Charisma ERP + AI Demand Forecasting — How to optimize stock and procurement (2026)
Charisma BI is reporting-based, not predictive. Learn how to layer AI over Charisma for SKU/location demand forecasting, inventory optimization, and supply planning.
Mihai Istrati — CTO Azuvio · 2026-05-30 · 9 min · ERP & Integrations
Why AI Forecasting matters
Inventory is trapped cash. Under-stocking leads to lost sales. Over-stocking leads to tied-up capital and expiration risks. The difference between a company with an 8x annual stock turnover and one with 4x is 50% less capital tied up for the same revenue.
Charisma BI reports on what happened (sales, stocks, turnover). However, it does not predict what will be sold next month per SKU per location. That requires AI / ML.
What the AI Forecasting (Smart Layer) does
Data inputs from Charisma:
24-36 months of sales history
Current stock levels per warehouse
Supplier lead time per SKU
Historical promotions (date, discount %, sell-out)
Calendar (holidays, seasonality)
AI Output:
Predicted demand per SKU per location for 4-12 weeks
Optimal order quantity (ROP — Reorder Point + EOQ)
Alerts for SKUs at risk of OOS (out-of-stock)
Alerts for over-stocked SKUs (slow-movers, nearing expiry)
Promotion mix recommendations (which SKUs to promote for maximum revenue)
Typical Accuracy
For SKUs with stable history (12+ months of sales): MAPE 8-15% (industry standard).
For new SKUs (under 6 months): MAPE 20-35% — still superior to manual gut-feeling.
ROI
Distributor with 5,000 active SKUs, average inventory of €3m:
Inventory reduction: 18-25% → freeing up ~€550-750k in cash
OOS reduction: -60% → recovered sales of ~€250-400k/year
Expired stock reduction: -70% → savings of €80-150k/year
ROI of the AI Forecasting Smart Layer: 2-3 months
How it integrates with Charisma
1. Nightly ETL — data extraction from Charisma (sales, stock, purchase orders) to the Smart Layer data lake
2. AI processing — ML models run daily, recalculating the forecast
3. Push back to Charisma — replenishment recommendations appear as "order proposals" in Charisma; the operator validates, modifies, or approves
4. Executive Dashboard — live KPIs (fill-rate, OOS rate, days of stock per SKU)
Typical Setup
Weeks 1-2: Charisma data extraction and cleansing
Weeks 3-6: ML model training on historical data
Weeks 7-8: Team validation and calibration
Weeks 9-10: Progressive go-live (10% SKU → 50% → 100%)
Cost: €10-25k/year (included in the Smart Layer package).
Conclusion
AI Forecasting does not replace the planner — it assists them. The final decision remains human, but the AI input eliminates 80% of repetitive calculation work for 80% of the SKUs.
See «Charisma Limitations».