By Jutta Stienen -
Part 3 – How to implement a deep learning chatbot in customer service
In this post, we walk you through a 2-step guide to effectively implement a deep learning chatbot in your customer service. We also talk to Sascha Kappeler, Head of eCommerce at BRACK.ch, about his experience with the implementation of this technology.
If you read Part 1 and Part 2 our 3-part series, it should be clear that advanced chatbot technology such as a deep learning chatbot is not a simple plug & play purchase.
Understanding chatbot technology for customer service optimization.
This blog post is part 3 of a 3-part series. In case you missed the other 2 parts, here they are.
Part 1 – Debunking 8 myths about chatbots in customer service >
Part 2 – Deep learning and machine learning chatbots: how do they work? >
What do you need to deploy deep learning chatbots in customer service?
To implement a deep learning chatbot in customer service requires a steady supply of pertinent and current data.
Luckily you have several options to effectively integrate automation into your customer service strategy. To guide you, we’ve simplified your options into a Step 1 and Step 2.
Step 1: Choose the chatbot software you would like to work with.
- Hire a team of programmers and choose the ideal chatbot from of the range of chatbot software on offer, OR
- You can choose the GUURU SmartBot, which also comes with a peer-to’s-peer live chat function.
GUURU provides automation out of the box. There are no hidden maintenance or tagging costs.
- GUURU offers customized models that perfectly fit your business, providing your customers with the best possible experience.
- Should the GUURU SmartBot fail (as every bot does at some point), the enquiry is seamlessly sent to the customer service Guurus, who act as a fallback.
- No programming or technology knowledge is required with this out-of-the-box solution.
Step 2: Provide your deep learning chatbot with a wealth of good data to constantly tap into.
Your chatbot should have continuous access to transcripts of agent and customer interactions. This enables your chatbot to acquire an understanding of how your customers ask questions and what responses they receive from your customer service agents.
When you add or change product or service offerings, the enquiries from customers change, as do your agents’ answers. That is why you need to provide an ongoing supply of relevant data for your chatbot to draw upon.
With GUURU, the chatbot continuously receives data from the Guuru’s live chats and automatically evolves.
Your chatbot needs to be able to learn the latest vocabulary, and its significance to remain relevant for your customer’s enquiries. Deep learning chatbots demand a continued supply of up-to-the-minute data.
Deploying a chatbot requires time, energy and huge amounts of data. Although many businesses implement chatbots to reduce costs, the reality is that it can often be extremely expensive. And your chatbot can ultimately prove to be ineffective without specialized data from your business.
An interview with Sascha Kappeler, Head of eCommerce at BRACK.ch
BRACK.CH, a leading Swiss online retailer, leverages GUURU’s customer interaction solution to meet their client’s demand for always-on customer support.
We asked Sascha Kappeler, the Head of eCommerce at BRACK.ch, a few questions on his experience with the implementation of GUURU’s deep-learning chatbot into BRACK.CH’s customer service department.
You have actively integrated your customer community into your customer support. How is that working out for you?
It’s true. Our best product users, so-called Guurus, advise customers via live chat as part of our customer service. Our goal was to offer live chat support on all our pages at BRACK.CH, 24/7 and at a reasonable cost.
We have managed this and have been able to thereby improve our KPIs enormously: both customer satisfaction and the sales conversion have improved. Waiting times are low, at under 40 seconds.
We compensate the Guurus for each successful chat response they provide to customers. The solution is easily scalable. The more enquiries received, the more Guurus there are who receive a push notification and vice versa.
This means that we have no idle periods to pay for.
You have now activated the GUURU SmartBot. What led you to that?
We can view chat contents on the partner portal. Over the course of time, we determined that certain questions came up repeatedly, such as questions regarding delivery times, whether we deliver outside the country, where the partner code must be entered, how to reset a password, or customers requesting a copy of the invoice.
We sought a way to automate so as to avoid paying to answer these repetitive questions each time. The GUURU SmartBot fits perfectly as it is part of GUURU’s customer interaction solution.
How did the implementation of the SmartBot go?
GUURU primarily completed the implementation of the automated solution. Based on the existing live chat conversations conducted by the Guurus, which can viewed any time in the partner portal, we can easily recognize and filter out the repetitive questions.
We have formulated and categorized answers to these questions, and GUURU added these questions/answers to the SmartBot’s knowledge base.
Is the chatbot intended to replace the community?
No, that is not our goal. I believe we are still far from being able to replace personal support with chatbot technology. Especially when it comes to questions about products, we should provide personal, authentic consultation.
We have been able to excellently cover this need with our Guurus. The SmartBot shall handle simple, repetitive questions. This enables us to optimize costs without negatively impacting customer satisfaction.