Built for operational intelligence
Work with forecasts, anomalies, risk signals, extracted fields, and document insights across practical AI workflows.
Operational AI
Turn Operational Data Complexity
Into Explainable Operational Intelligence
Operational AI helps teams turn operational data, documents, and business signals into practical AI workflows for forecasting, anomaly detection, fraud detection, data extraction, and document intelligence.
Explore Operational AI through the Pro Console and choose the workflow that fits your operational challenge.
Need an AI workflow for unique data, documents, risks, or operational processes? Talk to us
What is Operational AI?
Operational Intelligence Built for Real Business Signals
Operational AI is MAG OptiAI's AI solution family for operational data, business signals, and document-driven workflows. It helps teams turn forecasts, anomalies, risk indicators, extracted data, and document insights into explainable, decision-ready outputs.
It is built for environments where teams need clearer forecasting, earlier signal detection, faster document understanding, and practical AI support without building a custom AI system from scratch.
Instead of treating operational data as scattered reports or manual analysis tasks, Operational AI turns it into a guided workflow. Teams can upload data or documents, choose the right workspace, review outputs, inspect evidence, and evaluate what the result means for the business decision in front of them.
The AI complexity runs behind the scenes, while users stay focused on the signals, risks, forecasts, documents, and decisions they need to understand.
AI made usable
Use forecasting, detection, extraction, and document intelligence through guided workspaces without starting from a blank technical project.
From signals to decisions
Review outputs, evidence, metrics, and model context before deciding what the result means for the workflow in front of you.
Useful for operations and learning
Support real business workflows while helping teams understand applied AI through practical, visual, and explainable experiences.
Operational AI products
Operational AI Products Built for Real Operational Workflows
Our five Operational AI products help you turn operational data, business signals, and documents into explainable, decision-ready outputs.
GenAI
Document Intelligence
DocAI / Knowledge Assistant
Ask grounded questions over project documents with supporting evidence.
Create a project, upload documents, ask natural-language questions, and review answers with citations, retrieved excerpts, conversation history, and exportable evidence.
Project-scoped knowledge baseCreate DocAI projects and keep documents, conversation turns, and evidence scoped to the selected project.
Document upload and indexingUpload project files and track readiness before asking grounded questions.
Grounded document Q&AAsk natural-language questions and receive answers grounded in ready project documents.
Citations and source excerptsReview supporting excerpts, source snippets, page/context anchors when available, and answer scope.
Persistent conversation historyRevisit saved conversation turns and export selected answer context for review.
ExtractAI
Turn operational documents into structured JSON and reviewable evidence.
Upload a document, run a persisted extraction, inspect structured fields, review raw evidence, and export JSON or source text for downstream workflows.
Project-based document extractionUpload documents into an ExtractAI project and run persisted extraction jobs.
Structured JSON outputReview extracted fields in readable form or inspect and download the raw JSON payload.
Raw evidence reviewInspect source text evidence when returned and understand input coverage notes.
Saved extraction runsOpen prior extraction runs as reference results and compare what was extracted.
Export-ready resultsCopy or download JSON and source text outputs for operational handoff.
Machine Learning
ForecastingFraudAnomaly
Forecasting
Train saved time-series models and generate planning-ready forecasts.
Create a project, train or stage a saved model, generate forecasts, review model metrics, inspect forecast charts and tables, and export results.
Time-series data validationLoad historical series, validate required columns, and summarize dataset readiness before training.
Saved model lifecycleTrain, stage, compare, activate, archive, and delete saved forecast model versions.
Forecast generationGenerate forecast points and intervals from a staged or selected saved model.
Metrics and provenanceReview model family, holdout evaluation, WAPE/RMSE/sMAPE/MAPE context, and run details.
AI summary and explanationGenerate bounded AI summaries or explanations grounded in forecast outputs and model context.
Exportable charts and tablesExport forecast charts, forecast tables, and JSON result context.
Fraud Detection
Train risk models, score rows, and review threshold trade-offs.
Train saved fraud models, score operational rows, inspect probabilities and evaluation evidence, tune threshold policy, and export scored results.
Training and scoring workflowsPrepare training data, train saved models, and score rows against staged or selected versions.
Multiple model familiesUse supported model families such as LightGBM, Logistic Regression, Random Forest, and XGBoost.
Threshold policy reviewReview threshold candidates, precision, recall, F1, alert rate, and operating trade-offs.
Evaluation evidenceInspect saved-model metrics, stability context, calibration notes, and feature evidence where available.
AI summary and explanationAsk bounded questions grounded in visible score distributions and saved-model evidence.
Scored-row exportsExport scored rows and JSON result context for downstream review.
Anomaly Detection
Detect abnormal time-series points and review what needs attention.
Load time-series data, choose a compatible detector, score anomalies, review charts and ranked points, ask AI-assisted questions, and export results.
Detector selectionChoose compatible detectors for univariate, contextual, or multivariate anomaly review.
Data readiness checksValidate time-series inputs and detector compatibility before scoring.
Anomaly scoringScore time-series points and review anomaly count, flagged share, threshold, and detector context.
Chart and point reviewInspect anomaly charts, ranked flagged points, scores, timestamps, and expected baselines when available.
AI-assisted interpretationAsk operational questions grounded in anomaly scores, point lists, and detector metadata.
Saved run history and exportsOpen saved scoring runs and export anomaly charts, CSV rows, or JSON results.
All Operational AI products run inside the Pro Console with guided workspaces, reviewable outputs, and AI assistance where supported.
Operational AI features
What Operational AI Helps You Do
Operational AI brings product selection, project setup, data and document intake, AI execution, result review, explanation, and export into one guided Pro Console solution area.
Shared Operational Intelligence Experience
One solution area. Five products. Clearer AI-supported decisions.
01
Product-Guided Workspace Selection
Choose the Operational AI workspace that matches the signal, model, document, or review task.
02
Project and Input Organization
Keep datasets, documents, saved model versions, extraction runs, scoring history, and conversations scoped to projects where supported.
03
Data and Document Intake
Upload CSVs, structured time-series data, scoring rows, or documents through product-specific intake flows.
AI
04
AI and Model Execution
Train models, stage saved versions, score rows, detect anomalies, extract fields, or ask grounded document questions.
05
Reviewable Results
Inspect charts, tables, metrics, evidence, extracted JSON, source excerpts, and run provenance.
AI
06
AI-Assisted Interpretation
Use bounded AI summaries or explanations where supported, grounded in visible outputs and saved context.
07
Export and Handoff
Download forecast tables, scored rows, anomaly outputs, structured JSON, source text, or evidence payloads for review.
One shared Operational AI experience across product workflows
Built to help teams upload, run, review, explain, and export with more confidence.
AI-Assisted
Interpretation is grounded in visible outputs.
Evidence-Aware
Review provenance, excerpts, metrics, and run context.
Built for Operators
Designed for practical business and operations teams.
Pro Console access
Start in the Pro Console. Scale for Real Operations.
Operational AI products run inside the Pro Console with guided workspaces and AI-powered tools. Start free to explore and evaluate, then scale when you need more capacity, usage, or operational support.
01
Choose Your Product
Select the Operational AI product that fits your operational challenge and the type of output you want to generate.
You can
- Forecast trends
- Detect anomalies
- Score fraud risk
- Extract documents
- Ask questions on documents
02
Explore with Guided Workspaces
Follow the product-specific workspace to set up your project, add data or documents, and prepare the right AI task.
You can
- Create a project
- Upload data or documents
- Configure the workflow
- Prepare models or detectors
- Index documents for search
03
Evaluate with Free Access
Run workflows with free access to experience outputs, explanations, and evidence-backed results.
You can
- Run forecasts or scoring
- Detect anomalies
- Extract structured fields
- Ask questions and review evidence
- Review AI explanations where supported
04
Scale for Real Operations
When you are ready, scale with higher usage, broader product access, collaboration, and operational support.
You can
- Unlock higher limits
- Use more projects and saved runs
- Collaborate with your team
- Prepare outputs for pilots
Built to start simple and grow with your operations.
Use Operational AI anywhere, anytime, with more projects, data, runs, and team needs. Scale when you are ready.
Need a custom or enterprise setup?
Contact usUse cases
Real challenges. Real models. Real results.
Explore how Operational AI helps teams extract knowledge, forecast demand, score risk, and detect unusual behavior with practical model workflows.
Multi-Document Contract Intelligence Room
See how DocAI indexed five contract files, answered renewal, security, SLA, and amendment questions, and kept every answer grounded in project evidence.
Contract approval slows down when obligations are scattered across multiple files.
Legal, procurement, finance, and operations teams needed to review a master services agreement, statement of work, SLA exhibit, data processing addendum, and amendment without losing the source trail.
Problem
Renewal terms, service credits, security obligations, and amendment changes were spread across five contract files.
Operational AI workflow
Index five documents, ask focused contract questions, and review grounded evidence for each answer.
Outcome
Exact terms surfaced with evidence: 120-day notice, 99.5% uptime, 72-hour incident notice, and credit caps.