Toyota's plan for contact center AI deployment

Imagine that you're tasked with managing the technology that helps support overall owner satisfaction and customer retention for the world's largest automaker. How do you go about turning on AI to support contact center operations -- responsible for answering more than a million queries annually from drivers, potential buyers, dealers and field offices -- without losing ground in customer loyalty?

That's the mission of Mitch Aubin, DevOps product manager, Dealer Consumer Technology Solutions at Toyota. This group interfaces with customers, dealerships and field offices to address inquiries and complaints. His group is engaged in a long-term plan to integrate AI into its operations using a stack built on Amazon Connect and Salesforce, among other systems.

We sat down with Aubin and colleague Bradley White, technology, innovation and automation manager at Toyota, to discuss their progress so far: 4% of inquiries handled by self-service, 20% reduction in call handling times, and 13% fewer call transfers.

Editor's note: This Q&A has been edited for clarity and conciseness.

What was the problem Toyota needed to solve with AI and Amazon Connect?

Mitch Aubin

Mitch Aubin: Our guests [customers] call Toyota's Brand Engagement Center about anything under the sun. It might be a bad dealership experience. It might be just a question of how to connect their Bluetooth to the car. "Hey, there was a recall last night. I heard about it on CNN." The agents need to be well-versed in everything to provide their answers as quickly as possible and get the guest on their way to enjoying their day.

We wanted to transition to be more of an omnichannel contact center, and to have the capability to route those guests to the channel that makes sense for the type of questions that they had -- certain interactions are more sensitive, certain interactions can be best handled via chat with image sharing, things like that.

We had reached end of life with our on-premises Genesys IVR [interactive voice response] system, and we had a newly implemented Salesforce CRM to manage our cases and track the interactions. The Brand Engagement Center was looking for a system that would enable the agents to receive real-time transcription of the call so they would not have to take copious notes during the call. That was something they had to do; it was distracting them from being fully present during the calls.

We also wanted a single pane of glass for our agents to keep them from having to swivel-chair to all the multitude of applications that we have across Toyota, which they might have use to solve any number of the types of problems that our guests call about.

How has the introduction of generative and agentic AI during your transformation affected your plans?

Aubin: We made the selection in 2023; we weren't thinking about generative AI. Generative was this new technology that had taken the AI -- natural language processing and machine learning that was there -- to set the stage to know about me and know-my-vehicle type capabilities.

We have a traditional AI component that pays attention to who calls in, and perhaps we recognize you, the guest. You have a Camry. Maybe we just issued a recall. We can notify them of the recall. Were you calling to learn about that, or would you like to learn about that? We can have IVR play that and even offer to connect you to your preferred dealership to schedule an appointment to take care of the recall. So implementing some traditional natural language processing AI was definitely on our roadmap.

Now, with generative AI technologies, we have also taken advantage of real-time transcription for generative AI call summaries. At the end of every call, Amazon Connect Contact Lens enables us to have the call summary fed directly into the Salesforce screen, allowing us to record details such as "Customer called with this problem, agents provided this answer, and customer was pleased," or whatever the case may be.

The agent can look at that and say, "Yep, that's it." They can add a couple of bullet points if they wish and then save it. The big value there is that now we have -- in standardized, non-jargon language -- a summary of the call. That means a world of difference to people who are looking at a call for the first time. Our legal team really likes that, because they're able to understand without needing a glossary to clarify specific terms. They also don't necessarily have to dig into the real-time transcription, but that is available to them.

What we're not hearing from you is that AI is going to replace humans.

Bradley White: Yes, that is our approach -- how can we enable that human touch, to empower them to do better?

By the time they contact us, customers have already escalated their issue -- not necessarily through Toyota channels: It could be just online searches. "I'm struggling to get my Bluetooth connected," or "I've been on the website, and I do not know what the new color for the Land Cruiser or the C-HR that is coming out next year. What colors are being offered on that vehicle?" Or "My experience with the dealer did not meet my expectations."

So they typically have a point of contact prior to reaching out to us. You don't want to send them back to full automation, right? We want to enable our agents to be more efficient so that we can engage in a more personalized way by providing them with the information to be able to do that.

What's on the roadmap?

Aubin: Toyota is a big partner of Amazon Connect because Amazon is never in a static state. They're continually improving their roadmap and their tools.

When we originally implemented this, we aimed for a 5% reduction in call transfers but achieved a 13% reduction. That means our customers are connecting with the right expert within our brand engagement center on their first call. It's improving their satisfaction for both the customer and our team because our team is getting questions that they know the answer to and can handle quickly. The customers love that their time spent addressing the issue is also significantly reduced.

We will continue to enhance our IVR so that when people call, perhaps that IVR can answer the question for them and make that a pleasant experience for the guest. Toyota is looking at how we can quickly resolve a guest issue with IVR instead of saying, "Hey, here, talk to the hand."

Toyota is looking at how we can quickly resolve a guest issue with IVR instead of saying, 'Hey, here, talk to the hand.'

How can we, if they need to talk to a human, get them quickly on the phone with the human? Get the human to be able to answer the call quickly? We want to provide a great guest experience, allowing them to move on with their daily life, and for the agent to move on to help the next customer who needs attention.

We've heard that AI IVR works well for small businesses, but it becomes more challenging the larger the business. What does the world's largest automaker need from AWS to improve its AI IVR?

Aubin: We need everything for making sure that we can handle multiple parties on the call at the same time -- we can do that, it handles that today -- but the part where we need some assistance and are working with them on this is to make sure that transcription supports and identifies those multiple parties: Sometimes we have the guest, the dealership, ourselves and [maybe] another party on the call. We need to ensure that we can clearly attribute any statements made on the call to each of these speakers, so we have a clear picture of what happened after the fact.

Don Fluckinger is a senior news writer for Informa TechTarget. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.