The Big AI Ordering Stumbles—and What Restaurants Can Actually Learn

Here's the detail that doesn't fit the usual story about technology: the companies that stumbled most publicly with AI ordering weren't scrappy startups cutting corners. They were some of the largest, best-funded names in the business, working with serious technology partners, on the most-studied menus on earth. If they hit walls, the tempting conclusion is that AI ordering simply doesn't work yet—and that conclusion is wrong. The stumbles are real, they're on the public record, and they're worth studying carefully. But read closely, they don't say "AI can't take an order." They say something much more useful about where and how AI ordering breaks—and almost all of it points away from the place AI ordering actually works best today: not the drive-thru lane, but the phone.

This is an industry-education piece, not a victory lap. Every case below is limited to what has been publicly reported, attributed to its source. The goal is to learn from expensive, well-documented experiments so the next restaurant doesn't repeat them.

The cases—just the public record

Four episodes have shaped how the industry talks about AI ordering. Here is what each one actually says, and nothing more.

McDonald's and IBM: a test that ran, then ended

McDonald's began an AI partnership with IBM in 2021 to test drive-thru "Automated Order Taking." In June 2024 it ended that IBM test and removed it from the more than 100 restaurants where it had been running. Publicly reported challenges included difficulty with accents and dialects affecting order accuracy. Notably, McDonald's said voice ordering may still be part of its future—as reported by CNBC and Restaurant Business, 2024. That last line is the part most coverage skipped: this was a company ending one specific test, not declaring voice AI a dead end.

Presto Automation: a regulator weighs in on the claims, not the tech

This case is different in kind, and the distinction matters. According to the SEC's enforcement order, January 2025, the U.S. Securities and Exchange Commission announced settled charges that Presto Automation made false and misleading statements about its "Presto Voice" AI product. Per the SEC, the company had claimed its AI eliminated the need for human order-taking, but in fact the vast majority of orders—about 70%—required human intervention. Presto was delisted from Nasdaq in September 2024 and disclosed "substantial doubt" about its ability to continue as a going concern—as reported by Restaurant Business.

Read the Presto matter precisely: the SEC charged the company with false and misleading statements about what the product could do. The lesson here is about over-claiming—the gap between a marketing promise of full automation and the reality that most orders still needed a person—not about whether AI can help take an order at all. Some coverage labeled it "AI-washing." That's a useful term for the pattern, and it belongs in quotes attributed to that coverage.

Taco Bell and Yum Brands: scaling up, and refining in public

Yum Brands rolled AI voice ordering out to hundreds of U.S. Taco Bell drive-thrus—reports cite 500+ by 2025. The rollout drew viral social-media clips of the system mishearing orders or adding unexpected items. Crucially, Yum has said it is continuing and refining the technology; it has not abandoned it—as reported by CNBC, 2024–2025. Whatever you make of the viral clips, the corporate posture is "keep going and improve," not "pull the plug."

Taco John's and "Olena": working at scale, removed where it didn't land

Taco John's deployed an AI voice bot called "Olena" that reportedly handles roughly 90–93% of orders without staff stepping in. At the same time, the chain removed the AI from three locations in smaller communities where customers didn't take to it—as reported by Restaurant Business. This is the most quietly instructive case of the four: high success rates in most places, and the discipline to pull it where the local fit was wrong.

Why the big drive-thru bets stumbled

String the four cases together and the failure modes aren't mysterious. They cluster into a handful of forces that stack up against a drive-thru bot all at once—and against an over-promised one most of all.

The drive-thru is a brutal place to listen

Start with the audio, because it explains more than anything else. A drive-thru lane is one of the worst acoustic environments imaginable for speech: engine idle, road noise, a window cracked into the wind, a kid in the back seat, music from the dash, and a microphone mounted outside in the weather. Even a human worker asks people to repeat themselves there. Ask software to get it right every single time and you've set an unforgiving task before the AI has done anything wrong.

Accents, dialects, and the full range of how people talk

Among the publicly reported challenges in the McDonald's test was difficulty with different accents and dialects affecting order accuracy. Customers don't speak in one tidy accent, and a system that only understands some of them isn't serving everyone. Early speech models were simply less robust across the full range of real speech.

Long, complex menus

A modern fast-food menu is a combinatorial maze—sizes, modifiers, substitutions, combos, "no this, extra that," limited-time items, and regional differences. Mapping a messy spoken sentence to exactly the right item and exactly the right modifiers is far harder than it looks, and small mismatches surface as the kind of "added an unexpected item" clips that went viral.

Mid-order corrections

Real ordering is not a clean script. People change their minds out loud: "actually, make that a large—no, two larges, and scratch the fries." A system that can't track the conversation and reconcile a correction against what's already on the order will get tangled exactly where a human cashier wouldn't.

Rigid automation with no graceful exit

Perhaps the most important structural lesson: a rigid system that tries to complete every order itself, with no clean way to hand a confusing one to a person, turns every edge case into a failure. The fix isn't a magic bot that never needs help; it's a system humble enough to pass the call to a human the moment it's out of its depth. Many early bets lacked that safety valve.

Over-claiming what the AI could do

Finally, the thread that runs through the Presto matter as the SEC described it: promising more than the technology delivered. When the public pitch is "no humans needed" but the reality is that most orders still require a person, the gap itself becomes the problem—legally, reputationally, and operationally. Honest scoping of what AI handles and what humans handle isn't a weakness; it's the difference between a system that holds up and a claim that doesn't.

So what actually makes ordering work?

Here's the encouraging part, and it's not hand-waving. Most of what broke in those drive-thru bets is about the environment and the claims, not some permanent ceiling on the AI. Change the channel and the honesty, and the picture changes with it.

The phone is a far more controlled channel than a drive-thru

This is the quiet, decisive advantage. On a phone call, the caller is usually speaking directly into a handset, in a quieter setting, without wind, engine noise, or a back-seat conversation fighting for the microphone. The signal arriving at the system is simply cleaner. The single hardest variable in the drive-thru experiments—chaotic outdoor audio—is largely removed before the AI even starts to listen. Phone ordering isn't a harder version of the drive-thru problem; it's a more tractable one.

Modern speech understanding handles messy, real talk

Speech models today are meaningfully more robust to natural, messy speech and to a wider range of accents than the tools available when these tests began. They track context across a conversation, so when a caller says "make that two" or "actually, scratch the last one," the system knows what they mean. That doesn't make them perfect—nothing is—but the floor is much higher.

Completing the order in the POS, not just transcribing

The systems worth using don't hand your staff a transcript to re-key. They place the order directly into the point-of-sale and fire it to the kitchen—either as part of the POS or by integrating deeply with one, such as Square, Clover, Loyverse, Epos Now, or Revel. Grounding the conversation against your real menu and modifiers is what keeps the AI from inventing items, and completing it in the POS is what closes the gap between "heard the order" and "got the order right."

Multilingual by default

Modern voice AI commonly serves English, Spanish, and Chinese among other languages, detecting the caller's language and switching automatically. The same menu and modifier grounding applies in each language, so a Spanish-speaking caller's order maps to the same kitchen ticket an English-speaking caller's would—directly addressing the "only understands some people" problem the drive-thru tests ran into.

The safety valve the rigid pilots lacked: graceful human handoff

This is the heart of it. A well-built phone system stays in its lane and transfers to a person whenever:

• The caller simply asks for a human—caller preference always wins.
• The order is unusually large, a catering job, or from a known VIP who deserves a personal touch.
• The request is genuinely unusual or outside what the system can safely complete.

The point of AI here isn't to replace your staff or to trap callers in a bot—the very promise the Presto matter showed to be hard to keep. It's to catch the routine, high-volume calls so your team can give full attention to the ones that need a person. A rigid bet with no escape hatch makes every hard moment a failure; a graceful handoff turns those same moments into a smooth experience. The Taco John's decision to pull the AI where it didn't fit is the same instinct in a different form: know the limits, and act on them.

How to evaluate any AI phone vendor honestly

These four cases are the best buyer's checklist you'll get for free, because together they tell you exactly which questions separate substance from a slick pitch.

• What happens after the caller hangs up? If the answer is "we send your staff a transcript" or "we create a ticket someone confirms," that's manual re-entry in a nicer outfit. Insist on completion in the POS.
• Is it grounded on my real menu and hours, or a generic script that can invent items and quote prices that don't exist?
• When and how does it transfer to a human? There must be a clean escape hatch—for VIPs, large orders, and anything unusual.
• What languages, and does it switch automatically?
• What does the vendor actually claim? Be wary of anyone promising the AI fully replaces people with no humans involved. The honest answer is that AI catches routine calls and hands off the rest—and the Presto matter is a public reminder of why the over-promise is a trap.
• Can I hear it myself—on a real line? This is the big one. A canned recording proves nothing; the entire point of these stumbles is that systems behave differently in the messy real world than in a polished clip. You should be able to call a live demo and judge it with your own ears.

That last point deserves emphasis. Don't buy AI ordering from a slide deck or an edited highlight reel—that's precisely how the gap between promise and reality hides. Pick up a phone, call a live line, talk like a real customer—mumble, change your order mid-sentence, ask for a human—and listen to how it actually handles you. You can call our live demos at /#try; they're real lines, not canned recordings.

An honest note: no system is perfect, and the handoff is the point

It would be easy to end with "and that's why our AI never makes a mistake." It would also be false, and falseness is half of what tripped up the cases above. No ordering system, human or AI, is flawless—people mishear orders too. KwickPhone is not perfect either. The right goal isn't a system that never stumbles; it's a system that knows when it's stumbling and hands the call to a person before a small confusion becomes a wrong order. The graceful human handoff isn't a fallback bolted on for emergencies—it is the design.

Read together, the four stumbles are reassuring rather than discouraging for an independent restaurant. The hardest part of the drive-thru experiments—chaotic outdoor audio—doesn't apply to your phone. The most damaging mistake—over-claiming what the AI could do—is one an honest vendor simply doesn't make. And the solvable parts—speech robustness, menu grounding, POS completion, and a clean human handoff—are exactly what a good phone system is built around today. Even McDonald's said voice ordering may still be part of its future; Yum says it's refining rather than abandoning. The industry isn't walking away. It's learning which channel, and which honesty, make it work.

Frequently asked questions

Did big restaurant chains have AI drive-thru failures?

Several publicly reported stumbles exist. As reported by CNBC and Restaurant Business in 2024, McDonald's ended its IBM automated order-taking test and removed it from the more than 100 restaurants where it ran, citing publicly reported challenges with accents, dialects, and order accuracy—while saying voice ordering may still be part of its future. As reported by Restaurant Business, Taco John's removed an AI voice bot from three locations in smaller communities where customers didn't take to it. None of this means voice AI doesn't work; the lessons are specific and largely solvable.

What happened with Presto Automation and the SEC?

According to the SEC's enforcement order in January 2025, the U.S. Securities and Exchange Commission announced settled charges that Presto Automation made false and misleading statements about its Presto Voice AI product. Per the SEC, the company had claimed its AI eliminated the need for human order-taking, but in fact the vast majority of orders—about 70%—required human intervention. As reported by Restaurant Business, Presto was delisted from Nasdaq in September 2024 and disclosed substantial doubt about its ability to continue as a going concern.

Did Taco Bell give up on AI voice ordering?

No. As reported by CNBC in 2024 and 2025, Yum Brands rolled AI voice ordering out to hundreds of U.S. Taco Bell drive-thrus (reports cite 500+ by 2025). The rollout drew viral social-media clips of the system mishearing orders or adding unexpected items, and Yum has said it is continuing and refining the technology rather than abandoning it.

Why is phone ordering more tractable than a drive-thru lane?

Much of what made the drive-thru hard is about the environment, not the AI. A phone call is a more controlled channel: the caller usually speaks directly into a handset in a quieter setting, without wind, engine noise, or back-seat conversation competing for the microphone. Pair that cleaner signal with completing the order in the POS instead of just transcribing it, multilingual support, and a graceful transfer to a human, and most of the hardest variables fall away.

How do I evaluate an AI phone vendor honestly?

Ask what happens after the caller hangs up—does it complete the order in your POS or just send a transcript? Confirm it is grounded on your real menu and hours, ask how and when it transfers to a human, and be wary of any vendor that claims the AI fully replaces people. Above all, insist on hearing it yourself by calling a live demo line rather than trusting a canned recording or a marketing claim. No system is perfect, so the clean human handoff is what matters most.

Related: why McDonald's ended its AI drive-thru and what it teaches about phone ordering and the complete guide to AI phone answering for restaurants.