The same patterns show up in almost every distribution business that hasn’t put AI on the ledger — and at least one of them is yours.
Reorders run on gut, not data
Buyers reorder on hunch and last week’s panic. Slow-movers pile up, fast-movers stock out, working capital sits on the wrong shelf.
Manual planning eats the morning
Planners spend two to three hours a day sequencing routes, building load orders and chasing standing-order changes — before any selling happens.
Exceptions surface after the fact
Short deliveries, returns and missed stops show up in next week’s report — long after the customer noticed and the supervisor could have fixed it.
Admin labour scales with volume
More routes, more SKUs, more customers — more clerks. Every new lane adds a person, because the planning, recon and commission work doesn’t automate itself.
How RouteMagic solves it
AI on the live ledger, not a side project.
Demand planning, replenishment and back-office automation built into the same database your drivers, planners and finance team already use — the forecast tells you what to buy, and the orders, routes and reconciliations follow on their own.
Neural Demand Forecasting
The N-BEATS model forecasts 30 days out at SKU, customer and route grain, with confidence bands. It gives you the sales forecast and the min/max on every line — the demand and sales-and-operations planning your buyers used to do in a spreadsheet — so you cut wastage and fewer fast movers stock out.
30 daysforward horizon
90+ dayshistory ingested
Per-SKUmin/max output
VRP Route Optimisation
GraphHopper’s VRP solver sequences routes in minutes, not by hand. Distance and time windows are respected, so planners approve the day rather than build it.
VRPGraphHopper engine
Time-windowaware
Minutesto plan
Auto-generated Orders & Documents
Order processing runs itself: the forecast triggers automated replenishment POs, and standing orders, load orders, service orders, credit notes and shelf-edge labels generate from upstream events. Nobody retypes what the system already knows, so back-office handles exceptions only.
Autostanding orders
Event-drivengeneration
0re-keying
End-of-Day Auto-reconciliation
Stock, cash and returns reconcile themselves from the day’s transactions. Variances arrive with a reason already attached, so supervisors just confirm on a screen.
Autostock recon
Autocash recon
Autoreturns recon
Commission Engine & Exception Alerts
The engine runs nightly against the live ledger and is returns-aware, so commission disputes drop from days to an hour. Exceptions push to the right supervisor with a one-line action attached.
Nightlycommission run
Returns-awareengine
Liveexception alerts
Scheduled Operations & Notifications
Pre-arrival SMS, gamification nudges and engagement workflows fire on the right trigger, without anyone clicking send. Customers stay informed and reps stay motivated, so ops stays quiet.
Pre-arrivalSMS / WhatsApp
Trigger-basedworkflows
Zeromanual sends
Outcomes
What changes once the AI runs the planning.
Representative results from live rollouts.
30 daysDemand forecast horizon2–3 hrsDaily planning eliminated~15 minTo reconcile the day~1 hrTo close monthly commission
Customer spotlight
How Barnies Foods cut wastage 30% with demand forecasting.
Barnies bakes fresh every morning and runs 10 vans across East Anglia — so every box that comes back is yesterday’s margin. RouteMagic now predicts per-store sell-through from each shop’s own history and matches production and delivery to demand, instead of baking on a hunch. Drivers invoice on the app rather than duplicate books, and the office sees every van and warehouse in real time.
The ones that come up most often. Book a walkthrough for anything else.
What is AI demand forecasting?
AI demand forecasting uses a machine-learning model to predict how much of each product you’ll sell, instead of relying on a buyer’s judgement and last week’s figures. RouteMagic trains a neural model on your own sales ledger and forecasts 30 days out at SKU, customer and route grain — then feeds that forecast straight into replenishment and order generation, so buying matches selling.
How is AI used for demand forecasting?
It learns the seasonality, trend and per-customer demand curve in your transaction history, then projects them forward with confidence bands. In RouteMagic the same forecast drives automated replenishment, min/max stock recommendations and the Forecast Sales screen, so the prediction becomes a purchase order a buyer signs off — not just a chart. It links tightly with inventory management for stock and replenishment.
How much does AI demand forecasting cost?
In RouteMagic it isn’t a separate add-on — the AI and automation layer is included from £29 per user per month, with the full platform at £39 (Core) and £59 (Premium). There’s no per-forecast or per-SKU charge and no standalone data-science project to fund; the model trains on the ledger you already run.
Will AI replace our demand planners?
No — it removes the spreadsheet work, not the planner. The model proposes the forecast and the replenishment quantities; a buyer reviews and signs off, and only the exceptions reach a human. Planners spend their time on judgement calls and supplier relationships instead of rebuilding the same numbers each morning.
Which forecasting model do you use?
A neural time-series model based on the N-BEATS architecture, trained per tenant on your own ledger. It predicts 30 days forward at SKU, customer and route grain with confidence bands, and re-fits continuously as new transactions land — no separate retraining project.
How much history does the forecaster need?
A meaningful fit starts at 90+ days of transactional history. Three to six months is the sweet spot for catching weekly and monthly seasonality. New SKUs and new customers cold-start from category and route priors until they have their own signal.
Does the VRP optimiser respect customer time windows?
Yes — time windows are first-class constraints in the GraphHopper VRP solver, alongside vehicle capacity, driver shifts, vehicle-customer compatibility and cold-chain requirements. Planners can pin stops, run what-if scenarios and override the proposal before publishing.
What exactly is “auto-recon” at EOD?
The system reconciles loaded vs. sold vs. returned stock, cash collected by denomination, cheque photos and digital references, and reason-coded returns — all from the day’s transactions. Supervisors confirm on a screen; variances surface with the most likely reason and a suggested action attached.
How does the commission engine handle returns?
It’s returns-aware by design. Commission runs nightly against the live ledger; returns and credit notes adjust commission in the same cycle, so reps aren’t paid out on a sale that came back the next day. Backdated returns and mid-month rule changes are both supported.
Which integrations does the AI layer feed?
AI-driven Purchase Orders, standing orders, credit notes and reconciled invoices sync natively to Xero, QuickBooks, Sage 50, Sage 200, Tally, Zoho Books and Microsoft Dynamics 365. Anything else — ERP, WMS, BI — via REST API. Exception alerts can also push to Slack, Teams, SMS or WhatsApp.
See it on your own ledger
Ready to forecast tomorrow and automate today?
Walk-through against a route shaped like yours
30 minutes
We’ll ask about your demand patterns, SKUs & current planners first
