Route Planning with AI: When the Dispatcher Stops Building Every Route by Hand

Why route planning eats so much experience and even more time

In many forwarding companies, the route-planning knowledge sits in one person's head: the experienced dispatcher who knows which road jams up in the morning, which customer can only unload before noon, and which driver prefers which route. That works — until this dispatcher is sick, on holiday, or retired. Then it becomes clear how much invisible manual work sits in daily route planning:

• Orders get sorted by hand according to postcode, time window and vehicle size — often in a spreadsheet or on paper
• Short-notice extra orders throw off the whole plan, because nobody recalculates automatically
• Customer time windows (unload only 8-10 a.m.) get overlooked, and routes arrive empty or too late
• Empty miles only surface at month-end when the fuel bill comes in
• With a vehicle breakdown or traffic jam, re-planning starts completely from scratch again

The core problem: route planning is an optimisation problem with many variables — capacity, time windows, distance, driver working hours, special agreements. People solve this intuitively well, but slowly, and with limits on how many variables they can juggle at once.

How an AI agent supports route planning

An AI agent doesn't replace the dispatcher — it delivers a fully calculated proposal as a starting point, and takes over the recalculation whenever something changes.

Step 1: Gather orders and constraints

The agent pulls open orders from the TMS or order list, plus vehicle capacity, driver availability and known customer time windows. If details are missing — say, a new time window at a customer — it asks specifically instead of calculating with wrong assumptions.

Step 2: Propose routes

On this basis, the agent creates an initial route proposal: which orders go together on which vehicle, in what sequence, factoring in driving and rest times. The proposal is a draft — not an automatic assignment without oversight.

Step 3: The dispatcher reviews and adjusts

The dispatcher sees the proposal, immediately spots anomalies, and adjusts based on what their experience tells them — for instance, that a particular customer always causes loading delays. Only after approval does the route go to the driver.

Step 4: Automatically recalculate on changes

When a rush order comes in, a vehicle breaks down, or telematics reports a traffic jam, the agent recalculates the affected part of the route plan and proposes an adjusted sequence — in minutes instead of half an hour of mental arithmetic.

Step 5: Make empty miles and utilisation visible

Along the way, the agent analyses where empty runs regularly occur or capacity goes unused, and delivers these patterns as a basis for better route planning going forward — not as a one-off report, but continuously.

Which systems get connected

The agent works with what's already in your forwarding company: TMS, telematics, spreadsheet-based route plans, customer time windows from email agreements. Where there's no modern interface, access gets set up through exports, files, or operating the existing user interface — a system change is not necessary, that's our core promise.

Data protection

Driver and vehicle data stays within your defined framework: operation on German servers or entirely within your own environment, with a data processing agreement and complete logging of every planning change.

What you can realistically expect

A typical result after implementation: the daily base plan takes noticeably less time, because the dispatcher reviews a fully calculated proposal instead of starting from zero. Re-planning after breakdowns or rush orders goes much faster, because the agent takes over the recalculation. Over time, the share of empty miles also drops, because patterns become visible that nobody systematically analysed before.

Important for expectations: the agent makes no autonomous decisions about vehicles and drivers. The final call stays with the dispatcher — especially for special cases, customer relationships, and short-notice human arrangements that no system can fully capture.

An everyday example

Say a mid-sized local delivery company gets three more rush orders around 2 p.m. on Thursday for the next day. In the past, the dispatcher would have re-sorted the entire Friday plan by hand — a good hour of work, often after clocking off. With a route-planning agent, they instead review an automatically recalculated proposal, shift two routes with a click, and approve. Drivers receive the updated plan that same afternoon.

Common objections from the field

"Our dispatcher knows the routes better than any algorithm." That's exactly why they stay in the decision. The agent delivers a proposal as a starting point — the experience feeds into review and adjustment, not into manual first drafts.

"Our routes are too individual for automation." The agent learns your customers' and routes' special rules over time and flags where details are missing instead of guessing. You start with the routes where the data situation is clearest.

"What if the software plans wrong?" Every proposal is a draft for review, not an automatic assignment. The dispatcher sees the reasoning behind the proposal and can intervene at any time.

"We already have planning software and it doesn't work properly." Often the problem isn't the software itself, but that time windows, special rules and current capacity aren't kept up to date consistently. An AI agent can take over exactly that upkeep — it reads changes from emails and calls and keeps the data basis current, the foundation any planning needs to be good in the first place.

Self-check: is this worth it for your forwarding company?

• Route planning hangs on one or two people with a lot of experience
• Rush orders or breakdowns regularly upend the day's plan
• Empty miles only surface at month-end with the fuel bill
• Customer time windows occasionally get overlooked or mixed up
• New dispatchers need months to understand the route logic

If three or more of these apply, it's worth taking a close look at your route planning.

The next step

We'll work out whether meaningful support for your route planning makes sense in a free intro call: we'll look at your current planning logic, what data is available, and where the biggest time savings realistically are. A pilot follows within a few weeks. For more use cases, see our industry page AI in logistics.