Chatbots

Overcome Chatbot Failure and Build Better Conversational Bot Experiences, Faster!

Conversational Artificial Intelligence (AI) experiences are a hot ticket item for enterprises as organizations realize the necessity for more automated digital engagement in the midst of doing business in a Covid and post-Covid world.

However, delivering good conversational experiences at scale through the implementation of digital assistants and chatbots is no mean feat. For some organizations, chatbot failure or poor performance has been discouraging. Yet, the demand for more automated and personalized self-service experiences is growing at a fast rate with a global conversational AI market CAGR of almost 22% per annum over the next 5 years.

So why are some chatbot projects unsuccessful and what approaches can businesses take to overcome problems with AI chatbots and deliver better bot experiences faster.

The Rise and Fall of Chatbots

Possibly one of the best analyses of where chatbots fall in terms of expectations is in Gartner’s Hype Cycle for Artificial Intelligence. In 2019, chatbots hit the peak in inflated expectations with a forecast that they would plateau within 2-5 years. Roll on 2021, and Gartner’s hype cycle for AI places chatbots at the lowest point in the trough of disillusionment.

So despite increased investment and adoption of chatbot solutions and resulting market growth, especially driven by the impact of the pandemic, chatbots are pretty much at their lowest ebb in terms of fulfilling expectations. Here are some of our takeaways on why chatbots have led to some disillusionment and what you can do to overcome these chatbot problems.

Four Top Reasons Why Chatbots Fail

1. Assuming that Chatbot = NLP

For chatbots to be valuable in an enterprise context they need to rely not only on the underlying NLP engine. The NLP is just one component to building enhanced conversational AI experiences. It is also increasingly a commodity and while advancements in NLP and powerful language models like GPT-3 (Generative Pre-trained Transformer) are enabling more human-like conversations, there are other critical components that go into creating and maintaining good conversational bot solutions.

A more holistic approach that includes data & system integration, security, identity, and access management, data storage, training, operations, deployment, and management is needed to create valuable solutions that can be enhanced and maintained efficiently over time.

NLP: The Tip of the Iceberg for Enterprise Conversational AI Solutions

A do-it-yourself approach to creating chatbots using just an NLP tool (e.g. IBM Watson, Google DialogFlow, Amazon Lex, Microsoft Luis,  Rasa, etc.) may work well for the initial basic Q&A bot but this approach becomes costly and complicated as more sophisticated solutions and more bots are deployed. The problem with building a solution just using NLP toolkits is that it requires good developer skills, either internal or outsourced.

This ends up being costly and is problematic for business analysts (or citizen developers) who want more independence in creating and maintaining their own conversational experiences. They don’t have the skills or time to build all of the additional components needed to turn a technical tool into a flexible solution that can be owned and run by the business. Also, because the business rules are embedded in the fulfillment and are hardcoded, changing them is an IT task and not a business task. Agility becomes an elusive goal and time to market increases.

Solution 1: Adopt an Enterprise Conversational AI Platform that provides a wide range of pre-built, reusable components that accelerate the development of enterprise-grade conversational bot experiences.

Learn about how a platform approach supports the creation of advanced and enterprise-grade conversational experiences while lowering the total cost of ownership. Download a free copy of our eBook: The Costs and Complexity of a DIY Approach to Building and Managing Conversational AI Solutions

2. Wrong or Poorly Scoped Use Case 

Another reason why chatbots fail is rooted in the choice and scope of the business use case for the bot. A chatbot solution needs to align with business priorities and goals and deliver measurable business results. Hence it is important for lines of business or business owners to have significant involvement in the concept design and operations. After all, they are responsible for the success of their self-service and digital investments. Chatbots that are solely in the remit of the IT organization can often become technology-oriented rather than business-focused solutions so business enablement is key.

A chatbot should be viewed as a solution to a business problem rather than just a piece of software to engage with customers or employees. The initial scope should be well defined and have guardrails. Overstuffing a bot, especially an early-stage project, is often a leading cause of poor performance. It is better to start with identifying some quick wins and a highly focused mission, for example, an initial bot may be designed to automate the top 8 most common customer queries with agent handover for less common requests. 

Solution 2: Seek subject matter experts to guide you in the planning & design phase of your chatbot project and follow some tried and trusted tips.

In this blog article about chatbot strategy, we mapped out 7 things to consider in designing and planning your chatbot project. These are very useful tips for businesses in the early stages of planning or creating chatbots. Working with seasoned professionals that can help you choose the right business case and get your chatbot to market quickly and deliver fast ROI is also a helpful way to approach the planning and design phase.

3. Operationalization Challenges 

Digital transformation initiatives tend to be long journeys and large-scale projects. Chatbot or digital assistants are great ways for an organization to get some quick wins along their DX journey but only if they can take an agile approach and move quickly from concept to production.

According to Gartner “the most significant struggle of moving AI initiatives into production is the “inability for organizations to connect those investments back to business value.” So despite increased investment in AI technology by enterprises, unless these projects can be accelerated from concept to production and linked to clear business outcomes they may be on a path to failure.

In a recent survey by the same analysts, their findings were that only half of AI projects make it from pilot into production, and those that do take an average of nine months to do so.

This touches on previously mentioned business enablement as well as having a well-defined business use case.

Speed to market is key to these quick wins and that is often down to the approach a business takes to building, scaling, and managing their bot projects. 

Take the example of our client, Cenlar FSB, the leading loan subservicing provider in the United States. When the pandemic hit in 2020, Cenlar recognized that the terms of the CARES Act (Coronavirus Aid, Relief, and Economic Security Act) would create an influx of calls to their contact center from distressed borrowers seeking information about mortgage payment options and that they would need to provide answers to borrower questions quickly and in the manner they wanted. A digital assistant was spun up in just over 30 days (a record-breaking speed for brand new technology)and quickly yielded positive ROI but more importantly led to increased borrower satisfaction.

Solution 3: Start with a low-hanging fruit use case and avail of low code tools and platform approaches that accelerate your projects from concept to production.

When it comes to enterprise chatbots solutions capabilities such as low-code developer tools, reusable skills and components, pre-built use case blueprints, out-of-the-box APIs and channel engagement adapters enable greater speed to market while enterprise-grade features around security, analytics, and integration are important to satisfy corporate governance and control requirements.

Discover the benefits of a Conversational AI Platform approach in this blog article.

4. Inflexible Solution Architecture

As chatbot use cases become more complex often a single-bot solution cannot support the experience well enough. For example, a company may build a digital assistant to handle common customer queries and roll this out in an initial phase. Over time, they may decide that the bot should also have the capability for the user to transact, bring them through a multi-step journey, and add more capabilities, content, personalization, languages, and skills. By adding additional capability, you can potentially erode the capacity that the bot has to deal with the actual use case i.e. answering the common queries. The concept of fitting your use case into the bot also helps explain a phenomenon that some companies are seeing, where the conversational experience declines when they expand the functionality.

This has implications for how you architect your bot solution to meet the requirements of your use case. Will a single bot be sufficient? If not, how will you architect multiple bots so that they can be coordinated and work together to fulfill the need? 

Solution 4: Think Multi-bot Architecture if you foresee a future where you may want to enhance your solutions and add capabilities or where you may want to orchestrate across multiple bots.

Our unique multi-bot orchestration architecture is leading the pack in helping businesses build and manage more advanced conversational experiences. To learn more about this you can watch the webinar recording here.

An Additional Note on Multi-bot Architecture for Scaling and Creating Advanced bot Experiences

Collaboration across multiple chatbots has emerged as both a requirement and a challenge for enterprises as they expand the scope and number of conversational AI projects across their business. Different approaches, NLP engines, tools, and conversational user experiences can emerge, impacting the ability to maintain a consistent brand experience and a single access point for the user no matter what their interaction may be. 

The way to overcome this is through multi-bot architecture that enables the business to break up the chatbot solution into more manageable or composable components. In this scenario, a number of skilled bots can be created, each with their own specific mission (and even with their own NLP that can be different from other bots). These bots can be orchestrated by a central virtual assistant or bot orchestrator that manages the conversation to and from the user, navigating across the skilled bots and routing to the appropriate bot according to the intent.

This represents more of a microservices approach rather than overstuffing a single bot with all the capabilities needed to execute the business use case (i.e. an unwieldy monolith). It also enables easier and faster scaling and maintenance of large-scale chatbot solutions, as additional skills can be added without breaking the experience.

Check out more Great Content about Chatbots and Conversational AI!

Where are you on your Chatbot Journey? To discover more please download our eBook: A Conversational AI Journey Guide

For those of you that would like a free trial of our Conversational AI Platform please sign up today.

We keep updating our content so please explore our blog and resources