Building a Chatbot is Much Harder Than You Might Think

In 2016, I founded Espressive on the thesis that enterprises were struggling to deliver help to their employees and needed a better approach. IT has been at the forefront of service management with legacy IT service management (ITSM) tools that are expensive and complex to deploy and use. That’s been okay with IT because they are highly technical and rely on these tools to run their daily operations, but it’s not okay for the rest of the enterprise. These other departments generally support employees with a variety of tools including knowledge bases, intranets, phone numbers, and email and are looking for a better way.

But whether it is IT using ITSM, or other departments using their own tools, they all share something in common – employees are not getting the help that they need.

Many CIOs tell me they believe artificial intelligence (AI) could help deliver a better approach for employee self-service. And most believe that chatbots are the answer. So, the question arises whether they should build or buy a chatbot. That is what I am going to explore in this blog along with the question of whether chatbots are enough to solve the employee self-service problem.

You Built Your Portal, But Has That Paid Off?

In the world of IT Service Management (ITSM), the build approach is well-known. Sometimes building makes sense, such as creating a workflow to automate an enterprise’s specific process, as each organization has unique processes. Where building doesn’t make sense is the employee self-service portal. Building custom portals is both expensive and time consuming, and research shows that it has not yielded a good return on investment with employee adoption of today’s portals at around 10-15% (see “Why Has ‘Shift Left’ Failed in ITSM”).

In order to build a portal, one needs to have coding skills plus visual and interaction design skills. Since those skills are generally not easily accessible to IT, enterprises often outsource portal design to an ITSM implementation partner. Not surprisingly, we’ve heard from these partners that they are having a hard time attracting and retaining top notch UX and design talent. This is a serious issue given we all know that building a true consumer-like app requires top talent.

Maybe Chatbots Have Come Along to Save Portals

AI is now one of the hottest topics in IT, and research shows that ITSM is one of the highest-ranking use cases being considered by CIOs. This makes sense because a large portion of the CIO’s budget is being used to fund the help desk which is bogged down by answering repetitive questions. AI is a promising technology for automating the resolution of common issues, which can reduce MTTR while allowing the CIO to reassign help desk resources to more strategic projects.

But the thing about chatbots is that they don’t replace portals – they are additive to portals. So, the expensive and time-consuming work that goes into building and maintaining portals does not go away. To add fuel to the fire, chatbots require a whole new set of specialized skills on top of those required for portals. In fact, it is just one more thing to build and fail if you don’t have the right experts involved.

More importantly, though, is that adding a chatbot to a portal that has experienced low adoption will not change the final outcome. The “build it and they will come” philosophy has historically not been successful, and history is likely to repeat itself in this case given your employees have already voted against the portal experience.

What’s So Hard About Building a Chatbot?

ITSM vendors provide toolkits and frameworks for building chatbots, plus there are third-party tools such as Watson or Agent.ai. So, what’s so hard about building a chatbot with all of those options? Well, it’s not about building a chatbot – it’s about building a chatbot that employees will want to use.

First, employees have a language that they speak, and they don’t want to learn a new one. Linguists are needed to ensure the chatbot understands common employee language and all of the possible permutations of a given question being asked or issue being reported. I have seen one too many chatbots that only respond if the question is asked precisely in the way it was configured, which will infuriate employees and quickly create an abandonment problem. Once deployed, AI experts and data scientists are needed to constantly analyze data to tweak the algorithms and configuration ensuring the chatbot is actually responding correctly to employees and to identify additional areas for possible automation.

Second, employees want an answer that is personalized to them – that’s why they call the help desk in the first place. Most chatbots are designed to play a game of 20 questions with an employee to narrow down the number of knowledge base articles served, but they are still knowledge base articles which are not personalized, are often technical and out of date, and might not even contain the answers employees are looking for. When the answers are not there, chatbots can open tickets on behalf of employees, but then employees are sent off on their own to figure out how to contact the help desk, and the experience feels like a transaction as a result.

Barista, Our VSA, Changes Everything

At Espressive, we knew we could go beyond chatbots to deliver a highly personalized experience that is not based on knowledge bases or portals using AI. Barista – our AI-based virtual support agent (VSA) – provides personalized and immediate answers to questions and issues in the language of employees. Barista is a multi-platform native app that provides an immersive experience that stays with employees until their issues are resolved. Because of that, employees get the help they need which develops a level of trust, ensuring they are willing to reach out again.

Barista doesn’t come with a toolkit for building dialogue trees that require experts and linguists to develop. Barista comes out-of-the box with an Employee Language Cloud which includes an extensive vocabulary tailored to the workplace so that Barista can speak the language of employees from day one. Barista is literally capable of understanding 15 million things without any customer input and then learns-on-the fly with each and every employee interaction to get even smarter over time. And because enterprises are global, Barista is multi-lingual, so your help desk doesn’t need to be.

Bottom line – why try to build something that requires significant time, highly specialized skills, and still relies on portals? Worse yet, why build something that is likely to fall short of your employee’s expectations and result in low adoption? With Barista, our VSA, Espressive customers report employee adoption of 50 to 60% with reduced call desk volume of 30 to 50%.

There are times when it is better to build versus buy. This is just not one of those times.

I am incredibly proud of the Espressive team for building the most exciting product I’ve ever had the opportunity to work with. I would love to give you a demo so that you can see why.