Last week I visited the Web Summit in Lisbon. A truly great event to meet new people and companies, get inspired by the world’s leaders in tech. One of the talks that I really like was the talk by Mark Curtis from Fjord, Accenture’s service design agency. Mark talked about the past, present and future of conversation(s). If you’re interested you can watch the video here.
(Chat)bots don’t have to act human
There are two perspectives in his talk that I want to talk about. The first one is that bots do not have to act human to be human centered. I could not agree more. Being human centered, is all about helping humans get their jobs done. And this is what bots should be all about: helping customers do what they need to do, with the least possible friction. And for bots to do that, they don’t need to act like they are human, they just need to be helpful in a meaningful way.
Conversation as market
Secondly, Mark claims that conversations are now not only the interface, they are also the market(place). He referred to the likes of WeChat being a full conversational marketplace. As much as that is true, I believe there is also another perspective here.
Conversations are interactions that enable a lot more value creation than just transactions. Conversations are the primary means by which knowledge and other competences are being shared and transferred with the purpose of getting (functional, social and emotional) jobs done. In this sense, conversations have always been the primary market(place) throughout human history. This is why I’d like to state that conversation is the primary interface to value.
From philosophy to practice
As much as I like to think about conversations in a philosophical way, I’m starting this conversation also to share our experience with the conversational user interface.
Much like AI, it is one of the biggest hypes in customer experience today. Conversational interfaces are sometimes considered to be the next big thing after sliced bread. We have been holding quiet about this for some time. Not because we were not interested, but because we wanted to experience first hand what works and and what does not.
We have been working with many clients on this over the past one to two years. I would like to share three stories and will summarise some key learnings, or patterns that we see, after that.
T-mobile Austria, conversational customer service chatbot Tinka
Our first client who chose to develop a smart customer service chatbot with a conversational interface was T-mobile Austria. Her name is Tinka and she was launched to the public with a PR-campaign that we had not see before when it comes to launching a self service solution. To us, this is proof that companies are actually making the shift toward customer experience as a differentiator. A shift that we believe is as important as it is inevitable. But I digress.
Tinka, before she was transformed into the conversational chatbot, was already an intelligent assistant or virtual agent. She used natural language understanding and processing techniques to provide answers to T-mobile Austria customers online. But it wasn’t conversational, it was Q&A, period. After her transformation to a conversational bot she has a lot more to offer. Some of the key features are:
- The conversational design (of course)
- Embedded rich media like video’s, pictures etc
- Seamless transfer to live, embedded in the interface
- Flexible understanding via spell checking (algorithm based)
- Memory: remembers chat history, across multiple sessions
- Integration of multiple knowledge sources (knowledge base, community platform)
And of course this is all nice, but what really matters is the results T-mobile has seen since. For one they have seen that the NPS of Tinka has improved from -36 to +16 in the first 6 months. As adoption grew btw, that normalised to an NPS of around 0. This is an incredible increase in customer appreciation, which is translated also in increasing usage of Tinka as a digital self service assistant.
OHRA, conversational conversion optimisation
With these promising insights we had several discussions with other clients. OHRA, an insurance direct writer, was the first who asked us to develop a conversational chatbot to improve conversion rates on their website. Together with them we decided to develop a minimal viable product (MVP) to test our hypothesis that a conversational assistant, when customers where seeking information on a pet-insurance, would improve conversion rates.
Long story short: it did. And not by a small margin. Over the course of the first 6 weeks, when OHRA also ran a big TV-campaign, conversion rates were 100% higher when the chatbot was offered compared to offering the standard website with a basic form. After the campaign we noticed the difference dropped to 35% higher conversion rates, which is still an incredible step.
Robeco, conversational advice
Robeco, a Dutch financial service company, developed a conversational advisor on their website to help customers decide what kind of product would better fit their needs. Also here, in an A/B test, it proved to be a very successful approach. Completion rates of the conversational advisor compared to the old form/checklist was 70% higher. This does not only help Robeco sell more products with a better fit, it also helps Robeco to comply with increasing demands from regulators. Which is important when it comes to proving that they have actually reached customers, not just send them stuff to inform them, about the risks involved with certain financial products.
Conversation drives real results
As you can see from the above three stories, conversational interfaces drive real results in multiple ways. And this is why we are further investing in our platform’s capability to design, manage and continuously optimise automated conversations between companies and their prospects or customers.
There are also some learnings that you should be aware of, before you start yourself:
1. Wait for the drop
We see a clear pattern that after the implementation of a conversational user interface results go up exponentially. And of course, people involved get excited and sometimes jump to conclusions. I would urge you to be patient and wait until you have rolled-out the interface to a significant part of your customer/user base. Because we see that, as soon as adoption is significant across all customer/user segments (including e.g. age-groups) that results “normalise” on a lower level. That “lower” level is still much better than the pre-conversational interface results as you can see in the stories above.
2. Limit open dialog
Some of our clients like the idea of having full automated, conversational and open dialogs with their customers. This means that they ask us to not use pre-defined option buttons in the chatbot conversations. As much as we feel for the idea, we advice differently. We strongly believe that customers do not desire to have conversations like this with a company. They like to get their jobs done and to go on with their personal lives. Pre-defined options make their journey faster and easier, because they don’t have to think and type (or talk). This is just one of those things that reduce friction for the customer. And that is what they want.
And of course, you do always need to have the “open”-option available. Just as much as you need to have the “escalation to live”-option available at all times. Preferably with a seamless transfer of the prior conversation to the live-agent who takes over the conversation.
3. Deploy omni channel
As tempting as it is to quickly design and deliver a conversation chatbot on Facebook Messenger, or just your website, we believe you should deploy the conversations that you have designed at all possible touch points or channels. Customer needs are often triggered by context. A context that, despite the vast data you have, is often not revealed to you. It would be missed opportunity if your chatbot cannot deal with product-advice related questions on the customer service webpages, in your service app, or on Facebook messenger, because you decided to deploy it on the product webpage only.
If we take all things into consideration I believe it is safe to say that the conversational UI is a great interface to value. I hope sharing our experience helps you capture it.