TL;DR (jobs + people, plain English)
- What happened: Dairy Queen is adding Presto’s AI chatbot to dozens of U.S. and Canadian drive‑thrus. The company says the AI will speed up orders and encourage customers to add items. Source: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
- What this means for people: Short, repeatable tasks at the speaker box—taking standard combo orders, confirming items, and offering scripted upsells—are most likely to be handled by the AI. Humans still run the kitchen, fix mistakes, and handle exceptions.
- How shifts change: staff will monitor the AI, verify flagged orders, and step in on substitutions, complaints, or unusual requests.
- Quick operational numbers to watch (recommendations, not company claims): aim for ≥98% order accuracy before scaling; run a continuous 2‑week check; keep one staffed lane for escalation during rollout.
- Short example: during a 10:30 AM rush, the AI takes a standard combo. It adds a prompted fries upsell. A customer asks for no pickles. The AI flags the substitution. A crew member reviews the order, confirms the change, and tells the kitchen to adjust.
What the sources actually say
The Verge reports that Dairy Queen is deploying Presto’s AI chatbot in dozens of drive‑thrus to speed orders and prompt add‑ons. The article quotes company statements about speed and upsells and places this rollout in the wider trend of AI at quick‑service restaurants: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
The Verge piece is a public announcement. It does not claim the rollout removes all on‑site staff or that every interaction is fully automated. Prior reporting (summarized in coverage threads) has shown similar systems sometimes use human reviewers or remote agents to assist or audit conversations. That background explains why human oversight may still be part of the setup, but the exact scope is vendor dependent.
Which tasks are exposed vs which jobs change slowly
Quick rule: exposed = tasks the AI can take on quickly. Slow = tasks that need hands, local judgement, or complex social skills.
| Task | Likely automation window | Recommended human guardrail | |---|---:|---| | Scripted order taking for standard combos | months (fast) | Human verification for flagged mismatches; immediate handoff for substitutions | | Order confirmation and routine upsell prompts | months | Daily audit samples and weekly prompt tuning | | Handling complaints, refunds, emotional customers | years (slow) | Staffed escalation lane and written repair scripts | | Cooking, assembly, food safety | years (slow) | On‑site trained staff retain these duties | | Lane management, cash/security, complex reconciliation | 6–24 months (reshaped) | Train staff on new reconciliations and keep human oversight |
These distinctions are grounded in the reporting that the rollout focuses on the speaker box. Kitchen and physical tasks remain human work: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
Three concrete personas (2026 scenarios)
Persona 1 — Maria, 22, Drive‑thru Crew Member (US)
- Before: Maria took most speaker orders, handled payments, and told the kitchen about rushes.
- After: The AI takes standard combos. Maria watches the order feed, intervenes on substitutions and flagged items, and completes payment when needed. She keeps a one‑page shift sheet with escalation phrases and a two‑line apology/repair script.
Persona 2 — Samir, 35, Franchise Owner/Operator (US)
- Before: Samir staffed peak hours by experience and gut feel.
- After: Paperwork shows higher throughput but a small set of flagged errors. Samir keeps one staffed lane as an escalation path. If accuracy drops below his threshold, he pauses expansion and asks the vendor for prompt tuning and an audit.
Persona 3 — Aisha, 29, Remote Quality Reviewer (contractor)
- Before: Call reviews were local and ad hoc.
- After: Aisha reviews conversations the system flags for edge cases. She tags transcripts and marks examples that need retraining. She follows rules to redact PII (personally identifiable information) before labels are reused.
Each persona reflects the public report that Dairy Queen is adding Presto’s AI chatbots to drive‑thrus and that human oversight can remain part of the flow: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
What employees should do now
- Learn the new flow: read the one‑page shift worksheet every shift. Know the 30‑second escalation script.
- Practice repair language: create a 15–30 second apology + fix line for wrong orders. Rehearse weekly.
- Build durable skills: focus on face‑to‑face upselling, conflict de‑escalation, kitchen speed, and multitasking. These skills stay valuable.
- Ask about privacy and review: request written answers on whether conversations are reviewed remotely and how voice data is stored and redacted. Use the public report when asking managers: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
Employee checklist (printable)
- [ ] Read the one‑page "AI lane" worksheet before shift start
- [ ] Run a 1‑minute escalation drill with a partner (weekly)
- [ ] Confirm who handles refunds/compensation on AI errors
- [ ] Report recurring failure modes to the manager
What founders and managers should do now
For founders / franchise owners:
- Require vendor disclosures on human review and data handling before signing contracts. Use the Verge report as a public signal to request those clauses: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
- Budget for retraining and reassignment. Plan 30/90 day followups.
- Publish an incident response plan for AI‑linked order errors. Keep a visible staffed escalation lane during rollout.
For managers / local operators:
- Set rollout gates and metrics. Example target: ≥98% order accuracy measured over a continuous 2‑week window.
- Keep a staffed lane during testing. Run weekly QA to surface patterns.
- Require vendor audits on privacy and service levels every 90 days. Ask for a documented correction timeline for prompt changes.
Decision table example (manager):
- If accuracy ≥98% and complaints <5/1,000 over 2 weeks → continue rollout.
- If accuracy 95–97% or complaints 5–15/1,000 → pause expansion; request prompt tuning.
- If accuracy <95% or complaints >15/1,000 → roll back AI lane and investigate.
France / US / UK lens
France
- Worker consultation: French labor law and works councils may require consultation before major work‑organization changes. Prepare to consult staff or representatives.
- Data rules: GDPR (General Data Protection Regulation) applies. Consider a DPIA (Data Protection Impact Assessment) for voice processing and human review.
United States
- Law is patchwork. Focus on wage and hour rules if duties change. Watch state privacy laws (for example, California) about voice data and recordings.
- There is less uniform requirement for formal worker consultation. Use clear contracts and local communication to reduce conflict.
United Kingdom
- Data protection law and guidance on automated decision making suggest a DPIA is prudent for voice processing. Disclose human review to staff and customers where required.
Compact jurisdiction checklist (France / US / UK):
- Set a data retention limit and enforce it.
- Post customer notice at the speaker box and include info on receipts.
- Record worker consultation steps where required.
- Complete a DPIA or equivalent privacy assessment.
These jurisdictional notes align with the report and with the fact that similar systems often pair AI with human review: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
Checklist and next steps
Assumptions / Hypotheses
- The Verge article is the primary public snapshot that Dairy Queen is deploying Presto’s AI in dozens of drive‑thrus: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto
- Prior reporting suggests human‑in‑the‑loop review can be part of these systems. The exact scope, location, and number of reviewers is vendor‑dependent.
- This brief assumes stores will keep on‑site staff for kitchen work and for escalation during initial rollouts.
Risks / Mitigations
- Risk: accuracy drops and customer complaints rise. Mitigation: enforce a rollout gate with a threshold (example: accuracy ≥98% over 2 weeks) and clear pause/rollback triggers.
- Risk: undisclosed remote human review creates privacy and labor concerns. Mitigation: require vendor disclosure and contract SLAs; run a DPIA where required.
- Risk: staff feel deskilled or threatened. Mitigation: provide retraining budgets, short‑term reassignment funds, and clear role descriptions for AI‑assisted lanes.
Next steps
- Employees: carry and use the one‑page shift worksheet every shift. Complete the weekly 1‑minute escalation drill.
- Managers: implement the rollout gate template (metrics: ≥98% accuracy, 2‑week window). Keep one staffed lane for escalation. Run a vendor audit within 30 days.
- Founders/owners: require vendor disclosure on human review and privacy. Budget retraining funds and set a 90‑day operational review cadence.
Immediate 30/90 checklist (assign owners and deadlines):
- [ ] 30 days: vendor disclosure obtained; one‑page staff worksheet published; weekly drill schedule set
- [ ] 90 days: first operational audit completed; rollout gate metrics reviewed; retraining budget allocated
Primary public source: https://www.theverge.com/ai-artificial-intelligence/913928/dairy-queen-ai-drive-thru-presto