Aito · the predictive database

Calibrated predictions from your live data, in one query.

Generative AI gave software language. Aito is the predictive database that gives it judgment: instant, calibrated decisions learned from your own business data, with a confidence score you can audit and no model to train. The same predictive operators serve your users, your workflows, and your agents.

docker pull ghcr.io/aitohq/aito

Free for development. Production license at usage scale.

Try a live demo →Talk to engineering →Run managed →

<200msquery latency

~16×smaller prompts, same answer

$pcalibrated confidence on every prediction

See the demos → recognize the pattern → talk to engineering

EU-hosted (Ireland) • GDPR-ready • AES-256 / TLS 1.3

Query

{
  "from" : "invoices",
  "where" : {
    "Description": "AWS Cloud"
  },
  "predict" : "Processor",
  "limit": 1
}

Response

{
  "hits": [{
    "$p": 0.908,
    "Department": "IT",
    "Name": "Carol White"
  }]
}

Run this query live

What it means

The intelligent application

Intelligent = Generative + Predictive

The problem.Most products shipped the generative half of AI first: a copilot, an assistant, an agent. The right first move. But fluency is not judgment: on the ten-thousandth invoice, ticket, or order of the month, a per-token, per-second, uncalibrated LLM call is the wrong tool, and there is no confidence number to decide what auto-processes and what escalates.

What's changed.Intelligence has two halves. Generative handles language, exceptions, and conversation. Predictive handles volume: the instant, calibrated, per-row judgment learned from your own data. Intelligent = Generative + Predictive.

The result.An intelligent application runs its routine decisions through the predictive layer at database speed and cost, escalates low-confidence cases to the LLM or a human, and writes every resolution back so tomorrow's judgment is sharper than today's. The same predictions surface as features on every screen of your product and as tools your agents call, and calibrated confidence decides what auto-processes, what routes to review, and what stays manual.

Read the architectural argument: The Predictive Application →

Predictive features on every screen

Pattern Discovery

Statistical patterns ranked by lift, mined live.

Honest Evaluation

The system scores its own predictions.

Demand Forecast

_predict units_sold, seasonality included.

Smart Search

Persona-conditioned re-ranking.

Recommendations

Cross-sell, similar items, for-you.

Smart Forms

Pick one field. The rest predict.

Categorization & Routing

Account, cost center, approver per PO.

Anomaly Detection

Low confidence on a normal field is the anomaly.

Multi-field Prediction

Catalog gap-fill. Predict many fields at once.

Browse all 45 use cases →

What it does

See Aito in Action

Real use cases with actual query syntax and production results

NLP Text Classification

Real-timeanalysis

{
  "from": "prompts",
  "where": { "prompt": "Which payment methods do you provide?" },
  "predict": "answer"
}

Try live demo →

The generative half

The faculty most AI stacks are missing

LLMs gave your agent reasoning. RAG gave it memory. Aito gives it intuition.

A neural network turns experience into instant answers, but its intuition is frozen at training time and learned from the whole internet, not your business. Aito does the same act live, over your own data, with nothing to train: ask about your customers, invoices, tickets, or codes, and it answers in milliseconds with a calibrated sense of how sure it is.

Measured against the standard stack, live and reproducible: ~16× smaller prompts for the same tool shortlist, ~10× faster ticket resolution than chained LLM calls, and structured matching where vector search picks the wrong customer 86% of the time.

A live agent calling Aito ops as tools — predict win odds, estimate effort, query references, recommend outreach — then proposing a confidence-gated email action

The live agent at agent.aito.ai calls Aito ops as tools: win odds, effort, references, the outreach that books the meeting, all calibrated from the firm's own history. It drafts the action and gates the send. Numbers an LLM can't invent.

Reasoning · the LLM. General, deliberate, handles the novel.

Memory · RAG. Recall of what was stored.

Intuition · Aito. The instant, calibrated answer from everything your application has seen.

See the live agent and benchmarks

The predictive half

From Insight to Automation — Analyze → Assist → Automate

Three ways to make every screen of your product intelligent: surface the numbers your users can't compute (Analyze), narrow and ground their choices (Assist), and let decisions run when the prediction is confident (Automate). The same predictive operators serve a product UI and an agent's toolbox alike.

A 360 company dashboard showing six KPIs per segment, each with its root causes and the lever that moves it

📊

Analyze

Make decisions from data — instantly.

Segment customers, forecast outcomes, monitor quality. Start with analytics, evolve into predictive KPIs.

Product analyticsSegmentationMonitoringForecasting

A predictive e-commerce UI re-ranking search results per shopper persona

🎯

Assist

Suggest the next best action — with confidence.

Smart search, recommendations, autofill and routing suggestions. Human stays in control; Aito boosts accuracy and speed.

Smart searchRecommendationsAutocompleteAutofill

A predictive accounting invoice queue with predicted approver and GL code and a calibrated confidence per row

⚡

Automate

Put predictions into production workflows.

Invoice processing, PO routing, classification, tagging. Powers predictive accounting, predictive ERP, and any custom RPA workflow.

Invoice processingPO routingTag predictionAnomaly detection

Apply all three across every screen — that's a predictive application.

Built for Your Industry

The same predictive layer, packaged per vertical, in production since 2018.

Predictive Accounting

Build predictive accounting SaaS — GL coding, approver routing, payment matching, anomaly detection — from one multi-tenant Aito database. 255-tenant reference at accounting.aito.ai. Predictive operators in Nordic enterprise AP production since 2018, and powering RPA partners like Q-Automate at end-customers including Jokiväri.

• Invoice data extraction
• GL code prediction
• Payment term classification
• Fraud detection
• Drop-in tool for your AI agent: GL code, approver, VAT in one _predict

Trusted by:

Learn more →

Predictive E-commerce

Build predictive commerce SaaS — persona-conditioned search, recommendations, demand forecast, inventory with revenue-at-risk-in-€, markdown decisions, churn ranking, win-back with 45× ROI. 16 production-ready predictive features from one Aito database. PetNord reference at ecommerce.aito.ai.

• Smart Search with persona-conditioned re-ranking
• Demand forecast and inventory with revenue at risk in €
• Markdown decisions via _estimate across discount levels
• Churn ranking and win-back with empirical revenue impact
• Recommendation and search ops your assistant can call as tools

Trusted by:

Learn more →

Predictive ERP

Build ERP SaaS where POs route themselves, smart-entry forms predict four fields from one supplier pick, and demand forecasts drive replenishment. 14 production-ready predictive features from a single Aito database. Same predictive operators that have run Nordic enterprise AP automation at 95%+ accuracy since 2018.

• PO queue with predicted account, cost center, approver
• Smart entry with multi-field prediction
• Anomaly detection via inverse prediction
• Demand forecast and inventory replenishment
• Agent-callable routing and smart entry, confidence-gated

Trusted by:

Learn more →

Predictive Application

The pattern across every vertical — a SaaS designed around the premise that the user should never make a decision the data has already made for them. Three open-source reference applications prove the architecture across ERP, accounting, and e-commerce. Your application is next.

• Predictive ERP
• Predictive accounting
• Predictive e-commerce
• Your application next

Learn more →

Try a Real Query Against the Demo Data

For developers — run real predictive queries against Aito's demo database, no sign-up. The shape of the API, visible in seconds.

Invoice Processing

Predict the GL code for a cloud services invoice

Aito Query

{
  "from": "invoices",
  "where": {
    "ProductName": "Cloud Services (AWS/GCP)",
    "TotalAmount": 5000,
    "InvoiceType": "Service"
  },
  "predict": "GLCode"
}

Or run it locally

Same predictive database, in a container on your laptop. No signup, no API key.

install

docker pull ghcr.io/aitohq/aito

docker run -p 9005:9005 ghcr.io/aitohq/aito

Then query it

curl localhost:9005/api/v1/_predict \
  -d '{"from":"invoices","predict":"GLCode"}'

10,000 rows per table, 50,000 rows total — free tier
Same query API as cloud — no code changes when you scale
Runs on your laptop, in your CI, in your VPC

Full self-hosted guide →

In production

20–168ms response (P95 < 200ms) · 10M+ records · 90–98% accuracy · 80% GL coding fully automated at end customers · Nordic enterprise AP since 2018 · 99.9% uptime SLA

Benchmarked vs the standard stack (live, reproducible)

~16× smaller prompts, same shortlist · ~10× faster than chained LLM calls · structured match where embeddings pick the wrong customer 86% of the time

Run the benchmarks yourself →

One engine, no models, nothing to forget

Upload and Query

No training or deployment cycles. Upload your data and start predicting immediately with SQL-like queries. Skip both the ML pipeline and the per-call LLM bill for routine decisions.

Continuous Learning

Real-time adaptation without retraining. Aito automatically learns from new data, eliminating concept drift and model maintenance overhead.

Multiple Predictions

One dataset, unlimited use cases. Predict any field from any other fields without building separate models for each prediction target.

It's crazy how well aito works out of the box for answering specific support question!

Tobias Vogel,

Chief Revenue Officer @Gridpane

Built for Enterprise Scale and Developer Productivity

Discover how Aito's predictive database architecture eliminates MLOps complexity while delivering enterprise-grade performance, security, and 90% cost reduction compared to traditional ML infrastructure.

🛡️Enterprise-Grade Security

Built for production automation: confidence thresholds, evaluation workflows, audit trails, and clear operational boundaries. GDPR compliant with EU data residency.

See what customers are doing with Aito

Predict accounting parameters for purchase invoices

Q-Automate's invoice robot uses Aito to predict the four datapoints (GL code, cost center, VAT %, approver) that otherwise stop the workflow from completing — proven on Jokiväri's Lemonsoft ERP at 80% GL automation.

Learn more

Find similar support tickets and match responses

Gridpane resolves user tech support tickets faster using real-time suggested solutions from Aito.

Learn more

What developers are building

Introducing the Predictive Application
A new architecture for B2B SaaS
Read more
How Aito Scales to Millions of Rows
Buyers ask whether a predictive database can hold up at SaaS scale. Here are the numbers from a 10-million-row invoice routing benchmark, run end-to-end through Aito's HTTP API the way production traffic actually hits it: low-hundreds-of-milliseconds predict latency, sub-linear scaling from 1k to 10M, and what the cold-start looks like before the cache warms.
Read more
Top 3 insights from using RPA+ML to automate invoice processing
ML in RPA 1) can be very rewarding, 2) it requires a different mindset and 3) it can be easy
Read more

Add the predictive half this afternoon.

docker pull ghcr.io/aitohq/aito

Address

Episto Oy

Putouskuja 6 a 2

01600 Vantaa

Finland

VAT ID FI34337429

Contact

About us Contact us Partner with us Join our Slack community

Calibrated predictions from your live data, in one query.

The intelligent application

Predictive features on every screen

See Aito in Action

NLP Text Classification

The faculty most AI stacks are missing

From Insight to Automation — Analyze → Assist → Automate

Analyze

Assist

Automate

Built for Your Industry

Predictive Accounting

Predictive E-commerce

Predictive ERP

Predictive Application

Try a Real Query Against the Demo Data

Invoice Processing

Or run it locally

In production

Benchmarked vs the standard stack (live, reproducible)

One engine, no models, nothing to forget

Upload and Query

Continuous Learning

Multiple Predictions

Built for Enterprise Scale and Developer Productivity

🛡️Enterprise-Grade Security

See what customers are doing with Aito

Predict accounting parameters for purchase invoices

Find similar support tickets and match responses

What developers are building

Introducing the Predictive Application

How Aito Scales to Millions of Rows

Top 3 insights from using RPA+ML to automate invoice processing

Add the predictive half this afternoon.

Address

Contact

Shortcuts