If chatbots or AI-powered conversational apps were cats, they would be fast approaching the current upper limit of their possible lives. Their obituary has already been written multiple times by now.
Clearly, they have not been able to live up to the initial hype (yet!), but does this mean that their future is doomed? In the post COVID-19 world, will cost-cutting across the board and lower business appetite to take risks with newer technologies kill the chatbots as business applications, or will they make a comeback as drivers of new growth coupled with unprecedented operational efficiencies and reach?
Will text remains the dominant medium of interaction with conversational AI, or with breakthroughs in speech recognition, will voice become the preferred medium of interaction with bots? In this multipart blog series, we will discuss what went wrong before the pandemic and what the future holds for chatbots as tools that solve real-life problems and deliver value commensurate with the initial promise.
Although chatbots have been in existence for over five decades now, their large scale adoption as tools that deliver value to businesses is quite a recent phenomenon. The adoption was primarily triggered by three factors: -
- Relative maturing of AI allowing to move away from brittle rule-based bots. This includes advancements in the field of voice recognition.
- Cloud computing allowing anyone across the globe to develop chatbots with little investment, leveraging algorithms being developed by large players
- Texting & Internet proliferation – Behavioral changes in people with the high adoption of conversational channels, including messaging and access to the internet, creating a significant user base for chatbots
In 2011, IBM's Watson made quite a few headlines by defeating two of the ‘Jeopardy!’ champions - Ken Jennings and Brad Rutter. In 2016, Google acquired API.ai, which then evolved into Dialogflow, a platform for building chatbots. Microsoft also jumped into the fray with GA, its own bot development ecosystem in late 2017.
The period from 2015 onwards also saw at least scores of startups launching their "No-Code" or "Low-Code" platforms, often focusing on a specific industry segment while the number of bots that went live could smoothly run into hundreds of thousands, if not millions.
What happened then?
Going by Google trends (which in my opinion are a good representation of our online lives & interests), the popularity of chatbots rose exponentially till about the middle of 2017, after which the trend has acquired a more sober, yet growing trajectory.
Depending on who you speak with, there are a variety of issues that resulted in the fizzling of the initial exponential trajectory of interest for both consumers and businesses. Some examples include the inability to understand what users were trying to achieve, the lack of useful use cases implemented end-to-end and not having any catalyst for change in consumer behavior etc. In our own experience, there are multiple other important reasons beyond technological limitations for a chatbot to not deliver the promised business value: -
- Approaching AI-based chatbot projects as any other conventional app projects: - A conversational AI application development has to be visualized as a journey with multiple iterations where AI its’re-trained’ by looking at usage data. While even conventional applications need a good amount of user feedback to be incorporated for it to succeed, the latter, as expressed by actual usage, is an integral part of AI-based chatbot development. Since there is limited scope for visual cues that can be given to a user in text-based chatbots (and no scope at all on voice bots), it is imperative to understand how the user intends to use it as opposed to how you want them to use it, which brings me to my second point
- Designing a bot as per the ‘product owner’s’ requirement as opposed to ‘user's’ requirement: - While exceptional product owners do keep their ears closer to the customer's voice, it cannot beat the insights that can be derived from actually monitoring bot usage and then modifying it accordingly. While product owners can decide on the initial scope for the bot, re-training and enhancement should be done keeping in mind the actual usage.
- Expecting customers to follow the flow that you designed: - Again, as the nature of interface makes it very open for users to take any path, they are more likely than not to not take your happy path.
- Not selecting the right use cases: - Implementing a toy bot is easy, but production-grade chatbots are challenging. Also, they cost money. Hence, deciding the right use cases that deliver demonstrable value very quickly and against pre-defined KPIs is essential. It is equally important to monitor and tweak the bot so that you hit the KPI (and use the outcome to convince management to sustain the funding on your bot project).
- Making it all about conversational AI only: - Last but not the least, treating Conversational AI as a standalone technology wonder that has an answer to every problem. The failure to embed other emerging technologies including other applications of AI like Computer Vision, Image Processing, Augmented Reality and others as part of your chatbot solution limited the effectiveness of chatbots in achieving the target they had set out to achieve. Let's have a look at how this one single factor, when addressed, can disproportionately impact the future of chatbots.
Why they will succeed when they have already failed?
Chatbot development should never have been only about application of conversational AI alone. If we apply this technology as a standalone, there is always a good chance that its utility will remain limited. However, when we augment Chatbots with other emerging technologies, sound processes and embrace them as an integral part of our larger digital ecosystem, that is when we will start seeing the true value that can be unlocked. We will talk more about these themes in upcoming parts of this blog, where we will see how different areas of business (like sales, marketing, service, operations etc.) need to approach their conversational AI strategy, and the key aspects to be considered in order to apply this extremely valuable tool in a manner that it delivers tangible business benefits.
There are several lessons to be learned in the last five decades of advancements in chatbots. From technological limitations to faulty execution to lack of persistence, there have been many reasons as to why chatbots have failed. While advancements in specific technologies including conversational AI will happen at an even faster pace, businesses that are able to orchestrate the interplay between a combination of multiple emerging technologies to augment conversational AI in order to solve real-life problems, will be better poised to achieve their intended business objectives. Whether it is to earn more revenue, decrease operational costs or provide a better experience to their customers and employees, conversational applications will live up to the promised expectations.
In the previous part of this series, we saw how conversational AI, along with other cognitive services, can be applied to customer service functions to provide a better experience while keeping the TCO low. To access the blog, click here.
To know more about our capabilities in this space, do visit our conversation AI site. To apply our framework to identify and execute the right conversational AI strategy for your organization's unique needs, contact us.