No Jitter is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Interactions Brings Conversational AI to Food Service

  • Interactions brings conversational AI to Pizza Pizza
    Artificial intelligence (AI) company Interactions this week announced an expansion of its product portfolio for the food service industry. Pizza Pizza, Canada’s number one pizza chain, is among the first restaurants to implement the company’s new product, Guest Experience Platform (GXP).
    Restaurants can implement Interactions GXP to enhance phone, digital, and drive-thru ordering in order to offer digitally transformed guest care, reduce operational costs, and boost overall revenue.
    This slideshow provides information on Interactions, the technology behind GXP and its other solutions, and the results companies like Pizza Pizza are enjoying with conversational AI.
  • Intelligent virtual assistant vs. chatbot
    Intelligent Virtual Assistant versus Chatbot
    AI is a term that has been in use for decades. As early as the 1950s, scientists, mathematicians, and philosophers explored AI’s technical and moral boundaries.
    In the world of enterprise communications, and specifically here at No Jitter, the first use of the term came in a 2008 post that posited a future capability: “voice recognition tied to artificial intelligence.” But only in the past two or three years have posts including the term as a headline grabber become a regular occurrence.
    Given the newness of AI’s application in communications and customer care, many see chatbots and intelligent virtual assistants as the same thing. However, there's a significant difference between the two that will be useful to understand as we explain Interactions’ GXP and how Pizza Pizza is using it.
    • Chatbots are generally for acquiring information such as store hours or simple product data.
    • Virtual assistants can assist in conducting business, like buying a product or asking for a return label. If you ask a chatbot for such virtual assistance, it will likely get confused, and ultimately keep asking the same questions for clarification.
    Both are considered conversational interfaces, yet each has a distinct purpose.
  • Interactions description
    Founded in 2004, Interactions is a global company headquartered in Franklin, Mass. The company offers intelligent virtual assistants that assimilate conversational AI and human understanding to enable businesses to engage with their customers in productive conversations.
    The privately held company reported that in 2018 revenue exceeded $100 million for the first time. Interactions’ five-year 40% compound annual revenue growth rate exceeds even the lofty growth most cloud contact center companies are delivering. It’s no surprise that the company has attracted top contact center talent, including CMO Jim Freeze, formerly CMO of Aspect Software; and Dave Rennyson, former general manager for cloud at Genesys.
  • How the restaurant industry is transforming
    The Restaurant Industry Is Transforming
    In its 15-year history, Interactions has been successful in a broad array of consumer-to-business markets, including financial services, retail, travel, and utilities. Food services is a specialized vertical within retail, with some characteristics common to other types of retail but also some unique requirements.
    I discussed Interaction’s rationale for building the GXP in an interview with Priyanka Tiwari, the company’s director of product marketing. “A lot of things are changing in food services. People are spending more and more money eating at restaurants. And they are actually preferring to eat off-premises, spending more money to buy food and taking the food outside the restaurant to eat,” Tiwari commented. Most importantly, from the perspective of Interactions and No Jitter readers, 95% of restaurant-goers agree that technology improves efficiency.
    Sources for the above data points are:
    1. 2019 State of the Restaurant Report
    2. Interactions internal data
    3. Nation’s Restaurant News
    4. The United States Bureau of Labor Statistics
    5. Access: Customer Engagement and Loyalty Statistics
    6. Restaurant Technology in 2017 Toast
  • GXP voice and data input
    GXP: Voice and Digital Input
    The GXP solution features modules that cater to the entire restaurant guest journey -- from digital ordering to care to social media engagement. GXP supports voice channels like phone, drive-thru and smart speakers, as well as digital channels such as chat, SMS, social media, and messaging platforms.
    Interactions offers this breadth of channel choice and modalities for all of its clients, in the food services business or any other. That said, Interactions stresses that for its customers, voice is the most difficult channel to offer using AI. As a result, voice is where the majority of the company’s research and development effort is put -- and patents received.
  • Adaptive understanding diagram
    GXP: Adaptive Understanding
    On the right side of the above diagram is a graphical depiction of Interaction’s Adaptive Understanding “secret sauce.” The technology blends conversational AI with human intelligence to provide understanding every time, even amid noisy restaurant environments or wind and rain at drive-thrus.
    Looking at the lower-right portion of the Human Intelligence block may lead you to think that humans listen in on all interactions. This is far from the case.
    AI may not always understand what a customer is saying. In those cases, Interactions offers an additional source of recognition to teach and improve the algorithms. In real-time, GXP will engage a human “intent analyst” who, within a fraction of a second -- as I witnessed during an analyst briefing at company headquarters in May -- reviews, identifies, and feeds the customer utterance back to the application. These intent analysts work on behalf of Interactions to offer an additional source of recognition to the algorithm. Intent analysts never engage with the end customer directly and never hear more than a few seconds of an interaction; each is clarifying hundreds of utterances an hour. Importantly, intent analysts never see all the pieces of sensitive data. As shown in the slide, Interactions adheres to industry standard security certifications. The company holds the privacy and security of its client’s data at utmost importance.
    Intent analysts also perform a crucial task of completing the machine learning loop, by regularly labeling and tagging the data and intents from data that the AI had low confidence recognizing. The correct answer is fed back into the AI, to improve the grammar and models. This way the system constantly learns from its experiences and gets smarter each time.
  • Kitchen AI diagram
    GXP: Kitchen AI
    The goal of the GXP is to take the place of a human without sacrificing quality by providing behavior like the restaurant’s best staff. Look to the center of the image above to see the restaurant-specific tasks that GXP can perform.
    Kitchen AI is a one of GXP’s food service-specific elements. Kitchen AI helps automate aspects of food preparation -- what Josh Bob, senior product leader at Interactions, called “the back of the house.”
    In a typical food establishment, kitchen staff put on a pair of gloves to prepare an order. They then take off the gloves to press a button signaling that the order is complete. And then repeat, over and over again. Replacing that button press with one verbal command such as, “Close out order 456,” can increase employee throughput by up to 15%, Interactions said.
  • Prediction recommendation engine schematic
    Predictive Recommendation Engine
    Another food services-specific element of GXP is a proprietary Predictive Recommendation Engine. In analyzing more than 100,000 customer interactions across a number of different types of restaurants, Interactions discovered that only 30% of customers received an upsell option, Bob said during our briefing.
    Of those, he continued, one quarter of the customers accepted the offer, resulting in upsells in 7.5% of interactions. If, by using GXP, restaurants offer an upsell 100% of the time, that revenue would grow by more than 300%. Put another way, Bob said, “restaurants are leaving money on the table... no pun intended.”
    In addition to ensuring that restaurants are consistently offering upsells, the Predictive Recommendation Engine pairs contextual cues and historical trends with restaurant-defined business rules so that suggestions offered are both of a type and at a time that a customer is most likely to accept them. For example, if a customer has said she wants vegetarian food, she wouldn’t receive a special offer on a meat lover’s pizza.
  • Current state of restaurant customer service schematic
    Making a Business Case for GXP: Current State
    While Pizza Pizza is the only customer listed in Interactions’ press release, several additional restaurant chains have come on board in the last year, Interactions told me. While Interactions anonymizes its customer data, it shared this real-life current-state example of a pizza restaurant with very high take-out and delivery volumes (the numbers have been rounded, but the percentages are accurate):
    • First, because every call has to be answered by a human, 10% of inbound calls go unanswered. Some of these calls may be at 6 a.m., or maybe the store was so busy that nobody had a free hand to grab the phone -- regardless, 1,000 calls become 900.
    • Those 900 answered calls often come in during peak times, so they’re likely to be put on hold.
    • With an average hold time of one minute, 17 seconds, another 50 callers drop -- possibly to order from another restaurant. One thousand callers have become 850 potential customers.
    • 170 of the remaining callers are just looking for information -- “What time do you close?” or “How long is the wait?” That leaves 680 orders.
    • At an average of $18 per order, that’s about $12,000 in call-in sales -- combined with online and dine-in, the restaurant brought in about $44,000 today.
    • And because the calls are coming at peak times, it’s unlikely that the employee is offering the customer any sort of upsell. Keep that in mind.
  • Future state restaurant depiction
    Making a Business Case for GXP: Future State
    Let’s see what happens if every inbound call is run through the GXP.
    • First, every single call gets answered. Conversational AI routes 140 calls through its informational system and successfully handles 560 orders from start to finish. That leaves only 300 calls -- 70% less volume -- to pass along to a human at the restaurant.
    • When the system handles a call, it manages to boost sales by an average of $1.50 per order --– so 560 orders jump from a value of $18 to $19.50.
    • The 300 human-answered calls have a queue time of just 34 seconds, meaning the abandon rate becomes virtually nil.
      • Of those, 20% are looking for information -- say, “I’d like to talk to a manager” or “I want to give Larry a message.”
      • The remaining 80% are orders requiring human assistance, due to complexity or other reason.
    At $18 an order -- assuming no upselling by the store employees -- that’s $4,000 coming in through call-in orders, with the bulk of orders now handled digitally. The use of GXP resulted in a 7% bump in same-store sales; $3,000 a month is $36,000 a year per location. For a 200-location chain, that’s more than $7 million in increased annual revenue.
    In addition, there’s a major reduction in average hold time and abandoned calls -- and restaurant managers know that every call answered is an opportunity to build a loyal customer.
  • AI is vertical story
    AI Is a Vertical Story
    One of the reasons this announcement intrigued me was the deep vertical specialization Interactions created to address a set of specific industry requirements. There are great AI use cases in a number of different industries, including healthcare, education, marketing, and retail, and each requires some level of industry expertise to perfect. Products like GXP will allow more companies to both see the very specific value AI can bring to their businesses and decrease the overall time and cost to make such systems deliver value.
    As it was in 2019, expect AI to be a hot topic at Enterprise Connect 2020, in the Contact Center & Customer Experience track I lead, and across the whole conference. (Mark your calendars for the week of March 30, 2020!)

An example of how to deliver industry-specific value to businesses through AI