8 steps you should consider when introducing a chatbot for utilities

Chatbots for energy providers spark your interest? Understandably, as customer experience and the associated customer loyalty is becoming increasingly important for energy suppliers. New competitors, including Tesla and Google, are entering the market. At the same time, consumer expectations are rising on a broad front. It is therefore more important than ever for energy suppliers to invest in their customer service and to continuously improve it.

Utility companies themselves have also recognized this problem. In a 2019 utility study (German and focused on Germany), 77 percent of energy suppliers stated that they are heavily involved in the areas of “sales/marketing/customer service”. Creating a chat bot is a journey, but also a good starting point to focus on customer-centric intelligent automation in these areas in particular.

In this article I’ll explain why chatbots are relevant for utilities, how to create a chatbot and give you 8 suggestions for how to run a chatbot project at utilities if the topic is new to you.

Why is a chatbot important for energy providers and what is it anyway?

Chatbots are defined as conversation systems with natural language skills of a textual or auditory nature. A chatbot examines user input and provides answers and questions based on routines and rules.

But why is a chatbot important? In 2011, Gartner predicted that by 2020, 85 percent of customers will have a relationship with a company without human interaction. My perception in 2020 is that in the age of online services, more automated contacts with companies are more accurate than ever.

On the other hand, a study by the XM Institute found that emotions are critical for businesses. Emotions are the component of the customer experience that has the greatest impact on customer retention behavior. Consumers who have given a company a high rating for emotion are more likely to buy from that company, recommend it, forgive it, try it and trust it than consumers who have given a high rating for success or effort. As a company, it is therefore advisable to improve the emotions associated with the company.

Even small changes count. Even modest improvements in customer experience result in higher customer loyalty. In particular, the probability of recommending a company and the intention to make future purchases can be significantly influenced by even small improvements in customer experience.

A chatbot can influence emotions through intelligent conversation design and use of corporate identity and is an additional channel through which customers can interact with the company in the way and within the time frame they prefer. On the other hand, this gives you the opportunity to get to know your customers’ language usage and questions and to use them step by step.

Mercury, for example, has seen an increase of more than 30 percent in the number of customer interactions via its online platform. It mentions that some of these interactions are new, but that they have not seen a decline, for example in telephone traffic, from which these interactions would have come earlier. This is desired by Mercury, as they want to address their customers on topics other than, for example, the bill.

”This is a positive aspect because we want the engagement with Mercury to be not just about the bill.

Kevin Angland – General Manager Retail & Digital at Mercury NZ

Tim Aynsley, head of ICT at Mercury, says chatbots help to reduce the burden on contact centres so that staff can focus on valuable conversations with customers.

In summary, chatbots offer the opportunity to achieve a high return on investment with minimal effort and are therefore worthwhile to consider.

Projectmethodology for a chatbot at a utility company

So far we have discussed why a chatbot should be important for utilities, but how do you run a successful chatbot project?

The methodology used should start with an understanding of value and ROI to prioritize use cases, minimize time to market and maximize performance and user experience. Our proposal is to divide a project into at least 3 phases, from design to implementation to production.

Design

The first phase is called the design phase. It is the most critical phase, as the success of the project depends on it.

1. Definition of business objectives and ROI

   First determine the ROI and business goals you expect at the end of the project. You should work closely with the business units to decide which KPIs can help monitor the ROI of the bot and the success of the project. It is important to understand that you should never start with the use case, but with the ROI. This will definitely help you reduce the scope of the project by focusing on what is important and adding value to the business.

   Mercury for example started with simple questions like “What’s my account balance” or “I’m moving – how do I get a new connection? They want to inspire and reward their customers and make it easy for them. You started with an MVP (minimum realizable product) for simple requests and plan to expand it step by step

2. Platform training
   

   Next, you should select a software solution and train your business and IT teams on the selected platform. This should familiarize the project team with natural language processing, a high level of understanding of bot architecture and the concepts of intentions, entities and capabilities.
   The benefits are that the team builds trust with the solution and understands that the platform is not a black box. It also makes sense that the people involved understand the concepts of the platform and that they think about intentions, entities, and capabilities when moving on to the next steps of the project, so that the process of building the discussion flow is faster, and finally, it will prepare the team to take over the bot at the end of the project.
   Make sure that the team is diverse and cross-functional – a chatbot project should not be a technology-driven project.

3. Designing the flow of conversation
   

   Next, you should narrow down the identity of the bot and the use case based on the ROI you want to achieve during the project. Again, it is important to understand that the use case is derived from the ROI and not the other way around. First define the persona of the chatbot and then design the flow of conversation regarding the use case with the department. For example, use diagrams to visualize the conversation flow.

4. UAT definition and common communication measures
   

   Continue creating UAT (User Acceptance Testing) conversations. The test conversations should contain conversations that are in line with the use case, as well as cover topics and errors that differ from the use case. You should aim to create between 5-10 UATs that describe the conversation between the bot and a customer. The UATs will validate the created chatbot. This is also the time when you explain the different communication strategies (social network, press release, websites, etc.) and create a communication plan for the launch.

5. Technical requirements
   

   Next, you should clarify the technical requirements of the project. For example, you need to answer questions about which channels the bot will be available on. Do you have access to all APIs needed to answer the questions? Is user authentication required or are these questions general?

When all steps are completed, you write the functional and technical specifications. All specifications should be signed and approved before development begins, to freeze the scope of the project.

Tips and advice on technical requirements

Technical requirements can drive up the cost of any project, so as mentioned above, you should first define and limit the scope. However, the technical requirements should influence e.g. the choice of platform, as synergy effects can arise.

Let us assume that you are currently running an online service based on SAP Multichannel Foundation for Utilities. The interfaces provided by MCF can be a good integration basis for the chatbot. If you already run the user interface on the SAP Cloud Platform and users are logged on to the backend system via principal propagation, SAP Conversational AI would be a perfect platform for your chatbot, because SAP CAI can also use the IdP configured on the Cloud Platform.

If you already use SAP Contact Center, SAP CAI is also an ideal candidate, because SAP Contact Center can serve as a fallback channel for SAP CAI, i.e. transfer the chat to an agent if necessary.

So when choosing a solution, think about possible synergy effects and avoid isolated applications.

Implementation

The second phase of the project is the development or implementation phase. This phase includes the steps required to implement the chatbot solution on the platform of your choice and the backend integration of the project. The duration can vary greatly depending on the number of APIs to be integrated and also depends on custom development of UX components within the chat or implementation of authentication.

Before implementing the Chatbot, a design review is required to better understand what you want to create. This includes, for example:

• What intentions, entities and skills do you need?
• A brainstorming session to question the design

After implementation, you should hold test and training sessions. Testing means playing with the UX of the bot, while training means testing the semantic understanding of the bot.
These sessions are essential for the viability of the project. To make the bot more robust, you need to confront it with potential end users to ensure that most recurring patterns have been identified and covered.

The goal should be to validate the user experience in one session and the chatbot’s understanding and response to unexpected input in another.
With and from the test results the chatbot can then be further improved.

Production

The production phase begins with the execution of UATs to validate the chatbot created. Once the UATs are validated, you should start the test sessions again. As a rough guide, you should calculate with 50 to 100 hours of beta testing to make the bot as robust as possible before it goes into production.

Once the testing sessions are completed, you will need to organize a knowledge transfer to the teams that will maintain the chatbot for the long term. This includes explaining the bot in the chosen platform, the intended intentions, the existing entities, and the capabilities. In addition, the conversation paths and any additional developments and integrations of the bots should be explained. This knowledge transfer should involve people from the business departments as well as from the IT team. The final step is to bring the bot into production.

Once the bot goes into production, one of the main tasks is to monitor the bot daily for at least the first 3 weeks.

A soft launch, such as Mercury did, can also be useful. This has the consequence that if an error has occurred during quality control, not the entire customer base is affected and the error can be corrected at an earlier point in time.

An important point to make sure you get support in obtaining resources, not only to provide the basic skills, but also for ongoing improvements.

Key Takeaways

The three most important things you should keep in mind after reading this blog post are

• The improvement of the customer experience and positive emotions towards the company have a positive effect on the ROI (Return on Investment).
• Chatbots are a tool that helps you to improve in these areas compared to pure online self services.
• Chatbots can achieve great results with minimal effort.

However, you should focus on your return on investment, deduce the use case and then embark on a cross-functional journey rather than starting a one-time IT project.

Remember that every customer interaction with your chatbot is an opportunity to learn, and that the right tools will give you the opportunities you need to improve those interactions.

If you are looking for support with your project, we are here for you. Just contact us and we will discuss your requirements and goals together.