Autonomous Manufacturing Intelligence — AI agents that auto-discover and connect fragmented factory IT/OT systems, then find the inefficiencies

THE PROBLEM

Manufacturing companies run on fragmented systems: ERP, MES, machine and robot controllers, CNC software, databases, spreadsheets. Data sits disconnected between them. Before you can analyze a factory's processes or optimize anything, you first have to integrate these systems — a technically hard, slow, expensive IT project. Many manufacturers simply lack the engineers or budget to do it. So factories keep operating blind.

OUR SOLUTION: AUTONOMOUS MANUFACTURING INTELLIGENCE

We build an agent-based integration and process-optimization platform:

• Automatically identifies the systems a factory uses and their integration capabilities

• Analyzes APIs, data structures, file formats and other access paths

• Generates the integration code and data transformations

• Connects heterogeneous systems into one data ecosystem

• Monitors the integrations and adapts when systems or schemas change

Once connected, AI agents analyze the production process as an interconnected system, not isolated KPIs:

• Detect bottlenecks and determine their root causes

• Analyze downtime and recurring failure patterns

• Assess utilization of machines, workers and production lines

• Find correlations between defects, machine parameters, materials, shifts

• Predict upcoming disruptions and capacity shortfalls

• Simulate what-if scenarios before implementing changes

• Generate concrete, data-verified recommendations

The key difference from dashboards and BI: the platform forms hypotheses about why inefficiencies happen, tests them against data, and recommends what to change. Not just "what happened" — "why, and what to do about it."

THE CORE R&D CHALLENGE: ONBOARDING

The main research goal is turning integration from a months-long custom IT project into a standardized, largely automated process. The system should understand a new factory's process structure, key KPIs and data relationships without lengthy manual analysis. This is what makes the platform scalable across factories — and it is genuinely unsolved today.

ALREADY IN PRODUCTION (basis for this project)

We are not starting from zero. We work with numerous manufacturers and have live production builds:

• Order intake agent — running live in production, connecting to ERP systems

• Quoting automation — production use case at a façade manufacturer, 100% delivered

• Blueprint/CAD analysis engine — parses engineering blueprints (PDF imports, I/O extraction), MVP2 shipped, V3.0 in development

• Active pilots with EU manufacturers in metal processing, industrial refrigeration, robotics and food production — two pilots completed this month, one agreement signed

The order intake agent is the first working instance of exactly this pattern: an agent that connects to an existing manufacturer system and automates a workflow. This project scales that pattern across the whole factory.

EXPERIENCE IN INTERNATIONAL R&D CONSORTIA

We are not new to collaborative R&D programmes:

• Eurostars 3, Call 1 — "Perfect Public Transport"

• Eurostars 3, Call 3 — "Audiobooks for Everyone"

• Eurostars 3, Call 4 — "Accessible Credit for All Businesses"

• Eureka Network Resilience 2024 — AIDER

• Collaboration with Vilnius University on international projects

We handle project writing and reporting for national and international agencies. We can lead this consortium or join an existing one that needs AI/ML expertise.

WHO WE ARE

AAI Labs — 50+ machine learning engineers (PhD and MSc specialists in AI, maths, CS, physics). We invest heavily in research. Clients across Lithuania, Sweden, South Korea, Israel and the United States.

CONSORTIUM WE SEEK (SMART Call 10)

• Industrial end-users: manufacturers with fragmented IT/OT landscapes who want factory-wide analysis without a multi-year integration project

• MES / automation / ERP providers: native integrations and go-to-market

• RTO: validation of the agent-generated integrations and process-analysis methods

• Machine/robot controller vendors: connectivity to equipment-level data

Target: 3-5M EUR, 36 months, 3-4 countries. Project Outline ready for SMART Call 10 (deadline Jan 26 2027)