Burq explains how delivery AI must earn control for each shipping mode

Burq says its delivery AI must prove itself separately on each shipping mode before taking control of orders. In a November 3 FreightWaves interview, cofounder Salman Habib described a system that earns authority through live deliveries and retailer-set rules. He said the approach is already built into a product customers use, rather than remaining a proposal.

The framework comes from Earned Autonomy: A Deployment Doctrine for AI in the Physical World, which Habib and cofounder Shaban Habib published on October 13, 2026. It is the first installment in Burq’s Intelligence Notes series. Its central rule: AI starts without decision-making authority, and a person must approve every promotion.

Burq’s software chooses how an order gets delivered, selecting among a retailer’s own fleet, gig providers, parcel carriers, less-than-truckload freight carriers, and drones. Habib’s argument is grounded in the consequences: sending an order to the wrong carrier or driver leaves an angry customer and reflects badly on both Burq and the retailer.

He said the doctrine grew out of sales conversations. Buyers who previously asked whether a vendor had AI now ask what it decides and what it returns to a person. Burq contrasts its approach with what it calls granted autonomy: giving a system full authority at integration, based on the vendor’s own benchmark.

Parcel experience does not buy freight authority

Under the doctrine, delivery modes are separate segments. Success handling parcel orders does not automatically qualify the AI to make freight or gig-delivery decisions, because their operations differ. A retailer could therefore allow autonomous gig dispatch while still requiring a person to approve every less-than-truckload decision.

Each segment moves through four levels. The distinction is not simply whether a person is present, but whether the AI is observing, recommending, or acting:

1. Manual: The AI reads dispatchers’ work without making delivery decisions of its own.
2. Shadow: It records the decision it would have made alongside the dispatcher’s actual choice.
3. Review: It proposes a decision, which a person can accept or reject before it takes effect.
4. Autonomous: It acts on its own, but only within the rules governing that segment.

Retailers set the promotion criteria. Habib offered dispatcher agreement with the AI on 90% of decisions as one possible bar for autonomy. That is an example, not a reported customer requirement or a measured success rate. Under Burq’s rules, a human must still sign off before the segment advances.

The reverse move works differently. Burq says a segment that breaches a hard rule immediately loses authority, with evidence attached to its demotion. Habib also rejects permanent human rubber-stamping as the goal: once dispatchers are confident in a segment, he says, they should be able to step back.

We don’t need that. We just need human to be a stage in the process.

Salman Habib, Burq cofounder, speaking to FreightWaves about permanent human approval

A trial without changing the dispatch desk

For a shipper considering a trial, Habib recommends Shadow mode. Existing operations continue while the AI reads decisions across order, transportation, and warehouse management systems. Burq can then produce a report showing where its choices would have differed over the first 30, 60, or 90 days, without putting those hypothetical choices into action.

Habib said Burq began even earlier with one large U.S. retailer, running three to six months of historical orders through its AI. The exercise compared choices across the retailer’s fleet, parcel, gig, and freight options. He declined to identify the retailer, citing confidentiality. Such a backtest earns no authority under the doctrine; that requires live orders.

Burq records decisions in its Decision Ledger, with one row per order identifying who decided, what information was available, which option was chosen, and the outcome. Customers can export the ledger as a CSV file. That gives the handover process an order-level record, rather than leaving the retailer with only an overall recommendation.

The financial or operational payoff remains unquantified. MarketScale’s November 6 account noted that no improvement figures were disclosed for the retailer backtest or Shadow-mode reports. Burq has described how a retailer can decide to surrender control, but has not supplied those results to show how much better its delivery decisions become.