ERP · AI · Supply Chain · Quantum Computing — insights from building a real enterprise learning lab
Two new local-AI tools on the portal — a structured error decoder (what it means, causes, fixes, prevention) and a contract analyzer that flags red flags and missing clauses like an absent liability c
Train real scikit-learn models in the browser — pick an algorithm and dataset, tune hyperparameters, and read genuine metrics, confusion matrices, named feature
The portal's agentic tool decomposes a goal into steps, routes each to a tool, executes in parallel, and synthesizes a summary — with an honest look at what a small local model gets right and wrong.
Following the work orders the S&OP plan created: the WO list, one order opened with its BOM materials and routing auto-loaded, and the release-to-complete lifecycle — with live screenshots.
Reading BOM-001 for the SmartBar Pro 500: the 4-component structure, Level 2 raw materials, on-hand vs available, and the live cost explosion — with live screenshots.
RTG-001 in Flow View and Full Detail: setup vs run time, compounding yields, departments and cost rates, and the exact match with WO-2025-0001's planned hours — with live screenshots.
Four work centres with live utilization — and a 119% overload on the PCB Soldering Line, the open work orders feeding the load, and the 130-unit S&OP order waiting to make it worse.
The three registries every transaction references: the Item Master with make/buy and margins, 55 customers with segments, supplier cards with terms and lead times — with live screenshots.
SHP-2025-0001 closes the loop from Tutorial 1; the supplier performance table with expedite economics and the 80% reliability line; and the Reports dashboard where inbound lateness shows up in the on-
The series finale: 505 decisions triaged, and SO-2025-0226 scored 67/100 — with every weak dimension traceable to a page from earlier in the series: inventory, suppliers, and a 119% work centre.