The biggest pitfalls related to chatbot implementation in enterprises

As a business customer, once you start thinking about automated customer service tools, you need to be prepared for potential challenges that you will have to face. Many end-users, once they start interacting with chatbots, immediately ask for a human agent. It is usually caused by past negative experiences on other websites. And honestly, we cannot blame users for that. Let us present a few mistakes made by companies and how we suggest dealing with them.

There are several reasons why chatbot implementations seem to fail:

• Unrealistic expectations towards automation
• Lack of skilled staff within the team
• Time pressure

The issues presented above are already problematic enough to, sooner or later, lead towards a sure failure. However, after many interactions with business clients, our action.bot team has been able to identify even more pitfalls you should be aware of to avoid a certain disaster.

Let us share some additional advice that might help you tackle them properly.

1. No attention to UI and dialogue flow design

What companies say

“Let’s automate the basic FAQs we have on our website. That should be more than enough”

“Let’s use any UI template and color it. The visual design doesn’t really matter”

“Our colleagues from the customer support will prepare the answers and design the flow. That shouldn’t be a big deal”

What we have learned

As you can see, the problems raised above share the same root: lack of experience in dialogue design and no attention to the visual side. The main reason why virtual assistants are inaccurate in their answers is not that their technology is lacking but because of poor design and technology implementation skills.

It is nearly impossible to predict all possible questions and behavior of customers in the chatbot window. However, if users end up stuck in error loops, it is most likely because the conversation s not designed correctly. This also applies to the User Interface (UI) design. Customers report that most of the chatbots simply look the same. Hence the UI often doesn’t seem to be an integral part of the channel or application. The look of the chat window is simply discouraging.

Would you entrust your bank account data or savings to someone who has an unpleasant appearance, wears a dirty t-shirt, and provides you with incoherent answers?

A few tips from action.bot team regarding design

• Check what your users really expect from you: based on your knowledge (historical transcripts of the chat, call center recordings, e-mails), you can easily identify most issues and requests from customers to know what should be automated within the chatbot dialogue
• Customers appreciate a nice look&feel: leave the design to the UX/UI/conversation design experts. They will be able to properly develop the dialogue flow, create the chatbot’s personality aligned with your brand
• Chatbots are not only about plain text: take advantage of UI elements like buttons, videos, attachments. They will help users go through the process and make the conversation more engaging. The visual elements will pay off with higher task completion rates and successful user outcomes
• Use proper tools to organize the structure of dialogues from day one: when you automate 20 intents, it is not a big issue (an Excel sheet is more than enough). However, as the bot grows, you will need to have a system (dialog creator) that allows bot builders and analysts to navigate and modify the scenarios.

Lack of knowledge and too high expectations

What the enterprises say

“Let’s get a $10 advanced customer service chatbot”

“The pricing is far beyond our expectations”

“Peter and Jane will take care of the chatbot’s deployment”

“We have various needs. We want a multifunctional chatbot to automate the sales process, operations, employees’ onboarding, and collaboration with B2B partners from day one”

“Chatbot, voice bot, and RPA — we want all in one”

“Our solution provider should prepare the dialogue on their own. That’s what we pay for”

What we have learned

Most corporate clients assume that creating chatbots is a piece of cake: big tech companies keep announcing their latest breakthroughs in developing intelligent AI systems. It might lead the first-timers to think that the AI will immediately understand and solve all their problems. Believe me, even minimal improvements in technology are not achieved by small teams in a short time. They require a constant process of improvement and development to reach a particular level of maturity.

If we are talking about a real digital assistant that is supposed to become a new communication channel with problem-solving capability and NLP training data, the challenge is a complex one.

A few tips from action.bot team

• Build an interdisciplinary team: dialogue design requires the involvement of several departments to agree on real business objectives and define users’ problems. Try to involve your customer service, contact centre, marketing, sales, and IT departments
• Get involved: do not expect that your solution provider will solve all your problems. Be present at every stage of the chatbot’s development. It will significantly improve project results.
• Decide on the proper scope: think big, start small — make a list of all the use cases/scenarios you would like to automate, prioritize them and choose a few quick wins that could be a good start in the automation journey. The best approach is to limit the scope to a narrow set of intents to get a quick win and then gain momentum
• Be patient: naturally, as a business customer, you want to see results as fast as possible. To achieve that goal, companies typically adopt a relatively short schedule to finalize the whole process (e.g., three months, which in the enterprise world is a brief timeframe). But keep in mind, 80% of the chatbot’s success depends on how you adjust and improve its performance after going live.
• Measure, measure, measure: You should keep analyzing the conversation data: usage, completion rates, user feedback, user paths. That will help you identify areas for improvement. For example, the conversational engine used by our solution (Watson Assistant) allows us to constantly track the data in real-time and react to conversations with weak understanding to increase the number of resolved requests.

If this article has not scared you off and you would still like to start your adventure with conversational automation, check our website action.bot or contact us directly. We will do our best to avoid these problems and provide you with a truly powerful digital solution.