Neuclon Intelligence
Neuclon is the operations intelligence and automation layer above your ERP. It learns from your own orders, receipts, machines, deliveries and conversations; predicts what will go wrong before it does; prescribes the best thing to do; and acts once a person approves.
How it works
Learned from your own history: orders, receipts, machine readings, deliveries, deals and conversations. Each model is tested on your past data first and only predicts when it beats a simple base rate; otherwise it uses clear rules and says so.
Every prediction comes with the reasons behind it and the best thing to do, chosen from the real options: a transfer or a purchase order, this carrier or that one, repair now or after the run.
One click does it: the purchase order, the maintenance window, the customer update, the task for the owner. People approve what matters, within the roles you set, and every action is recorded.
What a prediction looks like
Build the load and book a carrier today. If it can’t be saved, tell Lakeshore Foods today, not on the day.
Expedite 74. A transfer from another site was checked first; nothing else arrives in time.
Plan the repair for November 11, before it crosses on November 13. A planned repair takes hours; a breakdown takes days.
Examples from Neuclon’s demo workspace, with sample data. Your cards come from your own records.
Eleven models, four apps
Each model lives inside the app its users already work in, so a prediction arrives with the record it is about.
| App | Predicts | Prescribes | Acts |
|---|---|---|---|
| Customer 360 | Who is ready to buy and who is drifting away; each deal’s chance to win; the discount that wins as often as a bigger one | Call this week; protect the price; offer 4% instead of 8% | A task for the owner; the discount set |
| Demand & Supply Planning | Real lead times from receipts; demand from the open pipeline; shortages across sites; whether a promise date holds | Transfer, buy or expedite, the cheapest option that arrives in time | A transfer or purchase order for approval; the lead time updated |
| Manufacturing Execution | When a machine reading will reach its limit, with the orders it puts at risk; measurements drifting while every part still passes | Plan the repair before the breakdown; check the tool before the first bad part | Maintenance planned; the operator alerted |
| Logistics, Warehousing & Fleet | On-time rate of every carrier on every lane; each order’s chance of shipping late or short, and why | Book the carrier that performs on this lane; release, replenish or warn the customer | The carrier rebooked; the customer told on their own channel |
Why you can trust it
Bring your own model
Point any probability model at your own model on Databricks, Azure Machine Learning, Google Vertex AI, Amazon SageMaker or any HTTPS endpoint. Neuclon sends each prediction’s inputs as rows and uses the probability your model returns, while the reasons, prescriptions, actions and approvals stay the same.
POST /serving-endpoints/late-delivery/invocations
{"dataframe_records": [{
"neuclon_model": "delivery",
"entity": "S00042",
"carrier": "Prairie Express",
"lane": "Plant → Chicago",
"lines": 2,
"neuclon_probability": 0.81
}]}
→ {"predictions": [0.86]}What it learns from
New workspaces can load a year of realistic sample history, so every model learns and every page has something to show on day one.
Questions
No. The models are built in and learn from the records Neuclon already holds for you. Your people read the reasons, choose an option and approve it.
Less than you might think, and each model says exactly what it is waiting for: five readings per machine for a failure forecast, three receipts per supplier and item for a real lead time, 40 closed deals for win probability, 200 shipped orders for delivery risk. Until then it uses clear rules.
No. Your ERP keeps the record. Neuclon reads it, joins it to email, Teams, WhatsApp, machines and carriers, and writes the actions back where they belong, such as a purchase order for approval in Odoo.
Only within the roles and approval rules you set. By default, actions that change records need a person with the right role, and every action is recorded with who approved it.
Yes. Ready to buy, win probability and delivery risk can be scored by your own model on Databricks, Azure Machine Learning, Vertex AI, SageMaker or any HTTPS endpoint, side by side with Neuclon’s or in charge, and both are graded against the same outcomes.
Neuclon Intelligence is part of the Business and Enterprise plans, and of the free 30-day trial.
Start free with sample data, or let us show you on your own records.