Learning AI planning, delivery areas and closed zones
Learning AI planning is the OptiRoutes method that divides the day's stops between your drivers and puts each route in a good order, learning from your recent plans and the routes your drivers actually drive. Around it you can draw delivery areas for each weekday, and closed zones that are always planned as lines of their own.
Updated:
What is Learning AI planning?
You give the system the day's orders and your available drivers. Learning AI planning divides the stops between the drivers, puts each route in the best order it can find, and gives every route a driver and a vehicle. It builds the distribution lines as far as possible from the experience of your recent plannings, the way an experienced planner would: it keeps cities apart where it can and keeps the distances inside each line short.
The method adapts to the way you work. It learns from the routes your drivers actually drive, so the more you plan with it, the closer the routes come to your own way of distributing. On every route card you can open the considerations behind that route, so a planner can see why the system grouped stops the way it did instead of trusting a black box.
Learning AI planning is the method the getting-started wizard sets up for every new company, and the guides recommend it for most days. You can still switch to another method in the planning wizard whenever a day calls for it — see route planning for the full list.
Which settings steer the AI planner?
The learning method works inside limits you set. The getting-started wizard asks for the essentials, and the planning settings hold the rest. Every value is a default that the planning wizard shows before each run, so you can tighten or relax it for one day without changing the company setting.
| Setting | What it controls |
|---|---|
| Vehicle type | Truck, car, motorcycle or on foot. Travel times and allowed roads are calculated for it; trucks have forbidden roads and lower speeds. |
| Working hours | The departure time of routes and the latest time a route may end; no arrival is planned after the workday end. |
| Drivers going out | The most routes the day may be divided into. Leave it at 0 and the system works out how many the limits need. |
| Stops, hours and kilometres per route | A route never gets more than these limits. |
| Minutes at each stop | Unloading and signing time added at every customer, in arrival times and in how many stops fit a route. A customer with a stop time of their own uses theirs. |
| Back to the business at the end | Whether the way back to the depot counts in each route's time and distance. |
| Capacity limits | Weight, cartons, volume and pallets for a route without a vehicle; a route with a vehicle uses that vehicle's capacity. |
| Special handling of large orders | An order that fills almost a whole truck gets a truck of its own, and large orders are placed so trucks fill better. |
| Check time windows while dividing | Customers' time windows are checked against real driving times; a stop that cannot be reached in time moves to another route or stays unassigned. |
| Split into rounds automatically | When a plan creates more than 50 routes, every 50 routes go to a separate round. |
Each driver can also have a work area — where they usually work. In Learning AI planning the system prefers a driver whose work area is close to the route; a driver with no work area is matched by their recent routes.

How does the system learn from your routes?
Learning AI planning draws on two kinds of experience: your recent plannings, from which it builds the distribution lines the way a person would, and the routes your drivers actually drive. That is why its plans get better over time and come closer to your real operation the more you plan with it, instead of staying a straight-line ideal on a map.
In practice the best way to start is simple: keep the default values, plan one real day, look at the routes, and adjust the limits by what you see. Fix what the plan got wrong on the routes screen — a stop dragged to another route, a driver changed — and keep planning with the same method. Lock the routes that must not change so a later run never touches them.
For companies on the current OptiRoutes server, which includes every company that signs up by itself, Learning AI planning calculates the routes in compact territories: the day's stops are divided into areas sized to the vehicles' capacity, each vehicle may work its own territory and the next one, and the result is then improved by moving stops to the nearest neighbouring routes whenever that saves distance.
Delivery areas for each weekday
On the Areas screen you draw your delivery areas on the map. Each area is a shape with a name and, if you like, a driver — and with the driver comes the driver's line number. Areas live in templates, one per weekday: template 1 is Sunday, 2 is Monday and so on to 7 for Saturday. When a day has no template, or its template has no areas, template 1 is used.
- Draw with a polygon, a line of at least three points or a circle; then attach the shape to the area.
- Copy a whole template to another day, then change only what differs.
- Check coverage by loading points onto the map: the deliveries between two dates, every customer with a map location, or your own CSV or Excel file with latitude and longitude columns.
- Export an Excel file of the template's areas, or of every loaded point with the area it falls in.
When planning is set to split by areas, stops are first divided by the day's areas and each area's route gets its driver and line. Loading a whole week of deliveries before changing a border shows where the work really is, not just today's orders.
What is a closed zone?
A closed zone is an area whose orders must always travel together and never mix with orders outside it. You create closed zones on the same Areas screen: give a template a name that starts with @, and every shape in that template becomes a closed zone. In each planning run, every order whose point lies inside a zone is planned only with the other orders of the same zone, as one or more lines of its own.
- Planned first — closed zones are carved out and planned before any planning method runs; the method you chose gets the rest of the orders.
- A fixed driver — when the zone has a driver, only that driver drives it; otherwise its line number decides, or any free driver takes it.
- Drivers stay busy — a driver who got a zone route is treated as busy by the rest of the plan.
- Nothing lost — orders a zone's driver cannot fit stay on the unassigned stops with the reason.
Typical uses are a remote town that a dedicated vehicle serves once a day, a large customer site with its own driver, or a district that should always be one line even when the algorithm would split it differently. Closed zones are on by default.
Learning AI, areas or closed zones: which to use?
| Need | Use |
|---|---|
| Routes that follow how your team already works, with limits | Learning AI planning |
| A fixed area per driver, the same every week | Delivery areas and planning by areas |
| Different areas on different weekdays | One area template per weekday |
| A place that must always be its own line, whatever method runs | A closed zone in a template named with @ |
| The lines already decided by your ordering system | Planning by the lines in the orders file |
The approaches combine: closed zones are planned first, and the rest of the day goes to Learning AI planning or to any other method you pick. After planning, the routes screen is where you review and adjust the result before it goes to the drivers.
Try Learning AI planning on your own orders — the free trial needs no card.
3-day free trial · up to 50 addresses a day · no credit card
Frequently asked questions
What does “learning” mean in Learning AI planning?
The method builds lines from the experience of your recent plannings and from the routes your drivers actually drive, so the more you plan with it, the closer its routes come to the way you work.
Can I see why the AI built a route a certain way?
Yes. Every route card created by Learning AI planning lets you open the considerations behind that route.
Do I have to draw delivery areas to use OptiRoutes?
No. Learning AI planning works without areas. Areas are useful when you want a fixed division, such as a regular driver per area.
How do I make a closed zone?
On the Areas screen, create a template whose name starts with @ and draw the zone in it. Orders inside the zone are then always planned as their own line, before the rest of the plan.
Can I still change the routes after AI planning?
Yes. Drag stops, change drivers, merge or split routes on the routes screen, and lock routes so later planning runs do not change them.
Start free trial — 3 days, no card
3-day free trial · up to 50 addresses a day · no credit card