Arc is a voice AI startup that takes drive-thru orders and optimizes business performance for fast-food restaurants. The company raised $10.76 million seed funding led by Andreessen Horowitz and is already working with two major fast-food chains with hundreds of locations. Drive-thrus represent a large share of quick-service restaurant revenue, with about 200,000 drive-thru locations in the U.S. The AI-in-QSR market is projected to reach $12 billion by 2034. Prior AI drive-thru efforts have struggled due to model quality and labor-focused goals. Arc trains brand-specific models using thousands of real drive-thru interactions and runs real-time A/B testing to measure accuracy, average order value, and order speed simultaneously.
"Arc trains brand-specific models on thousands of real drive-thru interactions (including specific accent and dialect exposure). Then, it runs A/B testing where operators can compare model variants in real-time, measuring accuracy, average order value, and order speed simultaneously."
"Arc's argument is that the previous wave failed for a specific, fixable reason: the players were using "80% good models trying to take out labor," MacLennan told me. "Chapter two-which I think is just starting now-is 99% good models focused on wins for customers, wins for employees, and wins for businesses.""
"Arc is a voice AI startup that takes orders and optimizes business performance at drive-thrus. MacLennan and co-founder Ali Hussain-both Square and Cash App veterans, where they built payment infrastructure-have raised $10.76 million seed funding, Fortune learned exclusively. Andreessen Horowitz led the round. The company is already working with two major fast-food chains, each with hundreds of locations."
"There are roughly 200,000 drive-thru locations in the U.S., with drive-thrus accounting for approximately 70 to 75% of quick-service restaurant (QSR) revenue. The AI-in-QSR market is projected to reach $12 billion by 2034. But the space is rife with failed concepts."
Read at Fortune
Unable to calculate read time
Collection
[
|
...
]