Five Live Demos
Three predictive applications, an agent demo, and a SQL demo. One predictive database underneath.
Open the demo closest to your domain, or scroll the use cases by capability. Same predictive operators across all five open-source demos. The three application demos showcase predictive ERP, accounting, and e-commerce in production-ready shape; the agent demo at agent.aito.ai puts the same operators in a live agent's toolbox and benchmarks them against the standard stack; the SQL demo at sql.aito.ai asks them as predict(), relate(), and recommend() inside a SELECT. More demos are at demos.aito.ai.
Five Live Demos
Three predictive applications, one agent demo, and one SQL demo. Each runs against a live Aito instance. Open the live demo to click through it, read the source on GitHub, or browse the full feature inventory on the per-demo page.
Predictive ERP
14 production-ready use cases across industrial maintenance, multi-channel retail, and professional services. 72% aggregate automation rate on the mixed profile.
Predictive Accounting
Multi-tenant by construction. 255 customer companies, 128K invoices, one shared Aito instance. Same operators behind 95%+ accuracy Nordic enterprise AP automation since 2018.
Predictive E-commerce
16 views on a 110K-row PetNord pet-store dataset. Smart search, recommendations, demand forecast, win-back, plus a deliberate honest-failure case so calibration shows through.
Predictive Agent
A live gpt-5-mini agent that calls Aito ops as tools: win-odds, effort, references, grounded numbers an LLM can't invent. The benchmarks behind the headline figures run live against the standard stack: ~16× smaller prompts for the same shortlist, ~10× faster than chained LLM calls, and structured matching where embeddings pick the wrong customer 86% of the time.
Predictive SQL
A 360° view of a machinery vendor's business, asked entirely in SQL: predict(), relate(), and recommend() inside a SELECT, so existing Postgres tools connect and work unchanged. Six cards give the root cause and the lever per question, one fails on purpose, and a confounded channel is cleared under conditioning. Behind them: an 827-cell driver map and a click-to-narrow explorer.
Twelve Use Cases, Organized by Capability
A sampling of what predict, relate, search, recommend, estimate, and evaluate do in production-ready shape. Each card links to the demo that showcases it best. The full catalog is at /use-cases/. The same operators plug into an agent stack as tools; see the live agent at agent.aito.ai.
Analyze
Pattern discovery, calibration, forecasting, segmentation. The decisions a user makes after seeing the data.
Pattern Discovery
🛒 E-commerceStatistically significant co-occurrence and persona-affinity patterns, ranked by lift. Dog dry-food cross-sells to dental treats at 2.72× in the PetNord fixture, mined live.
See in demo →Source →
Rule Mining
🧾 AccountingDiscover statistical patterns (category=telecom & gl_code=6200 → approver=Timo, 15.8× lift). Review each with its confidence and evidence; promote it to a rule or dismiss it.
See in demo →Source →
Honest Evaluation
🛒 E-commerceSystem scores its own predictions on held-out data. Return Risk reports +0.0 pp gain over baseline and renders as a red row.
See in demo →Source →
Demand Forecast
📋 ERP_predict units_sold blended with seasonality factors from same-month historical data. Drives replenishment without a separate forecasting model.
See in demo →Source →Assist
Search, recommendations, smart forms, conversational. The human stays in control; the system boosts speed and accuracy.
Smart Search
🛒 E-commercePersona-conditioned re-ranking. Same query returns different rankings for different shopper segments — derived from the data, not from a curated rule.
See in demo →Source →
Recommendations
🛒 E-commerceCross-sell ranked by lift, for-you tile rankings, bought-together pairs. All from the same predictive operators against the same data.
See in demo →Source →
Smart Forms
📋 ERPPick a supplier and four fields predict in parallel — cost center, account code, project, approver. Tab to accept, Esc to reject.
See in demo →Source →
Conversational Help
🛍 Grocery referenceCTR-ranked help articles via _recommend against click history. Users read what other users with similar queries clicked.
See in demo →Source →Automate
Categorization, routing, anomaly detection, multi-field prediction. Calibrated confidence decides what auto-processes and what routes to human review.
Categorization & Routing
📋 ERPEvery purchase order arrives with predicted account code, cost center, and approver — three confidence tiers, bulk-approve for rule-matched rows.
See in demo →Source →
Invoice Processing
🧾 AccountingGL code, approver, payment method, cost center predicted per invoice with $why factor decomposition. The full AP automation surface.
See in demo →Source →
Anomaly Detection
📋 ERP · 🧾 AccountingInverse prediction — low confidence on a normally-predictable field is the anomaly. Mis-coded account, amount spike, unknown vendor; surfaced before posting.
See in demo →Source →
Multi-field Prediction
📋 ERPCatalog gap-fill on workflow-blocking products. Predict category, HS code, unit price one-shot across the catalog. Same pattern as smart forms, applied to data rather than input UI.
See in demo →Source →Implementation Details & Use Cases
Explore the complete source code, technical documentation, and performance benchmarks
{
"from": "impressions",
"where": {
"context.user": "larry"
},
"recommend": "product",
"goal": { "purchase": true }
}
Why Engineers Choose Aito
Production-ready AI without the complexity of traditional ML infrastructure
SQL-Like Queries
- Familiar syntax for predictions
- No feature engineering required
- Real-time results
Zero MLOps
- No model training/deployment
- Automatic schema inference
- Self-improving accuracy
Measured at scale
- Latency by data size, measured and published: aito.ai/docs/api/v2/benchmarks/scaling
- Enterprise security
- Single node today; clustering in development
Evaluate in 15 Minutes
Complete technical evaluation path for engineering leaders
Address
Episto Oy
Putouskuja 6 a 2
01600 Vantaa
Finland
VAT ID FI34337429
