Customer Story VodafoneZiggo

Customer Story: VodafoneZiggo

VodafoneZiggo has successfully innovated their customer service, creating fully human conversations with customers, enjoying motivated agents and managing lower costs in operations. Learn how they did it.

In short

VodafoneZiggo switched its approach from automating some conversations fully to all conversations partly, using Deepdesk AI technology
Deepdesk is successfully deployed with hundreds of agents of VodafoneZiggo since 2018. It currently automates 20% of digital conversations
With human-centered AI, VodafoneZiggo empowered the human agent and fast and fully human conversations with customers
Vodafone saved 15% in costs on human agents

The Customer

VodafoneZiggo is a Dutch company offering fixed, mobile and integrated communication and entertainment services to consumers and businesses. VodafoneZiggo is a joint venture of Liberty Global, the biggest international TV and broadband internet company, and Vodafone Group, one of the world’s biggest telecommunication companies.

The challenge

Customer acquisition costs in telecommunications are high. So customer retention and upselling is incredibly important. A great customer experience is key.

This is especially important when customers are facing complicated technical problems. In telecommunications, there are plenty of those: customers are dealing with failing WiFi, a new SIM card that doesn't seem to work or issues while switching from provider. (Learn more about our service in telecommunications here)

Great customer care is expensive, though. No wonder that many companies had high hopes for automation and robotic process automation. VodafoneZiggo also has invested in this area, deploying chatbots in repetitive tasks.

But a lot of conversations in the telecommunications industry are incredibly hard to automate without alienating customers. Chatbots can do their (limited) part, but almost 80% of conversations falls back to human agents.

1.6M conversations

chat and messaging

600 Agents

now all working from home

12M messages

for B2C on Fixed & Mobile

Still, parts of conversations that are non automatable with chatbots, are still highly repetitive and agents lose a lot of time typing the same sentences over and over again.

The challenge for VodafoneZiggo in cooperation with Deepdesk was threefold:

to find ways to make human agents faster; to bring down the average handling time of conversations.
to automate repetitive tasks that cannot be handled by chatbots
to guarantee a great customer experience while doing so

The Strategy

In a joint effort with Deepdesk, VodafoneZiggo did something innovative and counterintuitive. They successfully switched their approach from automating some conversations fully to all conversations partly.

The strategy is not more automation or less automation. The strategy is smarter automation: augmenting customer experience by improving the quality and speed of conversations, using AI.

Instead of having a chatbot in charge of communicating with the customer, with the help from Deepdesk, VodafoneZiggo deployed AI to support their human agents. This makes the conversation with the client faster, more reliable, more accurate and, last but not least, more personal, than either man or machine could accomplish by themselves.

Deepdesk successfully enables our strategy in which we’ve switched our approach from automating some conversations fully (with chatbots) to all conversations partly. It’s giving an enormous boost to our customer service.

Robin Clements - COO Vodafone Nethelands

The Solution

Virtual assistant

So, how does this work? Deepdesk built a virtual personal assistant for agents. They give the agents possible replies, useful links or tutorials or anything that helps the customer to get the answer he needs. Noteworthy: the Virtual Assistant doesn’t give the answer, it suggests the answer. The agent selects the phrase, video or link he believes is appropriate. This speeds up the tempo of the conversation and the human agent is still in control.

Smart Customer Service

Text and solution suggestions are tailored to live conversations and are personalised with agent chat history and personal tone-of-voice.

Even without writing, there are instant reply suggestions. These reply suggestions, are the top-3 suggestions out of then-thousands available that match the conversation and the agents tone of voice. When typing, sentences are autocompleted, reducing typing time.

With a single click a specific link to the website can be added, that matches the customers' need. And that's just some examples of a list of features that are available to the agent, when needed, to help the customer swiftly.

Flipping roles: autoflow

When everything goes smooth the agent can decide to let the virtual assistant do the job, while at the same time keeping an eye on how the conversation is developing. The roles have flipped now: the agent is helping the Virtual Assistant to talk to the customer, standby to come to its aid anytime.

The Results

Concurrency

All the above features together enable the agent to do something that is widely considered to be the holy grail of cost reduction in chat conversations: having multiple conversations at the same time. The Virtual Assistant keeps track of the different conversations, suggests answers and will do some of the conversations itself. The agent can now pay attention to the tricky part of the different conversations and, more important: focus on the customer.

Proving value

We have made efforts to carefully calculate the impact of Deepdesk AI on agents' performance. There are many factors that contribute to agents' handling time, that this is not an easy thing to do. We were able to prove that the use of Deepdesk had a great impact on AHT. For VodafoneZiggo we managed to bring down AHT with at least 15%. Read more about how we proved our value

Agents happier with less redundancy

In the end the agent is crucial in making this success. Agents happily adopt their Assistant with a 7/10 rating in surveys. What they especially like is that they don’t have to keep typing the same thing all the time.

We have analysed over two million chats and e-mails and searched for logic in all questions and answers. Deepdesk supports our employees to help our customers faster and in turn make the technology equally more sophisticated.

George de Visser - Director Customer Care Ziggo

15% cost reduction

on human agents by bringing down average handling time

up to 50% automated

Top agents now only write half manually, the other half is assisted by Deepdesk.

8% Messenger traffic handled by Assisted Intake

Increased self service with no human interaction needed.

over 200 million characters automated

8% messenger conversations fully automated with the Assisted Intake

1 billion characters analyzed

The amount of text that human agents used to write each year.

7 AI models & 10 recommenders

Every two weeks all models are updated. Processing GB of chat data.

Innovating together

In close cooperation with VodafoneZiggo, we have continued to developed new features and a fresh approach to AI in customer service.

Even more personal with AI

On request of the agents, our algorithms now know the preferences of the agent in question and will give suggestions that honour the idiom of the agent, thus making the conversation even more personalized, exactly the opposite of what chatbots accomplish. We also developed a feature for agents to manage their own personal collection of frequently used messages. Allowing agents to use their favourite messages with the ease of Deepdesk.

Knowledge base integration

Custom VodafoneZiggo knowledge bases are now connected to help identify the customers' problem faster. And soon agents also have the customers Next Best Action available to directly share with in the conversation.

State of the AI technology: Transformers

Ziggo is the first customer in the Netherlands to run a Dutch GPT-2 model in production. The technology developed by OpenAI was used to train the first Dutch GPT-2 base model. This Natural Language Generation model allows us to still recommend text in complex sentence structures.

We continue working with VodafoneZiggo, if you are interested we can join up as well. 🙂