.
- Human decision. The platform informs and recommends. The decision stays with the person responsible.
Full controls are described in Secure and governed AI.
Where Active IA runs
| Environment managed by Active BI | In your company's environment | |
|---|---|---|
| Business data | At the company's source. The platform queries, it does not copy. | At the company's source, everything inside the perimeter. |
| Identity and sign-in | The company's Entra ID, no separate password. | The company's Entra ID, no separate password. |
| Who operates it | Active BI, with a monthly dashboard and support. | The company's infrastructure or Active BI, as set in the contract. |
| When it makes sense | Proof of concept, first area, companies without a dedicated infrastructure team. | Perimeter policies, regulated data and operation at scale. |
Agents, rules, permissions and history move from one model to the other without rebuilding.
How to start: one area, one source, one decision
No one needs to approve a large project to use Active IA. Each stage has a fixed scope and a decision by the company.
Proof of concept
Two to four weeks: one agent on a real source, in the chosen area, with accuracy criteria agreed before it starts.
Evaluation
The criteria are measured with the people who used it. The company decides to continue, adjust or stop.
Deployment
The platform runs in the chosen environment, with corporate sign-in. The proof of concept agent carries over without being rebuilt.
Expansion
New agents, sources and alerts by area, on the same platform.
Want to understand Active BI's full approach to AI? See Artificial Intelligence for enterprises and AI Agents.
Frequently asked questions about Active IA
What is Active IA?
Does Active IA work with the Power BI we already have?
How is it different from Copilot?
We already use other AI solutions. Do we need to replace them?
Is our company's data used to train models?
How long does it take to get started?
See Active IA with questions from your business
Our team runs the demo with sample data. Bring the question your executive team asks every week and we will show you the path to the answer.
- You describe your situation in two quick steps
- A specialist replies within 30 minutes during business hours (Brasília time)
- A 30-minute conversation to understand your situation and recommend a path
No commitment and no follow-up call sequence.
Schedule a demo
2 steps, under 40 seconds
Step 1 of 2