Hey, robot, let’s cut the cackle!

Konstantin Sadekov
March 17th, 2021

Mailiis Ploomann, Elisa Eesti

This somewhat rude sentence was said by an Elisa client to service robot Annika, when the robot was unable to understand the client’s issue with their Homepod not operating correctly. Since Elisa does not sell or service Homepods, this is hardly in any way Annika’s fault. However, the statement is still noteworthy, writes Mailiis Poolmann, Head of Telecom Services at Elisa Estonia.

What this statement means is that Estonian automated service has entered a phase where communication with a robot is something that the clients have become accustomed to. At that, communication with Annika can be equalled to communication with a human operator. The expectations are the same.  This can only be viewed positively and in order to celebrate the Estonian Mother Language Day (March 14), it might be a good idea to summarise what can be done with the help of Estonian language technology.

Chatbot Annika started work in spring 2018. At first, only carefully and in a small scope, but constantly learning and taking on more and more tasks that became increasingly more intelligent.

In 2020, through chat on the Elisa web page, Annika independently serviced 18,000 contacts on 70 different subjects – finishing all these contacts successfully. At that, it is important to stress that Annika is not a simple question-answer “solution” that would simply send the contact a link for finding a solution on their own. Annika’s thought patterns highly resemble the complicated ones of a human operator. This enables to make navigating in various information systems easy and logical for the client. To prove the latter, Annika can show its transactional NPS (net promoter score), which is normally between 30 and 50. In comparison, human operators tend to average between 40 and 60, whereas an NPS over 30 is considered a good result. Therefore, Annika’s result is completely satisfactory. Today, we direct nearly a fifth of all incoming contacts to chat as a service channel. The models affirm that Annika is able to recognise over 70% of them. Meaning, that the writing Annika’s also shows potential for growth.

More than half a year of working time saved

In 2019, Annika started to operate in the call centre, answering the client’s calls, listening to the issues people turn to Elisa with and directing the client to the human operator most competent to find the best solution. During 2020, Annika serviced over 300,000 clients in this manner. Considering that Annika’s work saves about a minute of (working) time for both the client and the operator, this means 5,000 hours of saved time. That’s more than half a year of net time that could have otherwise been spent by both parties pushing various buttons in IT systems.

But Annika’s goal in the call centre is not just saving time, but also to match the caller with the human operator whose competences are best in solving the client’s problem. Elisa offers very many different services and technologies. In general, clients turn to us for more than 800 reasons. It is unthinkable, if not even inhumane, to assume that one person would be able to help in all of these equally well. Often, clients are annoyed, when the operator is not informed about the topic. This causes stress to the operator as well. However, Annika is able to assure that such situations wouldn’t happen and that all humans – both clients and operators – would be treated well. Annika is excellent in this job.

Our initial concern, that certain clients – or a certain generation of clients – are not ready to talk to a robot, has been proven futile. Of course, there are also clients who don’t like Annika. But after ten years in the field, I am still to meet the human operator that every client likes. One of the ways to measure this is the abandoned call rate, which has lowered since we have started using the robot. An abandoned call rate shows how many clients end the call when they hear that the call has been answered by a robot. At that, we must also consider that there are also clients who just change their mind about calling or have misdialled.

During Annika’s first “year of life”, there was some confusion created by the fact that clients had not met an intelligent Estonian-speaking robot before, and thought it necessary to use codes or keywords instead of speaking with full sentences. Today, most call transcriptions include grammatical and fluent sentences. In all fairness, even a human would find it easier to understand what is meant by “I need help with changing my new Wi-Fi rooter password” than just by hearing “Wi-Fi” or “Internet rooter”. Considering the fun fact that by the second half of 2019, clients had taught Annika to curse (in many languages), we can truly claim that people have gotten used to Annika the callbot.

Annika helps to detect faults

At the beginning of 2020, an Estonian-speaking callbot was put into its third highly valuable usage. Annika started to detect faults and disorders and send the information on that in real time to our monitoring team. All entrepreneurs want to offer services that are constantly faultlessly operational and never fail, but in IT and networking such ideal worlds do not exist. Servicing over 600,000 clients, who own thousands of various appliances and are located all over Estonia but also worldwide (including underground or above ground), something is always liable to break down. Luckily, not all at the same time, but something is always happening. Sometimes we are the faulty party, sometimes it’s the client, but it can also be a third party or even some “evil spirits” or “pixies” because there are no other plausible explanations. Independent of who is at fault, the client’s problem must be fixed. Sometimes this can be done by the operator, but sometimes a network or IT specialist must interfere. In order to find out who has to do what and where, it is extremely important to understand fast where the fault originated from. That’s where Annika comes to play.

Within seconds from a client calling and complaining that there are issues with Internet quality, before the call is even connected to an operator, Annika initiates two models: network anomaly detection and fault classifier. In such cases, a dot is added to a real-time map to monitor where the client’s issue is. Based on that, it is possible to calculate the likelihood that the issue is caused by the Internet provider. Annika will add information to the specialist that the client complained about the quality, and didn’t claim the Internet to be lost completely. This might seem a trivial addition, but such information travelling from all clients directly to the specialists enables to detect such faults within minutes. Pre-mapping such issues took hours of research and correspondence between operators and specialists. In 2020 Annika managed that more than 1000 times. That means 1000 cases of nobody flipping out and time saves both for Elisa and its clients. Additionally, Annika is a great bridge between monitoring and customer services by connecting people who tended to take opposing roles and feel a certain animosity towards each other.

Annika searches systems to find answers for human operators

In 2021, when Annika’s role as a chatbot and as a callbot had reached the phase of stable scaling, another – the fourth – feature was added. Now Annika started to help front-line staff. Until now human operators had been involved in gathering training data, but now Annika has been fixed in place and has an obligation to help the operators in return. Elisa’s new internal tool for front-line staff has been built up in a way that enables AI to guide the operator in situations when the client turns to the human operator. The symbolically named KITT – which is a client information desktop for the operator – searches various systems and registers, among other things, also the reason for the contact. 

Yes, robots can take away jobs from people, but that doesn’t mean that people will not have jobs to do. The only jobs that are taken are those that were robotic to begin with. Today in Elisa, those robotic tasks are overtaken by Annika. Most importantly, we’ve seen how this turns the jobs that people will continue to do much more humane. Now people are engaging with tasks that require emotional intelligence, creativity and critical thinking. These are the kinds of tasks that people actually like doing, these offer challenges and learning opportunities.

Annika has another stunning ability that has perhaps not been talked about enough. The robot overlooks certain obstacles that often obstruct work in medium-large companies. Since Annika is client-based and her task is to make sure the client’s problem is solved, the robot is not concerned in the least about whose job description purports which task. In order to create value for the client, everyone must come together: business, technology, product owners, service, e-channels and call specialists.

Within the past three years, we have repeatedly rejoiced how much synergy is created by a robot that directs our work. Annika is able to connect people from various units of the company and make them work together for the same goal. This is the best possible application of innovation. We have not spent time to come up with something no one has ever thought of, but we apply technology to provide value for the client in a smarter way than ever before. This was not created by a brand new start-up in a fancy area of town, but a 30-year-old company of 1000 employees.

There’s one thing that is worth stressing: Annika is not a little bit of code that has been gracefully given to the Estonian branch of a big Scandinavian company. Annika is a unique fruit of cooperation that has been created through the synergy of two Estonian companies. Elisa’s partner in AI, MindTitan, has come up with all the models, whereas Elisa’s employees have undertaken the practical application of these. This has all taken place using the Estonian language, which was possible thanks to our language scholars’ world-class work at the University of Tartu as well as at the Institute of the Estonian Language. Our Scandinavian colleagues have expressed their high levels of jealousy and are working to follow our footsteps.

With that, we wish all Estonians a happy Mother Language Day. We promise that Annika will continue to care for the practical application of Estonian language technology and hope to be able to report soon on the robot’s next achievements. We are proud to lead our neighbours in innovation, wherever it would take us.

Written by the Head of Telecom Services at Elisa Estonia, Mailiis Ploomann

ai plan execution