← Back to use cases
A retailer tests AI delivery choices against past orders
Added 2026-11-03 · Reviewed 2026-11-03 · Unnamed US retailer (Burq customer)
Read the source ↓What the company did
Burq cofounder Salman Habib told FreightWaves that the company evaluated AI delivery selection for an unnamed large US retailer by replaying past orders. It produced a report showing how the AI would have chosen among the retailer’s in-house fleet, parcel, gig and LTL options.
How it went
The November 3 interview says the replay covered three to six months of order history. That is the data window, not the trial duration. No order count, agreement rate, cost saving or delivery-performance result is supplied, and live autonomous dispatch is not confirmed for this retailer. The account establishes a reported evaluation, not a proven operational improvement.
Worth knowing
- The retailer is unnamed, and the model, input fields and integration method are undisclosed. Burq’s wider product capabilities do not establish which controls this particular evaluation used.
- Practical takeaway: Check whether each alternative provider, rate and delivery window was available at the original decision time, including capacity and service restrictions. Keep hypothetical choices separate from actual bookings, and require evidence from live orders before granting dispatch authority.
Sources & provenance
Delivery AI should earn its authority one segment at a time ↗
FreightWaves · Published 2026-11-03 · Reviewed 2026-11-03
“running three to six months of its past orders through the AI”
Supporting evidence and review
Human review required
- Check whether each suggested provider, rate and delivery window was actually available when the original dispatch decision was made.
- Review service constraints and record missing information before comparing the AI choice with the actual booking.
Evidence & limitations
- Original trade interview reports the provider’s account of an unnamed customer evaluation; AI & Freight did not perform or independently verify the trial.
- Three to six months describes the history replayed, not the trial duration or number of orders. The evaluation date is not supplied.
- No agreement rate, cost saving, delivery improvement or confirmed live autonomous dispatch is reported for this retailer.
- Exact model, fields, candidate-option history, constraints and integration method are undisclosed.
- General product modes and illustrative 90% agreement thresholds in the article are not achieved results for this customer. Historical replay cannot by itself establish real-time dispatch readiness.