AI for ERP
Forecasting, anomaly detection and matching applied to the data your ERP already holds, measured against what actually happened rather than presented as magic.
Contact Us
Where the value actually is
Three places, consistently. Demand forecasting, because ordering decisions made on last month's sales and instinct tie up working capital in the wrong items while the right ones go out of stock. Anomaly detection, because a duplicate payment or a price that was keyed with an extra zero is cheap to catch automatically and expensive to find later. And matching, because reconciling payments to invoices where references do not quite align is work a model is genuinely better at than a rule. Each of these has a measurable before and after. That matters: AI in an ERP should be judged on stock turns and error rates, not on whether the demonstration impressed anyone.
Measured against what happened
We backtest. Train on data up to a point in the past, predict forward, compare against what actually occurred. That produces an honest accuracy figure before anything reaches a planner's screen, and it establishes whether a model genuinely beats the simple heuristic you already use. Sometimes it does not, and we will tell you. A seasonal naive forecast is a surprisingly strong baseline for stable products, and beating it meaningfully is the bar a model has to clear to be worth its complexity. For products with real variability the improvement is usually substantial, which is why segmenting the catalogue matters more than picking an algorithm.
A recommendation, not an instruction
Forecasts inform a person who knows things the model does not: a customer mentioned a tender, a competitor closed, a road is shut. The planner should see the suggestion with its reasoning and be able to override it without a fight. We build for that. Suggested order quantities with the drivers shown, anomalies flagged for review rather than blocked automatically, matches proposed rather than posted. The measure of success is a planner who trusts the suggestions enough to accept most of them, not a system that removed the planner.
What you get
Demand forecasting
Seasonality, trend and lead time, so stock sits in the items that move rather than the ones that did.
Backtested before it ships
Trained on the past and compared against what actually happened, so accuracy is known, not claimed.
Compared against the simple option
A seasonal naive forecast is a strong baseline. A model has to beat it meaningfully to earn its complexity.
Anomaly detection
Duplicate payments, prices keyed with an extra zero, quantities that do not fit the pattern.
Smart matching
Payments to invoices where references do not quite align. Genuinely better handled by a model than a rule.
Suggests, with reasoning shown
The planner sees why and can override without a fight. They know things the model does not.
Technologies
Frequently asked questions
How much history do we need?
For seasonal forecasting, ideally two to three years so the model sees each season more than once. Less can work for trend and anomaly detection. If your history includes a genuinely abnormal period, that needs handling deliberately rather than being averaged in.
How accurate will the forecast be?
It depends entirely on how predictable your demand is, and anyone quoting a figure before seeing your data is guessing. The backtest gives you a real number for your products before you commit. For genuinely erratic items, the honest answer is sometimes that no model will help much.
Does this replace our planner?
No, and a system designed to would be worse. The planner knows about the tender, the closed competitor and the shut road. The model handles the volume and the arithmetic so their judgement goes where it is actually needed.
Put the data your ERP already holds to work
Contact UsGet Free Estimation
Have a question or want to discuss a project? We'd love to hear from you. Get in touch and we'll respond as soon as possible.
Contact Information
Fill out the form and our team will get back to you within 24 hours.
Location
Serbia