How to create truly automated customer service?

Kristjan Jansons
April 12th, 2021

automated customer service


We will discuss how good customer service can be achieved by making it an integral part of the product with help from AI which enables us to create truly automated customer service.

Why is customer service getting worse?

“What a wonderful customer service experience!” is, unfortunately, a sentence that is not said too often. Research shows that experiences are getting worse. Rather counterintuitive in a world where we are used to things getting better together with technological progress.

Firstly, this is because contact centres are treated by organisations as cost centers which up to some point is a correct approach. Yes, customer service needs to be delivered cost-effectively, but if KPIs get optimized without sufficient sanity checks, you will see unwanted behaviour. One might optimize short-term KPIs without keeping an eye on the long-term ones. For example, when minimizing the number of contacts, you might end up hurting your brand NPS, if done wrong. Or when maximizing the deflection rate, you might leave a lot of the contacts actually unresolved. So what happens in the real world? In order to minimize contacts, contacting an organization might be made so difficult that people simply give up. To minimize call length, agents might hang up to meet their call length targets. To avoid solving costly contacts, organizations might keep forwarding the customer around until the customer gives up. The list of horrible examples goes on and even if such customer service might be profitable, not many can afford it. In the long run, only companies who are in a monopolistic or near-monopolistic position can afford bad customer services as their customers simply do not have any good alternatives.

Secondly, there is a lot of technological advancement, mostly in the form of chatbots, but when used incorrectly you will read stories like this that claim that bots are ruining customer service. Indeed, when FAQs are simply put inside a chatbot, the bot does not understand the customer or gives vague answers, there will be serious resistance from customers who will simply start looking for ways to get to a human. They will not trust the bot. This is what happens when people think about automating customer service without thinking if it really solves customer problems.

Thirdly, we should not forget that customers simply have increased expectations. Even if you are not a competitor of Google or Amazon, people still have the expectation that they will get an answer by just typing a question in a box and pressing enter. They expect to get their product with one click. Let’s be honest, there are many areas where the world has become more complicated and it is easier to ask instead of investigating it yourself.

Of course, this list is not conclusive. The good news is that almost all of these issues are curable with the correct tools and processes without the loss of profitability.

What is expected from customer service?

Of course, everybody wants the best, but as an organization you should ask the following. How large is a customer service part of your value creation? What do customers value and what are they willing to pay for? In a top-notch hotel, customer service might even be 80% of the value. In a bank where you get advice for a loan, it might be 30% because people still care more about the loan conditions. For a telco, people might just want the service to work and customer service might only amount to 10%. For a tech startup, even only 1%. The higher the number, the more you should think about how to make every interaction special, create a wow-effect and give your employees the freedom to act. The lower the number, the more people want things to happen automagically, and customer service is just something they need if things go wrong or something is confusing – they simply want a solution and an occasional apology. Of course, whatever the case, people expect customer service to respond as quickly as possible.

Let’s focus for a moment on quick responses, getting a solution and automagical experience. A quick response time is realistically only possible with automated or semi-automated customer service because hiring a lot more people would not be feasible. Getting a solution is solvable with training, using manuals if the product/service is rather simple and changes almost never – this is increasingly difficult in our modern complex world. Manuals tend to be somewhat vague (there are simply too many cases to cover to be specific), are often out of date, or even open to interpretation. People want expert solutions which truly solve their problems. This expert knowledge as a principle does not exist on the customer service side, it needs to come from the product side. For example, if your package delivery policy changes, it is the product manager’s job to deal with people who need to know about it, whereas customer service deals with the consequences of things not working out as expected. And finally, the automagical experience – people want things to work out with minimal effort from their side. Customer service might hear about the complaints and deal with them, but let’s be honest, the product side needs to understand where things are failing and fix it.

Should you care?

As stated above, giving bad customer service is sometimes simply profitable because you have to manage fewer contacts if people eventually give up contacting you. But in the long run, not many can afford customers who hate you. Dissatisfied customers tell an average of 7 people about their bad experiences. This may get heavily amplified by social media, Yelp and Google reviews. Through that, your brand’s NPS and other KPIs will get a hit in the long run. But is that the only thing that people think about?

Not really. We mentioned that customers expect solutions and an automagical experience. Delivering those is actually more about product management than customer service. In reality, nobody has a crystal ball to know what it means to be competitive in your market in the future, but you can be sure that it is crucial to be close to your customers, being able to understand what the impact of your decisions is, and knowing what is working and what is broken. Competitiveness is increasingly defined by having the ability to quickly adapt and make decisions – the vision needs to be executed by teams who are close to the action and who have the necessary power and tools. Product management deals with product development, processes, pricing, promoting, finances and much more which affects customers, but historically they do not have a clear feedback loop from customer service to be close to the action.

Let’s think about the alternative – your product management is out of the loop. At first, probably nothing happens and the status quo might remain the same for a long time. This does not mean that the status quo is maximizing your profitability. It also does not mean that there would not be a growth in the number of contacts – especially with product lines getting more complex over time. If product management is out of the loop or barely in the loop, it means that there is no clear ownership structure and product management doesn’t really know what the customers think or struggle with. Yes, product management might conduct surveys, interviews, collect data and they should keep on doing it, but this does not replace the direct and clear feedback from customers, especially from those who are rather annoyed.

In conclusion, you should care a lot if you want to improve your product/service and if you care about your customers’ opinion.

How do customer service and product management interact today?

As shown in figure 1, product management and customer service have an apparent brick wall between them. Product management deals with product-related activities such as product development, processes, promotion, pricing, supporting the frontline, handling finances and dealing with customer contacts when things go wrong. Customer service deals with handling the contacts according to their training, manuals and gut feeling. A lot depends on the interpretation of customer service agents who decide if customer contacts got resolved or not. Product management and customer service mostly come in contact on two occasions: when product management has created/updated a manual and customer service needs to be (re)trained based on it; or if something goes so wrong that product management needs to come in and save the day or deal with the consequences.

product management

This setup which is mostly based on manuals has some obvious issues. The first is human interpretation – the agent decides if the customer was satisfied. Unfortunately, this does not necessarily equal to the customer getting the answer they should have received. In addition, manuals by design are somewhat generic because these can’t cover all the edge cases. This means that these manuals are open to interpretation. Being open to interpretation is especially a problem with new agents. Also, manuals get updated, but agents might remember how things were done based on an outdated manual. Depending on the sector and the general situation, it might not be possible to keep a manual up to date – for example, when COVID-19 lockdowns started, many contact centres shared outdated information because it took too long to update manuals, and conduct relevant training.

The actual biggest drawback of this setup is the fact that there is no clear ownership. The contact ownership is shared between customer service and product management. If nobody truly owns it then nobody can truly fix it. Many parties (product manager, customer service manager, even an agent) might be held accountable, but all might sense that they do not really have the power to change anything significant.

How should customer service and product management interact?

So who has the power to change the situation and the process? An agent can suggest changes, but they do not have the power to change it for everybody. Customer service as a whole might eventually be able to figure out how to answer contacts, but they lack the broad and up-to-date product perspective for maintaining it. For example, how should they know if there was a change in the pricing strategy made yesterday and what the reasons behind it were? Customer service might analyze the chats or calls (even transcribe them) to understand where things go wrong, but this is fighting a lost battle because they are dealing with the consequences, not the root cause. This is why product managers need to step in. It makes sense, if they own the contact, take the responsibility and feel the pain. They have the power to change processes and solutions if they see fit. Instead of just observing customer service and supporting them, they have the opportunity to make customer service part of their product. 

What would this world look like? Would manuals disappear? Not exactly, manuals would still survive because there are plenty of situations that require expert intervention, non-standard approaches or the wonderful human touch. In every other case where a standard solution is possible, product management should create a step-by-step flow for how to solve those cases – a solution flow is what we call them. Creating solution flows is not a one-time feat. The flows need to be improved, analyzed, revised and updated regularly. Maintaining them is just as important, and possibly even more important, than creating them. This has returned as product managers can learn from them, find out what is broken and what is working, be close to their customers and understand their mistakes to improve the product and processes. All of this is visualized in figure 2 below – instead of a brick wall between the two parties, they form a symbiotic relationship. We call it product management 3.0. It’s product management where customer service is part of the product.

product management 3.0

Figure 2. Product management and customer service working together in a world where customer service is part of the product

How to actually create truly automated customer service?

All of this sounds great, but how to actually make all of this work. To recap, we want to provide quick responses, give a solution and create an automagical experience. Also, it is crucial to get product management involved. Quick and automagical sounds like a job for AI – say what you need and it will understand. Giving a solution sounds like a job for product management. With these two we can create a successful bot and truly automated customer service. The first part is to understand what the customer wants with AI, and the second part is providing a solution with solution flows. This scheme as brought out in figure 3 works successfully both with chats and calls, albeit it is more complicated with calls. Even if a human agent takes over the chat or calls at some point, they will most likely still benefit from a clear solution flow to solve the customer contact.

automated customer service

Basically, we learned that we need AI-powered customer service to understand the customers and tools for product management to design solution flows, analyze the success of those interactions and maintain them over time. The AI part is something that has been solved by some through chats and in fewer cases through calls. The product management side is something that we have only seen solved fully in our own TitanCS solution. Here are some key questions that should get an answer from the product managers by using the tool:

  • Do I as a product manager know what to do next to improve my customer service KPIs?
  • Do I as a product manager understand if I’m held back by data, AI, the solution flows or something else?
  • Do I as a product manager have a clear system to design solution flows and maintain them over time?
  • Do I as a product manager have a clear process to review interactions with customers to understand where the interactions do not work out as planned?

Putting this kind of a system together is not easy, but we’ve proven that it is possible in both chats and calls.

Once it works, you will understand that customer service is not just about serving customers. It is about product management having the intelligence and ability to fix broken processes, to reduce the number of contacts, to solve these correctly on the first attempt, to be able to quickly redefine how issues should be solved, to understand what is working and what has an impact, and minimizing human errors and interpretation issues.

Webinar: how did Elisa automate customer service?

Webinar: how did Elisa automate customer service?
Mailiis Ploomann

Find out how Elisa, one of the leading telcos in Nordics, handles customer service automation during the crisis.


Customer service is broken in many companies and sometimes this is done knowingly. However, only some can afford this luxury and most others should strive to fix it. This is important both in the long run and in the short term to improve your product and processes. With the help of AI and correct product management tools, it is possible to automate customer service. Both chats and calls can provide quick responses, actually solve your customer contacts and create an automagical experience. If you wish to learn more about TitanCS, you can always book a demo with us.

Kristjan Jansons


Kristjan has been studying and working on machine learning projects for more than 5 years. After acquiring a Master’s Degree in Computer Science and Machine Learning, he started working at Milrem Robotics as the Team Lead for Autonomous Vehicles, helping to build self-driving vehicles. Kristjan also has experience in building intelligent systems for data centers, robots and electric formulas; also with computer vision and image recognition.

Talk to us

kristjan jansons