Four Live Demos, Twelve Use Cases
Three predictive applications and an agent demo. One predictive database underneath.
Open the demo closest to your domain, or scroll the use cases by capability. Same predictive operators across all four 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. More demos are at demos.aito.ai.
Four Live Demos
Three predictive applications and one agent 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.
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). Promote to rules with audit trail; dismiss to record the decision.
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
Production Ready
- 20-168ms range (P95 < 200ms)
- Enterprise security
- Horizontal scaling
Evaluate in 15 Minutes
Complete technical evaluation path for engineering leaders
Address
Episto Oy
Putouskuja 6 a 2
01600 Vantaa
Finland
VAT ID FI34337429
