The Drive-Thru Bot Does Not Widen the Lane

Bojangles is putting AI order-taking in more of its drive-thrus. Nation's Restaurant News reported on October 22 that the chain extended its agreement with Hi Auto to cover its 890-plus-unit system. Restaurant Dive, covering the same announcement, put the near-term plan at about 100 more drive-thrus by mid-2027, on top of roughly 500 already running, which it said brings coverage to about two-thirds of the chain.

Our view: this is a labor and accuracy story, not a site story. Software at the speaker post does not add a foot of stacking. What a drive-thru pad can do is still set by how many cars it holds off the public road and how cleanly its lanes come back together before the window. The two features Bojangles is adding say as much.

What the new features say

The first is upselling keyed to the line. NRN reports the system now adjusts its recommendations in real time based on drive-thru queue length, offering quicker-to-prepare items when the line is long, with the goal of adding revenue at peak without slowing service or adding labor. That is a sensible tool. It is also built for a line that already exists. It manages the queue. It does not shorten it.

The second is a picture of the car. Both outlets report that each order is now paired with an image of the vehicle that placed it, so staff can match orders when dual drive-thru lanes merge. A chain does not build software for the merge unless the merge is where orders go wrong. On a site plan, the merge is a line on a drawing. In operation, it is where a dual-lane drive-thru gives back some of the capacity the second lane was meant to add.

What we underwrite

When we look at a drive-thru pad, we count the stack from the pickup window back to the point where a waiting car would block a drive aisle or the curb cut. We look at where two lanes come together and how much room is left after the merge. We look at whether the line, at its worst hour, stays on the property. None of that changes when the order-taker is software.

We would not credit a site with extra capacity because the tenant's ordering is automated. If a store takes orders faster than the kitchen and the window can hand food out, the extra speed shows up as a longer line. The length of that line is a real estate question.

The adoption question

There is also reason not to treat voice AI as settled. NRN notes that Burger King is slowing its own drive-thru voice test after finding some customers don't like it, and that Taco Bell slowed its rollout last year after an ordering glitch went viral. It cites Technomic data showing 23% of consumers find ordering from an AI drive-thru bot appealing and 46% find it unappealing.

Bojangles and Hi Auto say the system handles more than 1 million orders a week, with what NRN reports as a 98% success rate. If that holds across more stores, it is a real operating gain for the brand and its franchisees. What we will watch for is whether brands start drawing shorter stacks on new prototypes on the strength of faster ordering. We would want to see window times, not ordering times, before we accepted that.