McDonald's × IBM Automated Order Taker (AOT) — ended July 2024 after 100+ restaurant pilot

NA-USA

Content

McDonald’s × IBM Automated Order Taker (AOT) was the most publicised US QSR voice-AI drive-thru pilot of 2021–2024. It deployed to more than 100 McDonald’s restaurants in the United States. McDonald’s announced the end of the IBM partnership in a June 17 2024 memo to franchisees (obtained by CNBC), with the AOT technology scheduled for shut-off no later than July 26 2024.

The discontinuation was attributed by CNBC’s sources to issues interpreting different accents and dialects, affecting order accuracy. McDonald’s public statement left the door open to future voice-AI work (and a Google Cloud partnership was announced at McDonald’s investor day in December 2023), but as of mid-2026 no new McDonald’s voice-AI drive-thru deployment has been publicly reported.

The IBM partnership followed McDonald’s 2021 sale of McD Labs (formerly Apprente) to IBM, and a separate 2021 sale of Dynamic Yield (predictive ordering / menu personalisation) to Mastercard. McDonald’s has therefore divested both its in-house AI ordering assets within a few years.

What this unit is doing in the taxonomy

The most documented discontinued high-profile generative-AI consumer surface in US agrifood consumption. Useful as the cautionary case in any talk comparing voice-AI deployments — Wendy’s FreshAI and Taco Bell × Omilia both scaled; McDonald’s × IBM did not.

Why it matters for talks

• Documents the failure mode (accent / dialect accuracy) that mainstream vendor narratives underplay.
• Documents the divestment pattern: McDonald’s bought Apprente (2019) and Dynamic Yield (2019), sold both by end of 2021, and ended the IBM AOT partnership in 2024. The corporate strategy is moving away from owning voice-AI.
• The McDonald’s × Google Cloud partnership announced at McDonald’s investor day in December 2023 has not produced a publicly reported voice-AI deployment as of mid-2026 — important gap to track.

Critical context the unit does not sanitise

• The accent / dialect accuracy issue is a digital-equity problem: AI voice ordering systems that fail on non-standard accents create a two-tier customer experience. McDonald’s × IBM pilots were concentrated in specific US regions; the failure was most pronounced on customer dialects under-represented in training data.
• McDonald’s divestment of Apprente / Dynamic Yield means the company’s voice-AI future is now dependent on third-party vendors (Google Cloud, IBM in adjacent products). Sovereignty over ordering data flows to those vendors.