FleetPlanning

Knowledge base

Which planning software can calculate and explain? An overview

By Maarten Alta ยท 30 October 2026

Almost every supplier of planning software now says "AI". Very different things sit behind that one word. For the question that matters, "can the system explain how it reaches a plan?", I see four kinds of solutions.

This overview is not a ranking and not complete. I mention examples I came across; the market is changing fast.

1. Solvers with built-in explanations

The solver itself keeps track of why the plan is the way it is: which rules are binding, how much each rule costs, and why an order was not planned. The explanation is in tables and scores, not in plain language.

Example: Timefold. An open-source planning engine that software developers use to build plans for trips and rosters, among other things. The score of a plan is broken down per rule, and two plans can be compared. This is mainly the foundation other software builds on.

Suitable if: you are having software built, or you want to ask your supplier where their explanations come from.

2. An optimisation model with a language model on top

Here a language model reads the outcome of the optimisation model and explains it. For a what-if question, the language model adjusts the calculation, has the optimisation model recalculate, and explains the difference.

Example: Microsoft OptiGuide. A research project from 2023 that does exactly this, and that Microsoft uses in planning for its own data centres. An important detail: company data does not need to go to the language model. It is not a package for hauliers, but it shows that the approach works in practice.

3. Route planners you connect your own AI assistant to

More and more route planners offer a standard connection (MCP) that lets an assistant such as Claude or ChatGPT work directly with the planner. The assistant turns your wishes into rules, the planner calculates, the assistant explains.

Example: Routing24. A route planner for delivery and service. For each order that is not planned, the assistant gets the reason, what would have to give way to plan it after all, and what that costs.

Watch out: many of these connections only let the assistant enter orders and start the calculation. The assistant then gets no explanation of the why. Ask about it.

4. A TMS with a built-in AI assistant

The large transport management systems are building their own assistants that answer questions and sometimes also carry out actions.

Examples:

  • Soloplan CarLo showed an AI assistant for planners at IAA Transportation 2026 that answers questions and can also carry out actions itself, plus a standard connection for external AI applications.
  • Transporeon (part of Trimble) is expanding its TMS with AI-driven planning.
  • Locus works with AI agents that handle orders, trips and carriers themselves, and with a conversational layer in which the planner asks what-if questions.

How far the explanation goes differs a lot. An assistant that gives answers is not yet an assistant that explains why.

How to test it in a demo

  1. Have an unplanned order explained. Then find the reason back on the screen.
  2. Ask a what-if question. Ask whether it was recalculated, or whether the language model gave an estimate.
  3. Ask what calculates and what talks. Costs, battery and driving times should come from the optimisation model.
  4. Test with electric trucks. Can the explanation cover battery level, charging moments and the grid connection?
  5. Ask where your data goes. Which language model, from which supplier, and what happens to your data?

More demo questions are in Choosing planning software: 10 questions for your supplier. If you are choosing a new TMS, also read Requirements for planning software with an electric fleet. And the background: Calculating and explaining.

Sources

What do these changes mean for your company and your planners?

A first conversation is free and without obligation. You will hear from me within two working days.

Get in touch