Burq’s AI must earn delivery authority one mode at a time

Burq’s delivery AI must progress through four authority stages before it can act alone on each delivery mode. Co-founders Salman Habib and Shaban Habib set out that deployment doctrine on Oct. 13. The stakes are immediate: a misrouted order reaches a customer and damages both retailer and vendor.

At the dispatch desk, Burq’s software chooses among the retailer’s own fleet, gig providers, parcel and LTL carriers, and drones. FreightWaves described the system in August as the “last-mile decision layer,” where an algorithmic mistake becomes a service failure.

Wrong dispatches can send an order to the wrong carrier or driver, producing an angry customer. Habib told FreightWaves: “Us getting an order wrong … is an angry customer, and that looks bad on the retailer, that looks bad on us as a company.” Unlike a wrong sentence from Claude or ChatGPT, the service failure reaches the customer immediately.

Vendors’ push for full trust from day one has left many retailers still deciding whether to use delivery AI at all. That hesitation forms the commercial background to Burq’s alternative deployment model.

What are retailers asking AI vendors now?

The doctrine, titled Earned Autonomy: A Deployment Doctrine for AI in the Physical World, appeared Oct. 13 as the first Intelligence Notes entry. It requires each business process to begin without AI authority; permission grows only from live-order evidence, with a person approving every step.

Habib said the doctrine grew from his own sales meetings. Until the previous year, buyers asked whether a vendor had AI. Their question then shifted to what the AI decided and when it returned a decision to a person.

Habib calls the opposing model granted autonomy: a vendor requests the whole operation on integration day, then evaluates it against its own benchmark after a polished demonstration. When the system errs, he said, retailers have no transparent way to reconstruct what happened, what went wrong, or why it chose wrongly.

How does Burq limit the blast radius?

Burq divides an operation by mode, so evidence earned on parcel does not transfer to LTL, gig, fleet or drone work. Each segment advances through four rungs on the company’s authority ladder:

• Manual: The AI reads dispatchers’ work without making decisions.
• Shadow: It logs the call it would have made beside the dispatcher’s decision.
• Review: It proposes a decision, and a person accepts it.
• Autonomous: It acts on its own within rules set for that segment.

Retailers set the bar for every promotion. In Habib’s example, dispatchers must agree with 90% of the AI’s calls before the retailer grants full autonomy. Once they trust that segment’s performance, they can step back and let it run.

Burq rejects the industry term human in the loop because, Habib said, it suggests continuous rubber-stamping. “We don’t need that. We just need human to be a stage in the process,” he said. Human review thus occupies defined stages rather than every decision.

According to Burq’s site, breaching a hard rule demotes the segment immediately, with evidence attached. The mechanism contains a weak segment without changing authority elsewhere.

What can a shipper audit before deployment?

Every decision enters Burq’s Decision Ledger, which records one row per order. Each row shows who decided, what the system knew, what it selected, and how the outcome unfolded. Customers can export the ledger as a CSV file.

Habib said he had not seen another company, in logistics or any other sector, show customers this level of visibility into AI decisions. For a shipper weighing a trial, Shadow mode offers a low-disruption starting point.

Operations remain unchanged while the AI silently reads dispatch decisions and the retailer’s OMS, TMS and WMS. After 30, 60 or 90 days, Burq reports what it would have done differently.

Burq began earlier with one of the largest U.S. retailers, replaying three to six months of past orders to compare how the AI would choose among the in-house fleet, parcel, gig and LTL options. The retailer remains unnamed for confidentiality. Historical replay can show what the system would do, but only live orders let the AI earn authority.

What changes for shippers, carriers and forwarders?

For shippers and retailers, the staged approach allows a trial without changing operations, then isolates each mode before granting control. The ledger and CSV export give procurement and operations teams evidence for promotion. Automatic demotion provides a defined response to a rule breach.

For parcel, LTL, gig, fleet and drone providers, the stakes sit directly at order assignment: Burq’s software chooses which operator receives the job. A wrong choice can damage customer experience and the reputations of both retailer and vendor. No participating carrier, shipment volume, rate or capacity figure was disclosed.

Freight forwarders have no stated rate, capacity or upstream-routing change in the doctrine. Its described control point is last-mile delivery selection, not ocean, air or linehaul planning.

Customers will be able to choose between two data arrangements. Retailers can keep their data exclusive to their own deliveries, or opt into anonymized performance data from the wider network, creating a trade-off between privacy and cross-network insight.

What happens next?

The doctrine already sits inside a product customers use, according to Habib. Burq now treats any movement of an order—gig, parcel, LTL or drone—as part of Burq Intelligence. That scope extends beyond traditional last-mile delivery and creates a broader base for testing segment-specific autonomy.

Trust must come before the technical and data questions advance, Habib said. The policy trajectory points toward broader autonomous decisions, but each mode must earn that status through live-order evidence and remain subject to demotion after a hard-rule breach.