Describe constraints in plain languageCopy link

A requirement like "the reefer van cannot do the bakery run before 6am" or "leave two low-value drops for tomorrow rather than pay a second driver four hours of overtime" has to end up in specific fields. The assistant maps the description onto those fields.

The translation, in one exampleCopy link

What you say:

Plan tomorrow's 62 deliveries. Two vans, both from the Mons depot,
drivers start at 7:00 and I do not want anyone finishing after 16:30.
The four pharmacy orders have to arrive before 11:00. Chilled orders
can only go on the van with the fridge. If it does not all fit, drop
the lowest-value stops rather than running late, and tell me which
ones you dropped and why.

What the assistant sets up for you:

Say the target, not the settingCopy link

The prompts that work best describe the outcome, not the field that produces it:

Cost, not distance.

I pay drivers 28 per hour and the vans cost 0.32 per km, plus 140 a day
each if I use them at all. Optimize for what this actually costs me, and
tell me whether the third van earns its keep.

Vehicle costs are solver inputs, see Vehicle costs and the COST metric. With them set, the answer to "is the third van worth it" is the COST difference between the runs.

Legal driving limits.

These are long-haul runs under EU rules: 45 minutes of break before
4.5 hours of driving, and it can be split 15 plus 30. Apply that to
both trucks and show me where the breaks land.

See Driver breaks and driving-time limits.

Capacity you keep hitting.

The van holds 18 roll cages. Let it come back to the depot to reload
mid-day if that serves more stops than a second van would.

See Multiple depots, open routes, and reloads.

Jobs that must stay together.

The scaffolding drop and the pickup an hour later are one job, same
crew, drop first. Same for the three site visits at Rue Verte, they
run in that order.

See Pickup & delivery transfers and Order sequences.

Ask for the trade-off, not just the planCopy link

The assistant can run the optimizer more than once and compare the results:

Run this three ways and compare cost, distance, and unassigned orders:
(a) as it stands, (b) with a third van, (c) with the finish time
relaxed to 18:00. Give me a short table and a recommendation.

Each run is a full solve on your data. It answers questions like "do I need another driver", "what does the 16:30 cutoff cost me", or "how much does the new customer's 2-hour window add".

What the assistant tells you about plans and limitsCopy link

Some constraints belong to paid plans: tags, pickup and delivery transfers, order sequences, force allow and deny lists, explicit reload depots, driver breaks, and the max distance, duration, and overtime caps. See Plans, billing, and free limits.

Requesting them through the assistant behaves exactly like the Optimize button:

• Your first 5 optimizations that use above-tier constraints run with those constraints applied.
• After that, the run still happens, but those constraints are ignored, and the assistant is required to tell you which ones.

Next stepCopy link

• Query, edit, and export a plan with AI
• The full constraint reference lives in Vehicles & Constraints.