01
Descriptive
What happened?
“How did monthly revenue trend this year?”
- Aggregations, trends and comparisons
- The right chart chosen automatically
- Plain-language summary of the movement
Sovereign AI Platform · Analytics & Forecasting Agent
Data Science Agent for Forecasts and Root Causes Without SQL
Ask questions in plain language and get governed queries, forecasts and root causes, with every analysis running on the Sovereign AI Platform.
Data Science Agent · Workspace
Sample
Illustrative sample. Figures, table names and results are demonstration values and will vary by environment.
Agent Workflow
The Data Science AI Agent at Work
Follow the Data Science Agent from the first question through analysis and audit to see how an answer is produced. Each step runs inside the sovereign boundary.
Data Science Agent · Run trace
Inside boundary
FDShow revenue by region with trend analysis
1 / 8
Illustrative sample. Table names, figures, response time and confidence score are demonstration values and will vary by environment.
Levels of Analysis
Look Ahead with an AI Forecasting Agent
Connect data trends to their root causes quickly using an intelligent AI forecasting agent.
02
Diagnostic
Why did it happen?
“What drove the jump in September?”
- Anomaly detection on every trend
- Drill-down to the dimensions behind a change
- Driver contribution in plain numbers
03
Predictive
What happens next?
“Where will revenue land by year end?”
- Statistical modelling and forecasting
- Confidence ranges around every projection
- Assumptions stated alongside the result
Illustrative sample values.
Transparent by Design
Inside the Data Science Agent’s Decision Path
Give auditors the concrete proof they require by deploying agentic AI for data science.
Answer
APAC revenue rose 34% in Q4, driven largely by one new $8.2M contract. Americas and EMEA grew a steadier 3% and 6% quarter on quarter.
Illustrative sample values.
The exact query that produced the numbers
SELECT r.region_name, d.fiscal_quarter,
SUM(f.net_revenue) AS revenue
FROM finance.revenue_fact f
JOIN dim_region r ON f.region_id = r.region_id
JOIN dim_date d ON f.date_id = d.date_id
WHERE d.fiscal_year = 2026
GROUP BY r.region_name, d.fiscal_quarter;
Generated by the sovereign SLM, validated against the schema and executed read-only.
Where every figure came from
finance.revenue_factNet revenue by transaction · permissions inherited
dim_regionRegion hierarchy · permissions inherited
dim_dateFiscal calendar · permissions inherited
dim_customerUsed for drill-down only · names masked
Sources are read in place through the governed data layer. Nothing is copied outside the enterprise.
How the analysis was performed
- Aggregate net revenue by region and fiscal quarter.
- Test each quarter against the trailing trend to flag unusual movements.
- Drill down into the flagged quarter by customer to find the driver.
- Forecast the next quarter with a stated confidence range.
The method is recorded with the result, so the same analysis can be repeated next quarter.
What was verified before delivery
✓ Regional totals reconcile to general ledger control totals
✓ Query re-run returned identical results
✓ “APAC grew 34% quarter on quarter” supported by the data
✕ “Growth was broad-based across APAC customers” contradicted by drill-down and removed
Question, SQL, sources, model version and checks are written to the immutable audit log.
Question Gallery
Team Queries Handled by AI Agents for Data Science
Deliver insights based strictly on user access levels through AI agents for data science.
Why did operating costs exceed budget in Q3?
Returns · variance drivers chartWhat is next quarter's revenue forecast by region?
Returns · forecast with rangeWhich product lines drove the margin decline?
Returns · ranked contribution tableWhich lines had the most unplanned downtime last month?
Returns · ranked bar chartWhat explains the drop in output at the northern plant?
Returns · root-cause breakdownHow has first-time fix rate changed by region this year?
Returns · trend lines by regionWhich SKUs are likely to stock out in the next six weeks?
Returns · risk-ranked SKU listHow accurate was last quarter's demand forecast?
Returns · forecast versus actualWhich suppliers show worsening lead times?
Returns · supplier trend tableWhich customer segments show rising churn risk?
Returns · segment risk scoresHow has average discount affected win rate this year?
Returns · correlation analysisIs pipeline coverage sufficient for the Q4 target?
Returns · coverage against target
Where It Connects
AI Agents for Data Science within the Sovereign Boundary
Deliver insights and statistical forecasts directly into workflows while keeping data within the sovereign boundary.
Sovereign boundary
Enterprise data
SQL databases
Data warehouses
ERP and procurement
CRM
HR systems
ITSM and ticketing
Data Science Agent
Privacy gateSensitive columns masked or aggregated
SQL generationSovereign SLM on enterprise GPUs
Statistics and forecastingRun inside the perimeter
VerificationReconciled, checked, logged
Into the business
Charts and dashboards
Forecasts with ranges
Written insight
Update a record
Raise a ticket
Send for approval
Backlog to Self-Service
Natural Language to SQL Ends the Data Bottleneck
Eliminate the traditional cycle of waiting for data extracts with natural language to SQL.
Typical request cycle
- Question raised
- Ticket to analytics team
- Waits in the queue
- SQL and data extract
- Report built and reviewed
- Follow-up question restarts the cycle
With the Data Science Agent
- Question asked in plain language
- SQL, chart and insight returned, verified
- Follow-up asked in the same conversation
- Decision made with evidence attached
Pairs with the Research Agent for market context and the Knowledge Agent for the policies behind the numbers.
Part of the Sovereign AI Platform
A Fully Configured Data Science AI Agent
The AppsTek Sovereign AI Platform provides direct access to the existing data layer without complex external integrations.
Platform layerWhat the Data Science Agent uses it for
Connectors and governed data layerQueries SQL databases, warehouses, ERP and CRM where the data already lives
Privacy gateMasks or aggregates sensitive columns before analysis
Knowledge layerHolds business definitions so "revenue" means the governed finance metric
Sovereign small language modelWrites SQL and narrative on enterprise GPUs, with no public model APIs
Agent runtime and approvalsRuns statistics and forecasts, then writes results back to workspaces and tickets
Immutable audit log and key custodyRecords every question, query and result, encrypted with enterprise-held keys
FAQ
Everything You Need to Know About the Data Science Agent
What is a Data Science Agent?
The Data Science Agent is an AI agent for analytics, forecasting, and root-cause analysis on the AppsTek Sovereign AI Platform. Business users ask questions in plain language, and the agent uses natural language to SQL to query enterprise data, return charts and statistical analysis, identify drivers, and generate forecasts with a written explanation.
How does a Data Science AI Agent work?
A Data Science AI Agent turns a business question into a structured analysis. It interprets the request, identifies the data it needs, generates SQL, runs the query against governed enterprise data, verifies the results, and returns the analysis. The generated SQL remains visible for review.
Do business users need to know SQL?
No. Natural language to SQL lets users ask questions in everyday language while the agent generates the underlying query. The SQL stays visible alongside the result, giving analysts and data teams a clear view of how each number was produced.
What data sources can a Data Science Agent use?
The Data Science Agent can work with enterprise SQL databases and data warehouses, as well as ERP, CRM, HR, and other systems connected through the governed data layer of the Sovereign AI Platform.
Can AI agents for data science work with sensitive enterprise data?
AI agents for data science on the Sovereign AI Platform operate within the enterprise environment. Source permissions apply to every query, while sensitive data can be masked or aggregated before analysis. Queries, models, and results remain within the defined sovereign boundary.
Does enterprise data leave the organization during analysis?
Queries, models, and results run inside the enterprise environment, whether deployed on premise, in a private cloud tenancy, or on a pre-built appliance. Public model APIs are blocked by design, keeping enterprise data and analytical results within the sovereign boundary.
How does the Data Science Agent verify its results?
Before delivering an answer, the Data Science Agent reconciles results against source tables, checks claims in the narrative against the underlying data, and removes claims that the data does not support. The query, data sources, and model version are recorded in the audit log.
Can agentic AI for data science explain why a number changed?
Yes. Agentic AI for data science can move beyond descriptive analysis to investigate the factors behind a change. The agent can drill into relevant dimensions, identify contributing factors, and explain the findings in plain language.
Can the Data Science Agent forecast future results?
Yes. The AI forecasting agent uses statistical modelling and historical data to project future trends. Forecasts include confidence ranges and stated assumptions so teams can understand both the projection and the conditions behind it.
Does the Data Science Agent replace BI dashboards?
The Data Science Agent complements existing BI dashboards. Dashboards provide established metrics and reporting, while the agent handles follow-up questions such as why a metric changed, what drove the movement, and what could happen next.
Built on the Sovereign AI Platform governance ring
Data sovereigntyRegion pinned, inside the perimeter
Inherited accessSource permissions on every query
ExplainabilitySources, method, model version
Privacy by defaultPII masked before analysis
Enterprise keysHeld in the enterprise HSM
Immutable auditEvery run recorded once
Deploy air-gapped on premise, in a private cloud tenancy or on a smart appliance. Explore the Sovereign AI Platform ➜
See It Live
Put Data to Work Inside the Perimeter
See a natural language to SQL walkthrough from question to forecast.