Skip to main content

9 Ways Restaurant Phone Artificial Intelligence Is Different Than Old-School Chatbots

Blog Image

Source: Canva

If you’re like most people, you probably cringe a little at the word “chatbot”. We all had terrible, clunky chatbot experiences in the 2010s. They couldn’t understand what we wanted, they were extremely limited to pre-written scripts, and in the case of voice assistants like Siri or Alexa, we got stuck in loops trying to pronounce words as clearly as possible.

“Order me a pizza, Siri. No! Not a trip to the Tower of Pisa. A P-I-Z-Z-A!”

Thankfully, artificial intelligence has experienced a seismic shift in the technology so significant, the AI-powered chatbots of today do not resemble their 2010s ancestors in the slightest. It only takes about three minutes of playing with ChatGPT to realize that the old-school era is over, and we’re in incredible, uncharted territory.

Let’s look at nine ways AI advancements have enabled restaurants to adopt automatic voice and text ordering without the limitations of old-school chatbots. In the article, we’ll cover…

  • How Natural Language Processing (NLP) tech is learning to understand nuance, accents, and word substitutions
  • Ways restaurants can train their phone ordering AI to understand their menu, ingredients, and customer preferences
  • Hooking up loyalty, rewards, and other systems to give the AI a sense for what individual customers want (and what upsell may work best)

The restaurant use case for AI in the phone and text ordering channel is hugely upgraded. Let’s dive right in.

Talking to AI-Powered Chatbots Sounds Natural, Finally

The old-school chatbots required very specific prompts, in very specific orders, to work. “Open the pizza store app. Add a pepperoni pizza. Add extra cheese.” It was absolutely terrible, because the limitations were unclear, and misunderstandings were frequent. But gone are the days of having to tell Siri or Alexa precise commands in order to get what you want. 

Modern AI chat experiences are far more fluid and natural than they were in the past. This is due to the development of large language models (LLMs), which are trained on massive amounts of text and voice data. These models can generate text and artificial voices that are more realistic and coherent than the text generated by previous chat systems.

In fact, modern AIs can sound so real, that some employees at Google recently thought that ChatGPT was sentient. It’s not (for now), but that doesn’t dismiss how impressive the technology has become. OpenAI’s GPT-4 large language model has successfully passed the Bar exam, AP exams, Medical school exams, and even a Sommelier certification. Simply put, this really is an intelligent system.

AI Voice Ordering for Restaurants Actually Works Now

These new Natural Language Processing capabilities have real-world impacts on how AI can be used in a restaurant setting to answer phones, text orders, or even run the drive-thru.

We’ll speak from our perspective developing OrderAI, an AI-powered phone and text ordering platform just for restaurants. By leveraging these new NLP capabilities, OrderAI is completely unrecognizable next to the chatbots of old—the experience is infinitely more natural, seamless, and user-friendly. Here’s a look at some of the things OrderAI can handle that old chatbots could not:

1. Multiple Speakers and Background Noise

Phone orders often feature multiple speakers: a primary speaker, and a person or three in the background making comments, like “Don’t forget to order my pasta”. OrderAI distinguishes between speakers and understands that when the primary speaker then says, “And we’d like an order of pasta too”, there weren’t two pasta dishes ordered—just one. 

2. Mid-Order Changes

Order alterations are common, especially during phone orders. A customer may realize that, no, actually they don’t want a medium pizza—they want a large. OrderAI identifies language and tonal cues that determine this to be an order change, not an addition. 

3. Order Sentence Variations

There are many ways a customer can say they’d like to order an item. Add in modifiers, and it gets complicated quickly. One common example is ordering a half-and-half pizza. Some customers will use language like, “A large pizza with half pepperoni and half sausage.” Others will say, “We want sausage and pepperoni. Make it a half-and-half.” There are many other variations as well. OrderAI learns to interpret these variations during training and onboarding. 

Also Read: How AI Is Humanizing the Restaurant Ordering Experience

4. Unclear Modifiers and Tie-Breakers

Complex orders with many modifiers require human order-takers to either ask the customer clarifying questions, or use context clues to infer the right meaning. For example, when a customer orders “veggie pizza with barbecue sauce”, it’s not clear if the customer wants the sauce on the pizza or on the side. OrderAI can ask questions to clarify the meaning, then identify trends over time to make assumptions like a human would.

Some orders become even more complex, like “sixteen wings boneless and sixteen regular wings with spicy garlic sauce”. Does the sauce belong on both sets of sixteen wings, or just on the second set of regular wings? OrderAI listens for clues and asks clarifying questions when needed.

5. Language Variations

A customer may use many words to describe the same dish. For example, some customers may use the word “regular” to describe wings that the restaurant calls “bone-in” on their menu. Or a customer may confuse a “pesto peach pizza” with a “pesto peach flatbread”. OrderAI learns the variations and synonyms to understand customer intent, even when the customer language doesn’t line up perfectly with the menu language.

6. Problem Identification and Escalation

OrderAI Talk has a 92% order success rate, demonstrating the sophistication of the artificial intelligence to understand the customer intent. However, there are inevitably times when heavy accents, frequent order changes, and other hiccups may challenge OrderAI’s understanding of the customer’s intent. OrderAI uses both language cues (repeatedly saying “No that’s not right”) and tonal cues (frustrated or impatient tones) for ongoing sentiment analysis to determine when the order is best suited for a human operator and can ring the store line for a human hand-off.

Even Better, AI Can Now Connect to Your Restaurant Systems

Legacy chatbots were notoriously siloed off, rarely capable of integrating into existing restaurant systems, like the point of sale or loyalty program. But thanks to how powerful the NLP capabilities are in modern AI chatbots, we can actually connect the AI to your restaurant systems quite easily, then train the AI on your own data, customers, and menu. 

This allows restaurants to not only have a natural and flexible phone ordering bot, but one that can react in real-time to customer or order data.

7. Answering Food Questions

During phone orders, customers often ask questions about food on the data level, like diet approvals, allergens, or calories. OrderAI can use any metadata attached to the menu to help customers understand their options. Questions that OrderAI might answer with the right metadata can include: “What do you have that’s vegetarian?”, “Can you remove all the dairy from that dish?”, and “How many calories does the chicken have?”

Also Read: 4 Ways Artificial Intelligence Is Helping Successful Restaurants Scale Faster

8. Mid-Order Upsells

OrderAI always offers customers an upsell, both during Talk and Text orders. Rather than a generic upsell for everyone, OrderAI experiments with offers based on order history to find offers that are highly likely to convert. For example, if OrderAI notices that a customer has ordered a soda five out of seven orders, but hasn’t in this order, it may suggest adding a soda. However, the bot will also see that the customer doesn’t have a dessert in the cart, and perhaps they want one. OrderAI will decide in real-time which offer to present, then use the results to train itself on customer preferences for the next order.

9. Text Marketing Offers

In the same way that OrderAI experiments with upsells, the bot runs tests with personalized offers at the individual level as well. OrderAI Text Marketing Feature offers also use the variable of time to incentivize customers to order more frequently. For example, if a customer places an order on Tuesday for a cheese pizza, the OrderAI Text Marketing feature will send them an automatic text message the following Tuesday asking if they would like to reorder their favorite cheese pizza.

Phone Order Automation Was Risky, But Now It’s Not

No restaurant leader in their right mind would have banked the phone ordering experience on the legacy chatbots of the 2010s. They were horrible experiences, and surely would have alienated and frustrated huge portions of customers. 

But today, these issues are largely resolved.

  • AIs can understand the gritty nuances of language and interpret unclear language, misspoken words, and other communication hiccups with shocking accuracy
  • AIs can bring data about a restaurant’s customer and menu into a phone order into the conversation in real-time to make it feel—dare we say—human
  • AIs can free up human teams to complete tasks that are best suited for humans, without forgetting to show up to work, forgetting to upsell, or having a rough day

OrderAI represents a monumental shift in the safety of phone order automation, unlocking these capabilities for growing restaurants that want to exist at the forefront of channel and customer experience innovation. And it’s already proven with over 3,000,000+ orders processed successfully.

Want to give OrderAI a spin for yourself? Reach out for a personalized demo.

You won’t believe how natural it feels to order from an AI.

The post 9 Ways Restaurant Phone Artificial Intelligence Is Different Than Old-School Chatbots appeared first on HungerRush.