Why did you chose to offer live chat and chatbots on the Icetro site?
In the foodservice equipment industry, making a high-quality, reliable product is only half of the game. The other half is providing good support for those products. There are a lot of manufacturers that flood a market with products but don’t provide proper support or make it easy to get parts, and so those machines just slowly die out.
Our goal has always been to provide reliable products and best-in-class support for those products. That means we make it easy for a field technician to get the information they need while they’re onsite servicing a machine. These service visits can be stressful for technicians because they’re trying to get a machine fixed as quickly as possible to keep their clients happy and so they can get on to their next call. If we can get them accurate information quickly, it makes their lives much easier.
One way we do that is by using Olark to offer ‘Tech Support Chat’ right on our homepage. This makes it easy for someone onsite to get online and ask a question. We use Aiden to field the initial inquiry. If it’s a simple part number or information that’s already contained in our documentation, Aiden can quickly recall that information so the technician doesn’t have to hunt for it. If it can’t answer the question, then it transfers the chat to one of our live agents who can work on deeper troubleshooting.
What challenges were you trying to solve with Olark?
- Reduce the amount of time our support team spends on the phone
- Reduce the number of simple tech calls (part numbers, etc.) so they can use more of their time on more difficult, in-depth calls
- Introduce more AI-powered technology into the Support team tech stack
Why Olark?
We had never offered chat or chatbots on our site before, so I talked to a lot of people to figure out which platform best suited our needs. Olark was highly recommended to us by Jason Wiser who runs Wiser Sites. The Aiden component of Olark - that it will work to answer a question first and then transfer to a live agent if it can’t - sounded perfect for what we were trying to do.
How was it getting Olark up and running on your site?
Very easy. Creating the knowledge base for Aiden was the most difficult part, and that was only “difficult” because we had to gather all of our content and documentation and put it into a question-and-answer format that the LLM could use to extract answers. Once we had all that information compiled, setting up the knowledge base and the contextual answers happened quickly.
When you’re getting started with Olark and Aiden, make sure you have all your documentation ready to load in. When you have that gathered, start to think about all of that information in a question-and-answer format. If you haven’t done that yet, or haven’t thought about that yet, start to spend some time getting your ducks in a row.
Also take the time to review the documentation you plan to upload and make sure that it’s accurate and free of spelling mistakes. When you build your LLM, you’ll only get out what you put in. If you add high-quality documentation, your bot will provide high quality answers. If you put garbage in, it’s going to put garbage out.
What’s been the feedback from your support team?
Olark’s been easy for our team to use. We had one team member who recently retired - this person was in the ice machine business for forty years - and I would say this person was not great at using software, but he got pretty good at using Olark to chat with people on our site.
The team has been impressed with some of the AI responses it has seen. Jose on our team said that even though questions can be worded confusingly, Aiden can extract the true question from the wording and provide the right answer. Because it’s not just the question being asked but the context behind it.
And Aiden’s ability to interpret context is better than other bots I’ve seen. Also, the way Aiden words its responses sounds more like a live person answering than a machine.
Or sometimes I’ll review the chats that Aiden is handling, or even monitor them in real-time, and I do see a lot of simple questions that Aiden answers where the customer is satisfied with the response they're getting.
What results have you seen from Olark?
Because we primarily use Olark for technical support, our return isn’t measured in sales or leads. I will say though that I think of Olark as an indirect sales tool for us. Businesses don’t buy ice machines very often—once maybe every five to ten years. When the time does come to buy a new ice machine, a business calls their local service person and asks which one they should get. If that service person had a positive experience servicing our machines previously, they’re going to recommend us.
There are a lot of great qualitative wins from our customer support team, but one win in particular is still fresh in my mind. One of our biggest customers is an enterprise company that leases our machines. I asked them, “What’s something you wish our chatbot could do to make your lives easier?” And they said they’d like to be able to quickly get part numbers when their technicians are in the field servicing parts. So I added a file in our knowledge base that has all the model numbers, part numbers, and descriptions of the parts. Now Aiden can quickly recall that data when someone chats, and that's been working really well.