Why Customer Experience Matters More Than You Think And How Virtual Assistants Can Help You Beat Your Competitors

Virtual assistants and chatbots came to the wider audiences in the spring of 2016 when Facebook announced the possibility to build bots into Messenger. After initial inflated expectations the chatbot fever somewhat declined. However, much like in the Gartner Hype Cycle, the importance of virtual assistants is bound to grow in importance. And the reason is that superior customer experiences enable companies to retain and upsell to existing customers.

This follows the rollout of Aggregation Theory impact on businesses. “First, the Internet has made distribution (of digital goods) free, neutralizing the advantage that pre-Internet distributors leveraged to integrate with suppliers. Secondly, the Internet has made transaction costs zero, making it viable for a distributor to integrate forward with end users/consumers at scale”.

Such impact can be clearly seen in the realm of digital businesses – news media, movie rentals, social media. However, as many traditional services are becoming commodities, customers have much easier tasks in switching their suppliers. Electricity in Northern Europe is a good example – you can switch your electricity service provider through the internet. Electricity as a product is a commodity. Switching to another mobile operator can also be done through the internet. Mobile network access is a commodity. Switching from one bank to another is increasingly easier with all the fintech startups out there. Access to financial products is becoming a commodity.

In the light of this one thing is clear – customer experience matters. And there are three distinct opportunities where virtual assistants can currently help companies:

  1. Make companies available 24/7 for their customers. Customers want answers Friday night 11pm and Sunday morning 7am. If you can serve them answers with virtual assistants during those times, it is much better than finding out that customers have to wait until Monday morning 9am to call customer service. Waiting kills experiences and drives customers away.
  2. Convert conversations to sales. Analyzing conversations with customers and providing guidelines for virtual assistants, they are able to upsell your products and services at the right time to the right people. As natural language algorithms detect intents, you can use follow-up dialogue in intents to trigger sales pitches and run this across thousands of conversations. This way conversations can be turned into increases in sales automatically.
  3. Human work gets more optimized. As the virtual assistant analyzes conversations and knows the skills and performance of human customer service agents, it can prioritize incoming requests from customers and route the most important ones (e.g. customer wants to leave the service or is interested in buying more products from the business) to the customer service representatives that can help those customers the best.

Customers have more choice than ever in selecting their service providers. As such competition will be more and more in the realm of providing superior customer experience. Virtual assistants are one way of standing out in the crowd and beating your competition.

Here’s a simple thing you can do. I advise to visit the website of your competitors. See how well they can be reached and how quickly they respond. Via phone in 3 minutes? Or through chat in 1 second? Think what you would want as a user.

Top 2 Questions From Customers Answered – What Topics Should My Bot Know and How Much Training Data Do I Need?

The content and structure will largely determine the value for your virtual customer assistant. It stems from the use case that you give to your bot. Hence, before starting to build the bot it is important to really understand what you need the bot to do and how.

Once, that is determined, then you need to figure out the content that goes into the bot. Some important points to keep in mind about the content for the bot that we get asked quite a lot.

We advise to select around 50-60 (max 100) topics for your bot. These topics should reflect the most commonly asked things from your customer service team and as such you should focus on them. Of course you can select more topics but the risk in that case is that the topics will be rather similar in most cases and bot will start referring questions to topics that is not intended. Also it helps to keep in mind the bots are all about volume and frequency – the 80/20 rule applies. 80% of automatable chats in customer service come from ~20% of topics.

For the size of the training data – if you have around 2000 conversations then that is good enough to start. By conversation I mean chat session with a customer. For most part the important aspect from the conversation is the first one or two questions that the customer asks. This gives you information on how customers ask questions which, in turn, is valuable information to your bot as you can train the bot to understand the different ways how customers ask questions.

Videos, images, GIFs and emojis all make the content more enjoyably. Especially if you are targeting a younger demographic. Screenshots with usability information (about apps, instructions for the self-service portal) are helpful in many cases and can help users better than instructional text.

Keep the answers to the point. Usually 2-3 sentences is enough. Use language that is clear, much like if you were to explain things to your friend who does not know much about your company. Also it is good practice to include links for further reference in the answers so that people can find more details if they wish.

Have additional questions about training your bot? Shoot me an email indrek (at) alphablues (dot) com, I’ll be happy to help.

How virtual assistants understand language

“My language is tough and hard to understand for AI”. This what I hear often when travelling to customers in Europe. What people are worried about in non-English speaking countries is that their language is more complex for the AI to master than English.

This is totally understandable. Most AI companies focus on English as it is one of the most widely spoken languages in the world. However, when I talk to people from Hungary, Latvia and Poland, their concern is widely the same. Language is complex and as of today no AI system in the world is fully capable of understanding meaning from language and when being spoken to. This is a blessing as it gives companies like ours the ability to build a valuable solution. It is also a curse, as customers of course would like to have an all-comprehensible AI that works flawlessly and nobody as of today has built such a thing 🙂

Before talking about how bots understand language is it important to differentiate between understanding meaning and speech-to-text / text-to-speech. Understanding meaning means that when a user is telling something to the bot (e.g. “Hey, what happened to my last order?”) the bot understands that the user wants to know about a past transaction, there has been a problem with it and that it needs to be solved. This is different from speech-to-text which takes the sound of the sentence and converts voice to a string of text e.g. “Hey, what happened to my last order?”. Speech-to-text does not aim to understand meaning behind that string of text. Similarly text-to-speech takes the string of text “Hey, what happened to my last order?” and converts it into voice. Speech-to-text and text-to-speech work with remarkable accuracy (beyond 90% in many languages) whereas any human-accepted answers beyond 20% of understanding meaning from free text queries is considered good.

There are essentially 2 levels in the way bots understand free text inquiries and natural language.

One, they look at the token level. What the bot is doing is it sees an incoming sentence from a user (“Hey, what happened to my last order?”) and it is then going to look at all the data it has ever seen and trying to match this sentence to that data to determine what the sentence might mean. If the bot has been data about orders before then it will most likely match the sentence to a topic related to orders. There are a variety of algorithms used to assess the match between incoming sentences and the knowledgebase and they work differently. Some look at the occurrence of words in the sentence, some at the combination of words in the sentence, some a the occurrence of letters in words and across the sentence. All in all, the bot performs pattern matching to relate the sentence to a meaning. Most AI-based solutions on the market today with custom NLP work this way.

The second approach is to use semantics. This means that the bot has knowledge beyond pattern matching. Typically, a computer processes free text as a sequence of symbols with no apparent relationships apart from the order in which they appear in a sentence. A human, however, understands the semantics. For example, a person knows that in the sentence “my older brother rides the bike,” the brother is a human being, the bike is an inanimate object, the bike cannot ride the brother and so forth. Something we have done at AlphaBlues is feeding virtual assistants with “semantic enrichment inputs” so that the chatbot better understand the messages it receives. Powered by deep learning, the virtual assistant will continue to learn as it receives more messages and applies the semantic knowledge it has been provided with.

The combination of a token based and semantic approach gives a neat approach for increasing the probability of capturing meaning from language. With such a dedicated system it is possible to successfully understand languages like Polish, Estonian and Hungarian.

Have you run into trouble with language when building your virtual assistant? Shoot me an email indrek (at) alphablues (dot) com, I’d be curious to know about the challenges you faced.

Tallink launches its virtual assistant Nemo

Tallink (Nasdaq OMX Nordic: TAL: TAL1T) has launched their virtual customer assistant, called Nemo, on their website www.tallink.ee Nemo assists Tallink customers with questions related to booking trips on Tallink passenger ships and general inquiries about the services. Nemo was built by AlphaBlues and is utilizing the Natural Language Processing of the AlphaBlues platform.

The aim of Nemo is to help customers get immediate information to their questions, allowing customer support agents to focus on more complex issues. According to Martin Mürk, head of Customer Experience and Analytics “virtual customers assistants serve an increasing amount of customers as the first-point of contact in the field of tourism and travel”.

Tallink is one of the leading passenger shipping companies in Europe, serving more than 9 million customers per year. More information can be found at this press release link here.

New feature release – user authentication in virtual assistants

One of the recent product features we’re proud of at AlphaBlues is the ability to authenticate users and provide detailed answers for the users. It is a new and exciting feature that allows our customers to offer a personalized level of support to their users through virtual assistants much like human customer support agents can.

In the previous post we highlighted how the lifecycle of virtual assistants is evolving throughout time. Those companies that have high chat conversations volume on a monthly basis and innovation capacity are seeking more and more for virtual assistants to actually conduct activities out on behalf of the users.  

Personalized and authenticated virtual assistants are on the rise. How do they work?

In a traditional chat automation environment, you have a virtual assistant (VA) that is having a conversation with a human. The user for the VA is anonymous, meaning that the VA does not know the name or identity of the user. This is the case with website based virtual assistants but also with Facebook Messenger VAs. Even if the VA knows the first and last name of the user from Facebook, it does not mean authentication because ultimately it is the company (i.e. the service provider) on whose behalf the VA is having the conversation with users that has to identify the user against its own database.

How does authentication work? In our case the way it works is that when the user asks a query that requires a transactional element to be executed (e.g. “What is the balance of my invoice?”), our Natural Language Processing chat server determines that this query requires authentication. As a next step the user is directed to the company’s (e.g. a telecom company) authentication system where the user will log in with the regular authentication method given by the service provider (e.g. ID card, mobile ID, username/password). After that the AlphaBlues server gets an access token from the company, which is stored with the user session. This way a link has been made between the user and the company.

When the user is asking for their invoice balance, then the chat server is sending that request directly to the company’s API where, the company verifies the validity of the access token. Once the token is validated, the transaction is carried out on the company’s side. After that the information regarding the balance is passed to the chat server via HTTPS and then displayed as confirmation to the user in the chat interface (e.g. “Your balance is €38.24”).

Such a system enables to use the full functionality of the company’s API in allowing the user to execute a wide variety of commands without the need for being in touch with human customer support agents. What’s best is that the user in this case has a simple interface with which it can access its information and conduct transactions without bothering to log into self service environments on websites or call human support.

It provides for a quick and easy way for accessing data and conducting transactions. This simple technical overview is a quick glance of how the authentication works. If you have interest in authenticated solutions, do get in touch at indrek (at) alphablues (dot) com

Chat automation projects’ lifecycle

 

Over time I’ve noticed a trajectory for companies in adopting virtual assistants. Companies start from any of the starting points and then slowly move up the trajectory as their chat automation project progresses. Needless to say, all the successful chat projects differ in speed but on average, the course is the same.

 

Understanding this trajectory helps you assess your own organization, where you currently stand and how to think about chat automation in the larger context of automation. Take this as an automation roadmap that you can use internally in strategy meetings or when planning out your activities and budgets for the next quarters. The trajectory goes as follows:

 

  1. First, companies assess the needs of customer support in terms of volumes of monthly transactions. This is for deciding whether to go for a bot in the first place. We usually advise to use bots if the monthly chats exceed 1,000 conversations. Less than that and you can survive with a human support agent.

 

  1. Secondly, when companies go for chat automation they do so for certain parts of the customer service channel. The first look is mostly at high volume parts of the customer support. These can be related to certain products or services (e.g. topics related to invoices or billing usually gain the most volume in customer support).

 

  1. The third step is to offer the support bots across all customer service related topics and perhaps bring in some sales related topics as well. This is done after a pilot when the initial parts of the bot have been tested. Usually there you see that the customer support managers and people leading business development have accepted that the bot can add acceptable value to the organization in the initial pilot phase and the decide for a more comprehensive roll-out of the support solution.

 

  1. And number four is where things get interesting. What some of the more innovative companies are thinking about are virtual assistants capable of authenticating users. Why is this important? When in the beginning the bots are mostly just giving general advice on topics, then authentication unlocks a lot of powerful features. With authentication, the bot can verify the user’s identity against an internal API and start providing information that previously only a human agent could provide. This can include detailed information about billing, status of ordered services or products, ability to change details related to the account and orders. Essentially anything, that human agent could do for the user or that the user could do in self-service. Yet, with a bot through the chat interface, such queries can be executed faster and in a more convenient manner. Add voice to this and you got a killer combination.

 

Needless to say, authentication is something we’re offering as well and more on this in the next post. In the meantime, feel free to share this trajectory in-house with your manager to get their buy-in for the automation project you are seeking to push forward 🙂 If you need help, reach out indrek (at) alphablues (dot) com

How To Align Your In-house Team For Executing A Successful Chatbot Project

Like with any project, good planning is the key to success also with your bot project. Having seen many chatbot and virtual customer assistant projects over the past years, there are couple of lessons learned and best practices to keep in mind. The key feature here is how to align your team inside the company so that the project gets delivered in time. The task becomes especially tricky when your company has thousands of employees and you are coordinating tasks across several departments and teams.

Here are the things to help you out.

1️⃣ Pick a leader. Choose the person who is ultimately responsible for the project. Essentially that person is like the “mother” or “father” for the bot. It is his/her task to make sure the project gets delivered. That person is also the one making final decisions on bot related developments.

2️⃣  Choose your success metric. Make sure the team is aligned on what is the ultimate goal of the chatbot you wish to deploy. It can be the number of deflected customer service tickets, increase in response time, saved time for human employees, increased product sales conversion etc. Whatever it is, make sure your whole team knows that is your North Star.

3️⃣ Build in-house or outsource development. Large companies have usually in-house developers that could be capable of building chat automation systems. If you have such teams at your disposal then it is possible to get started there. However, in many cases such teams have plenty of things to do and usually it is easier to run first chat automation pilots with an external developer at smaller scale to see their capability and also measure how well chat automation can help your company.

4️⃣  Have your team be ready for the long term. Understand that your bot is not perfect on day 1. It takes a few weeks first to see how well the bot is performing. In next iterations which take another few weeks you can start measuring how people respond and what are the things you need to change. It can take a few months to get a good picture of how best to optimize your bot so get your team ready to be in it for a few months to really see the results.

What I recommend you to do is communicate this to your team beforehand. It will make the project less turbulent and you are more likely to maximize the value of your invested time and effort. And if you need help with these decisions, then reach out indrek (at) alphablues (dot) com and I can provide some quick insight and second opinion.

New customer – Enefit

Glad to announce that Enefit has launched their virtual sales assistant on the Finnish market. Enefit is Eesti Energia’s subsidiary in Finland. Eesti Energia is an energy group which consists of more than 20 companies and employs approximately 5 800 people. It is the largest producer of energy in the Baltics, incl. one of the biggest renewable energy producers. Enefit is currently in Latvia, Lithuania, Poland, Finland and Sweden.

Enefit recently entered the Finnish market and built a virtual sales assistant, called Sähköbotti (Energy Bot in Finnish). The bot was built on the AlphaBlues platform and integrated on the www.enefit.fi website. The bot helps new customers choose the best energy package right from the website.

Enefit chose the bot to differentiate from the competition and offer its customers new and easier ways of purchasing energy plans. As the energy markets are highly competitive, such differentiation helps Enefit stand out with its digital and user-friendly customer experiences.

#AI and #fintech Demo Days with Polish corporations

At the beginning of the year we joined the PwC Startup Collider program in Warsaw Poland as our first steps into that market. From our company point of view the past few weeks have been exceptionally exciting as we have been able to attend several corporate-startup matchmaking events in Poland and met with a large variety of companies to understand their needs when it comes to AI and automation.

In mid-March we attended The Heart Demo Day at the awesome Warsaw Spire building. The Heart is a leading corporate-startup matchmaking incubator in Eastern Europe and this demo day was focused on marketing and communications related startups. The level of the participants was high as most startups were more mature and some having raised already multi-million dollar funding rounds. Was great to see companies pitch their products and really emphasize their use cases with existing customers. In the B2B world, use cases and references are everything.

Indrek pitching at the demo day to Polish corporations.

In addition we attended at the end of Marh the ING Bank startup demo day at their Polish headquarters. ING has setup an Innovation Lab at their HQ and the event was focused on startups in the AI and fintech space. The lab is run by a great team of innovation professionals. Startups were again of high quality and it was a great example of how large corporations are really doing focused innovation work with technology players in the market.

With the ING Innovation Lab team that made the Demo Day possible.

 

5 Things Successful Companies Do When Setting Up Their Virtual Assistant

You’ve become interested in having a virtual assistant for your company but are unsure how to start. What should be done? How? What is important? What is less important?

These are all questions we get when we meet with companies. They wish to know what is the best practice in setting up a bot and what to keep in mind. Having deployed dozens of bots we have a learned a thing or two and we’ve highlighted some of the key lessons to keep in mind below.

Each of them deserves more attention than we can give in this short post and I’ll follow-up in the next posts with more insight into them.

That being said, here are the key things to decide and keep in mind when thinking of setting up a bot:

1️⃣ – Pick a use case for your bot. Is it customer service? Is it selling products online? Build a case what is that you want to achieve with the bot? Many want to offer a 24/7 channel for customers to get in touch or make their staff more efficient or understand what customers actually want to know about your company. All are good things to consider a bot. Pick yours.

2️⃣ – Where do you wish to put your bot? A good option is to put the bot on your website inside a chat window. Another is to implement in Facebook Messenger. Or if you have a mobile app, you can include it there as well. Consider where do you currently have customer traffic coming in and pick that channel.

3️⃣ – Pick your language. It is easiest to start to have the bot in one language. If your website is in English then train a bot in English. If your website is in Latvian and Russian then pick the language that gets used the most. When you start out with one language you can learn how users interact with your bot and you then take these lessons to your next iterations. If, however, two languages are crucial, then include both bots in the same channel through language detection.

4️⃣ – Automate the 20% of user questions that give you 80% of volume. Automation with chatbots is all about getting the most value out of volumes. Go for the low hanging fruits. The questions and issues that hundreds and thousands of your customers have. These are also the questions that take most of the time for your colleagues to answer. Or if you run a sales bot on your website, build it so to respond to the most common questions about a certain product or topic.

5️⃣ – Have bot-to-human handoff available. Bot cannot solve everything all the time. Humans are still needed in the process. Make sure your bot is capable of directing unanswered questions to humans for follow-up. Alternatively build your bot so that it is capable of giving people directions how to get in touch with human support.

There you have it. 5 things that you need to keep in mind when starting a bot project. Make sure your whole team understands these topics when you get started with your bot project and ask your tech provider about these topics.

Resolving them early on will make your project better and more likely to result in positive outcome. If you have questions, reach out 👉 indrek (at) alphablues (.) com and we’ll get back to you in 24 hours to offer our feedback.