Stockly

• Right-sized inventory, autonomously — AI agents continuously recalibrate buffers using stochastic simulation, so you carry enough cover without trapped cash. Indicative impact: 20–40% working-capital efficiency improvement and 30–60% faster decisions (results vary by operation).
• Supplier risk caught before it hits the line — per-supplier delivery-risk verdicts, so you find out about a slip before you're expediting.
• Plain-English decisions with cash impact — every recommendation comes with a reason and a dollar figure; planners approve every action (human-in-the-loop, full audit trail).
• On top of your existing ERP — connectors for SAP, Oracle, QAD, Epicor, Infor, NetSuite. No new system, no data re-entry.
• Proven model, agentic execution — built on the Ultriva pull-Kanban foundation: 25 years in production, 200+ plants, 20,000+ connected suppliers, 30 enterprise customers.
• Fast time to value — 2-week proof of value on as few as 20 SKUs; deploys in weeks, not quarters.

Problems addressed:

• Supplier slips go undetected until the shortage has already hit the line
• Inventory buffers set months ago quietly tie up cash or cause stockouts — nobody has time to recalibrate every SKU
• Planning takes days: one planner, thousands of SKUs, a spreadsheet that's stale before it lands
• No fast way to model disruption — tariff or supplier-outage decisions get made on instinct

Key features — four built-in agents:

• Stockout Agent — triages SKUs at critical stockout risk into a per-SKU action briefing
• Supplier Agent — RAG-based delivery-risk verdict for every supplier
• Signal Agent — narrates elevated external signals and their business impact
• Kai — multi-turn conversational advisor: ask anything about any SKU or scenario, with charts
• Digital Twin — natural-language "what-if" simulation of supplier, tariff, and demand shocks
• Two-tier architecture — agents orchestrate natively on OutSystems; demand, lead-time, and stochastic simulation run as containerized external ML microservices. Human-in-the-loop approval on every action; sized within default license capacity (150 AOs, 100 users).

Architecture
Stockly is a two-tier AI solution built on OutSystems:

• Agent Workbench (OutSystems-native): Four AI agents run natively on the platform — Stockout (per-SKU risk triage and action briefings), Supplier (RAG-based delivery-risk verdicts), Signal (external signal narration), and Kai (multi-turn conversational advisor with charts and contextual queries). Agents orchestrate and narrate insights; every recommendation passes a human-in-the-loop approval gate with an immutable audit trail.
• Intelligence Engine (external microservices): Demand modeling, lead-time analysis, flow optimization, and stochastic (Monte Carlo) simulation / Digital Twin run as containerized ML microservices. Hyperscaler-portable; AWS used for current deployments.

Integration

• Data Fabric ERP connectors: QAD, SAP, Oracle, Epicor, Infor, NetSuite
• Ingests ERP/MES data, supplier performance data, and external market signals
• Secure key management via Library; workflow and audit on platform-native capabilities

Sizing & deployment

• Sized within default license capacity: 150 AOs, 100 users
• All Critical and High priority Code Quality findings remediated
• Proof of value deployable in 2 weeks on as few as 20 SKUs; scales from one line or site to plant- or enterprise-wide

Modules
Core base (Control Tower, Supplier Performance, Settings) · Digital Twin (natural-language what-if simulation) · Advanced capability (expansion modules on roadmap)