15 Hardest Chatbot Implementation Challenges (and Overcoming Them)

Is your organization planning on implementing a chatbot? As someone who has been a chatbot product owner, I am going to share with you real experiences that I’ve had implementing a chatbot across over 40,000 employees.

I wanted to share this with you so that you understand the real life challenges of implementing one and how you can potentially overcome it.

Chatbot Implementation Challenges to Overcome

1. Implementing an Affordable Chatbot – challenges for Building & Maintaining

Depending on how you implement your chatbot, it can be expensive to not only set-up, but also to maintain. Currently, every single company is offering a chatbot solution for their platform. If you are an organization that uses multiple platforms to manage your business, chances are your human resource, communications, data lake store, and support platforms probably have their own chatbots. Having to piece meal all of these different platforms to have one main platform may be a huge endeavor if you want one cohesive chatbot.

The other option is that you can create your own custom chatbot, but that will also be expensive because you will need an engineer to custom code your chatbot and also maintain it. Having a team to support it when things go wrong is also something to consider as well.

The solution to having an affordable chatbot is understanding your first big use case as well as understanding big picture what you are trying to achieve with your chatbot. For example, are you trying to achieve ticket deflection to reduce headcount or are you trying to get data to use resources efficiently so you can save money.

2. Leadership Buy-In

At the C-Suite level, I’ve often found that it takes a long time for them to understand the value behind a chatbot. The conversation always seems to be around “how do we use a chatbot to reduce headcount or money”, when the actual real value is in the DATA that a chatbot can provide. Chatbots provide an IMMENSE amount of information. If your chatbot users are using a chatbot, they are hoping to solve a problem because the current available venues that they are aware of do not provide content. Thereby, you can use that information to your advantage by knowing WHERE to invest your resources to improve content, which will then help your audience.

3. Privacy and GDPR Considerations

If you are an enterprise organization, you are probably on the up and up with GDPR. However, if you are not up-to-date on these regulations, you need to ensure that the data that you collect from the chatbot conversations are compliant, especially for users in Germany and most of Europe. As you develop your chatbot and data collection strategy, ensure that you are reviewing your collection practices with your legal or privacy team. An architecture and data analytics review may be needed to ensure that you are masking private health information or even discerning the specifics of who your audience is.

4. Understanding Your Audience

One way to understand the audience is using data that you may already have with any search engine data you may have. This is a great way to discern what users are already asking about so that you can start creating skills and FAQs based on actual data. Often times, organizations may not understand how to use search content within their own enterprise search tool. If you hire an SEO (search engine optimizer), they may be able to provide insight on understanding your audience through intent based search.

Understanding the Audience Intent

One of the biggest challenges is to understand the intent of the user and being able to decode the intent hidden inside the chat conversations.

5. Finding Stakeholders for Your User Stories

Whether you are building an internal chatbot or an external chatbot, it is important for you to have USERS that are providing some sort of insight on your user stories and requirements for the skills you will be creating. Even if you are niched down to the skill you want to develop, often times it is helpful to have a pre-design meeting to ask them, “If you needed to look for X information, how would you start a dialogue with a chatbot like Alexa?”

This will hep provide the initial utterances and intents that you need to develop for your chatbot.

6. Adoption of Your Chabot

When you get stakeholders, you’ll also need to make sure that the stakeholders have buy-in from upper leadership if you want to continue with innovating your chatbot. As you continue to create skills with your chatbot, you’ll learn about the different needs of your audience based on which skills are used the most.

If you get upper leadership buy-in, they will have the authority to help promote the chatbot and/or the skills it has so that you can increase adoption/usage, which will allow you to see the return on investment on both the time and resources you utilized in creating your chatbot.

7. Ensuring the Chatbot Fits with Your Brand Identity

If you are going to name your bot anything other than your company’s name, ensure that you are following any branding guidelines or at least reviewing the branding provided from your team. There is a perception out there of an AI bias of having a virtual “assistant” being female. You’ll find some of the more popular chatbots do have male versions as a counterpart, but often with the female bot leading the way. In an effort to avoid a bias towards females as being only labeled as an assistant, your chatbot should have a gender neutral name.

In my own experience, the leadership before me utilized a female name for our chatbot, and we were able to discern feedback from a few users that it seemed derogatory to have a female represent the company’s chatbot. A brand overhaul was much needed went our proof-of-concept turned into large scale enterprise platform.

In addition, create a framework for your chatbot’s personality and define how it will behave during a conversation. Will it be quirky, professional, fun? Outlining these personality traits will help you when you start expanding and having other teams utilize the chatbot for their own skills.

8. Analytics

While you are developing the skills and the questions your chatbot will be answering, you’ll want to determine what type of analytics you are wanting to gather from the chatbot. Often, with the platform you choose, the data may be enough. However, if you are looking for detailed data, ensure that your platform that you are using for the chatbot will provide that. Examples include:

How do you determine failures or successes in chatbot conversation?
What do you refer to as a session, a conversation, or a message?
Are you able to tie back the user with an email, their demographics, or organization?
What will you do with that information? Do you plan on sharing this with content owners?

These are questions you should spend time answering BEFORE implementing your chatbot so that you have a database that can house this data.

9. Providing an Intuitive User Interface

In order for users to actually adopt and use your chatbot, it MUST be intuitive. It is important to hire a designer or a human factors designer to help with the conversations with your chatbot. With the skills that you implement, the design must be consistent from skill to skill so that your users can have an understanding of how to interact with the skill.

What you will run into is natural language understanding, which will then require a data scientist to help implement algorithms to understand the context of that user.

10. Managing User Feedback and Support

An idea for managing user feedback and support is by implementing a feedback loop cycle. Within the chatbot, you can implement a sentiment analysis to understand when it is a negative sentiment. Being able to use the data to label these sentiments and to review these sentiments will allow you to improve your bot and figure out where the chatbot is getting confused.

I also recommend creating a round table for feedback if you are able to get close to your users and getting open ended feedback.

11. Creating Algorithmic and Automation Features – Infusing NLP and Machine Learning

The beauty behind a chatbot is that you can implement small apps inside of the chatbot that can launch other small apps and skills other teams maintain. The great thing about this as you create processes in place to review the data, use that data to continually re-learn content that is being refined to continue feeding it to the chatbot to relearn in the future. This can be done via an automation tool or great content management system that feeds into the chatbot.

12. Limiting Technical Debt & Technical Currency

If you are building a custom chatbot or using a platform where are developing custom skills for embedding into a chatbot, my recommendation is to make it platform agnostic. With the changing landscape of technology and chatbots becoming the flavor of the month, there are lots of new platforms where there isn’t really A SINGLE platform that has stood out as being the best at EVERYTHING. You’ll find that you will land on your first platform, innovate, and realize through your iterations that you’re missing a key feature.

Making your coding agnostic to a single platform will allow you to move platforms easier and without a lot of tech debt.

In my experience, the technical currency that we had to manage included how often we had to upgrade the framework, which was not even the platform, it was just the version of the platform. After delivering over a dozen skills, having to go back to working on old skills felt like it took us a few steps back at a time when we were just making progress on increasing adoption due to how fast we were releasing new skills.

13. Measuring Success and Providing ROI

It’s really important that you determine from the beginning of the chatbot and also any additional skills released how you will MEASURE the ROI of the chatbot. The real value of a chatbot is not just reducing labor and support. It is truly the DATA that is inside of these queries within the chatbot conversation that will help dictate what strategies your business needs to take and what your users are asking for.

When we met with our financial team to get agreement on the financial value statements for our internal chatbot, implementing this chatbot was determined that there was more of an experiential gain and user efficiency into finding the information.

14. User Resistance

If you are developing an internal chatbot, you’ll often feel resistance because chatbots have a preconceived notion of being able to takeaway jobs. Being able to message to employees that the goal is to help reduce redundant work in order to allow employees to focus on higher level work will make them more productive and doing the work they want to be doing.

15. Human Handoff

You’ll need to determine whether your chatbot will have human handoff. If you do agree, this will be another expense to consider. Being able to transfer the conversation to another human during a conversation should feel seamless without a need to moving to another platform, YET be able to allow the user to discern the difference between talking to a chatbot versus a human.

Summary of Problems with Chatbot Implementation

Chatbots are becoming very popular with both big and large business because of the insights that are housed in the searches, queries, and conversations within a chatbot. Businesses that are addressing the importance of gathering this data and using it towards their business strategy will be in tune to listening to what their clients and employees are asking for. Being able to address these challenges head on in the beginning will allow businesses to succeeded past these challenges of implementing their first chatbot.

15 Chatbot Implementation Challenges to Overcome