McDonald’s Failed “AI Drive Thru” Is Wake-Up We Need

There’s a pervasive narrative in Silicon Valley and the broader tech industry: AI is coming for everyone’s job. From doctors to lawyers, coders to baristas, and even brain surgeons. No one is safe.

And then you have the shills; the people hyping AI, pushing the hyperbole. I follow a lot of people on X, subscribe to myriad newsletters. Believe me: the hyperbole is real, and more often than not it comes from three distinct sources:

1. Those that make AI models (Google, Meta, OpenAI)
2. Those that sell them (snake-oil SEO types, new startups, SaaS companies)
3. Those that promote it (affiliates).

But when you look at actual implementations in the real-world, outside the bubble of newsletters and tech bro posts on X, this narrative quickly unravels.

And perhaps the best example of this is McDonald’s messy dive into the world of AI automation with its Drive Thru.

What Happened With McDonald’s AI Experiment?

McDonald’s, the company that basically invented process optimization, decided to test AI-driven voice ordering at more than 100 of its U.S. drive-thrus.

The idea was simple: let an AI system, developed by IBM, take orders in place of human workers.

Customers would speak their order, and the AI would process it. The idea being that this would supposedly speed up service and reduce labor costs.

Overheads walk on two legs, after all.

But the experiment was a disaster, and not just the occasional-messed-up-order kinda deal. It literally couldn’t have gone any worse.

The AI frequently misunderstood orders, especially when faced with background noise, varied accents, or even simple requests.

Social media exploded with viral videos of the AI’s failures: bacon was added to ice cream, $211 worth of chicken nuggets appeared on a single order, and customers struggled to get just a vanilla ice cream and a bottle of water, instead receiving multiple sundaes, ketchup sachets, and butter.

The system’s slow response times actually increased wait times, frustrating both customers and staff.

By mid-2024, McDonald’s had instructed its franchisees to remove the technology, ending the trial and admitting that, while there were some successes, the system was not yet reliable enough for widespread use.

We use AI a lot to run the site and manage things. But we never let AI do its own thing; it’s always inside a system that operates away from anything important. It is never allowed to mess with anything on the site or any of the content.

We use it to parse data inside AirTable, pull information from Sheets, and assign tasks to the team, and always with a human in the loop.

AI left to its own devices always goes haywire. You can see this in real-time using any modern LLM. Push it a little too hard or get it to do something without explicit data or prompting and it’ll just do whatever the hell it likes.

This applies to small setups like us, SMEs, and large billion dollar SaaS companies. No one in their right mind would give an AI carte blanche inside their business. And if they do, well… there’ll likely be hell to pay.

Think about it: If McDonald’s, with its legendary focus on efficiency, can’t get AI to reliably handle a basic drive-thru order, what does that say about AI’s readiness to replace jobs that require nuance, flexibility, or judgment?

Big Tech’s AI Push: Hype Over Substance

So where is this narrative coming from? As I said earlier: it’s hyperbole, plain and simple. Big Tech needs billions and billions of dollars of growth between now and the next quarter, and the only way to do that is with something new, something big.

A decade ago it was phones and apps, now it’s AI. This is the real reason it is everywhere, not because it is some kind of salve that will solve all the world’s ills.

Companies like Google, Meta, and Microsoft are investing heavily in AI because they need a new growth story.

The smartphone, search, and social media markets are saturated, and AI is the next frontier for investor excitement and stock market gains.

The push is often more about maintaining momentum and market valuations than about delivering real-world value.

This isn’t to say AI isn’t powerful or massive useful; it is, especially when used in the right context.

But the current splatter-approach to its deployment, whereby AI is lumped inside everything and does everything, is NOT the context Isaac Asimov likely had in mind.

It’s just marketing, word-soup. We saw it with apps, with 5G, with the new Star Wars. It’s just marketing to make a buck.

Most current AI apps and use cases, including the endless stream of trendy, vibe-coded tools popping up daily, are pointless. They solve no real problems, offer no real value, and will never gain meaningful market traction.

The most successful, commercial AI venture so far is ChatGPT. And the reason? It’s useful and it’s easy to use. Soon

AI’s Failings Are Creating New Jobs

Contrary to the fear-mongering, AI isn’t just replacing jobs, it’s also creating new ones and changing the nature of work.

For example, Ikea announced it would phase out call center work in favor of an AI bot called Billie.

But rather than simply laying off workers, the company is upskilling thousands of call center employees to become interior design advisors, opening up new career paths.

The global people manager at Ingka Group stated that AI would lead to new jobs and development opportunities for existing workers, not just job losses.

Similarly, studies by the World Economic Forum predict that by 2025, AI could create around 97 million new jobs across sectors like healthcare, technology, and finance.

These new roles will require different skill sets, often necessitating training and development programs.

While AI is adept at automating routine tasks like copying, pasting, and transcribing, complex strategy, judgment, and critical thinking remain firmly in the domain of humans.

Real-World Failures and the Limits of AI

The limitations of AI are evident beyond McDonald’s. Chili’s paused its test of server robots after realizing they didn’t enhance the guest experience as hoped.

AI chatbots in customer service often frustrate users with irrelevant or scripted responses, turning simple queries into confusing ordeals.

Even Amazon’s AI hiring system was scrapped after it was found to be biased against women, highlighting the dangers of over-reliance on unvalidated algorithms.

In creative fields, AI-generated content often requires significant human editing. Agencies that replaced copywriters with AI tools found themselves scrambling to clean up SEO-damaging content, ultimately hiring editors to fix the mess.

This isn’t efficiency: it’s a cycle of cutting corners and paying twice to get the job done right.

Why You Shouldn’t Worry

Today’s AI is good at specific, narrow tasks but still struggles with basic errors, context, slang, and tone.

It can “hallucinate” (generate false or nonsensical information), and its decision-making processes are often opaque, a “black box” that even its creators don’t fully understand which is why it has no place in search engines (hello, Google, what the heck are you smoking!?)

AI lacks common sense and can’t reliably handle situations that require human intuition or judgment.

The McDonald’s drive-thru debacle is just one example of how AI can fail in the real world. If AI can’t reliably take a burger order, it’s unlikely to replace jobs that require complex reasoning, empathy, or creativity anytime soon.

Final Thought

If AI can’t get your cheeseburger order right, it’s probably not going to be writing your will or managing your finances anytime soon.

In the meantime, learn how to use AI, find out what it can do for you and how you can use it to expedite boring tasks and work more efficiently. For this, the current LLMs we have and platforms like n8n are brilliant.

For everything else? Keep a level head and don’t drink the Kool Aid. AI is huge, make no mistake, and it will have huge implications (good and bad) for the world inside the next decade. But right now? I’d say Big Tech’s ego is writing checks its AI models cannot cash…