The Myth of the Pre-Built Conversational AI Use Case

Plenty of vendors will tell you otherwise, but pre-builds are of little use when it comes to maximizing conversational AI

Download the PDF here.

With growth in the conversational AI marketplace continuing to skyrocket, it’s no surprise that everyone wants to take advantage of the associated technologies  in meaningful ways.Pre-built point solutions cater to a desire for easy implementation, but the hard truth is that conversational AI isn’t easy. In fact, it’s so complex that no matter how customizable a point-solution appears to be, organizations quickly outgrow them and need functionalities their vendor can’t offer. Those organizations are stuck waiting for outside development cycles to put a solution into action, sacrificing valuable time and potential market share.

Pre-built automations represent one of the great myths about conversational AI—the myth that it’s something that can simply be layered over top of your existing organization structure. That’s not how conversational AI works. Getting it right requires an organization to restructure itself around the sequencing of multiple technologies, which requires an open and flexible ecosystem. The false promises of pre-built conversational AI solutions can cause organizations to waste precious time in an undertaking where speed is critical.

Pre-Built Use Cases Fall Short

The conversational AI marketplace has made massive leaps in recent years, thanks to advancements in Natural Language Processing/Understanding (NLP/NLU), low- to no-code development tools, and a glut of viable products in the market. Hyperscalers like Amazon, Microsoft, Google, and IBM, as well as cloud and enterprise solution providers, have brought increased attention and investment to the space. As development tools have become more accessible and licensing costs have gone down, the barriers to entry are lower than they’ve ever been, something reflected in industry adoption rates. 

Pre-built, downloadable use cases have played a significant role in market growth, as they can be used out-of-the-box with little to no configuration. According to Gartner’s projections from 2020, chatbots will see more than a 100% increase in adoption rates over the next two to five years and represent the leading AI use cases being applied in enterprise settings today. But just because an enterprise adopts chatbots, that doesn’t mean their customers will. Gartner also forecasts that despite these advancements in functionality and affordability, 90% of today’s chatbots will be discarded over the next three years. Pre-built use cases will be a major contributor to this abandonment for the simple reason that if they were really gateways to success, there would be many more success stories out in the marketplace.

The Reality of Conversational AI

The reality is that the biggest success stories in Conversational AI–including those touted on every vendor’s website–were not based on pre-built templates, flows, or intent models. In successful use cases, the user experience was carefully designed, implemented, tested, and iterated over time to drive the outcomes being touted.

User experience is the hardest element to get right, but it delivers the biggest ROI. Consider the advantages that brands like Amex, Apple, Zappos, and Ritz-Carlton have enjoyed in their highly commoditized markets (credit cards, computers, shoes, and hotels). These leaders in each segment differentiate by focusing on customer experience.

With hotels, the five-star experience takes the standard process of booking, checking in, staying, and checking out and makes each step along the way unique, and personalized wherever possible. A welcome email one week before you arrive asks for details about your visit and if you have any special requests. The concierge wishes you a happy birthday at check-in and gives you directions to the museum you came to visit. There’s a handwritten note and a piece of birthday cake waiting for you in your room. At check-out the concierge invites you to enjoy a complimentary meal in the hotel restaurant on your next visit.

The components themselves (booking, check-in, etc.) are well-worn commodities. Customizing those components using available data to make them memorable and rewarding is where the ROI lies. This is true for conversational AI as well. There’s nothing more worn in than conversation as a means for communicating information, but even if a bot can understand a vast multitude of words and phrases, if it’s not part of an experience that uses conversation to leverage technology intelligently behind-the-scenes, it’s not worth much. Wildly successful deployments like Bank of America’s virtual agent, Erica, succeed because they put user experience at the fore. 

Here’s how Bank of America CIO Aditya Bhasin described their goals: “When we design or enhance a product or service, including Erica, we are guided by a deep understanding of our clients’ needs and what capabilities will help them to lead better financial lives. We strive to ensure our client journeys are smart and seamless, no matter where or how they choose to interact with us.”

As users increasingly turned to these online resources during the pandemic, Bank of America gave Erica the ability to understand more than 60,000 phrases and questions related to the virus, including answers about federal stimulus programs. According to Voicebot, Erica can now respond to more than a million unique financial questions, four times the initial number at launch in 2018. 

While other banks could certainly make some headway using a pre-built solution based on Erica, each one would reach a point where the template’s limitations would begin constricting any meaningful growth (hence the 90% abandonment rate Garter is predicting). Conversational AI presents massive opportunities for delivering personalized experiences that fuel user adoption, but capturing that power requires the flexibility to sequence multiple technologies and iterate on experiences quickly and efficiently. It’s critical to constantly be testing and customizing your solutions, which is something pre-built use cases often can’t accommodate. The way to succeed with these emergent technologies is by creating an environment where you can build your own use cases—as needed and often.

How to Succeed with Conversational AI

A winning strategy for implementing conversational AI starts small—often internally—with the automation of simple tasks and processes. These initial automations serve as the building blocks for future automations that continue to grow in complexity and the ability to deliver more personalized experiences. 

Famed Harvard Business School strategist Clayton Christensen advises that a company’s best source of competitive advantage is to focus on the least commoditized part of its value stream.  

In conversational systems the most highly commoditized components based on pervasiveness of their availability in the market—and level of competition on feature parity—include NLU models, pre-built processes, and intent libraries. The least commoditized part of your value stream is the way that you can orchestrate these kinds of technologies to provide users with increasingly rewarding experiences.

Based on Gartner’s findings, the use and adoption of conversational AI is essential to driving ROI. Solutions that don’t get adopted get abandoned. When strategized and executed properly, user experience drives adoption up and gives organizations a clear competitive advantage. New entrants to conversational AI should expect to prioritize their investment of time and resources around the creation and implementation of user experience.

The commoditized NLU models, conversation flows, and intent libraries are a valuable starting point, but in order to drive usability, adoption, and ROI, it’s essential to make a good first impression on your users. Success with conversational AI is all about user experience, and your top priority is to deliver solutions that drive adoption and loyalty.

The strategy of starting internally allows you to identify tasks and processes that will benefit your team members. The very team members who will benefit from the automations are active participants in their design. Automating internally allows you to test solutions and iterate on them in a controlled setting. The faster you can build, test, iterate, and deploy, the better. Creating an environment that is more agile than agile is essential. 

It’s fine to kickstart a project with existing models but a kickstart is all you’re going to get, which is somewhat minor in terms of moving fast enough. You can save more time and get exactly what your company needs by training your model on actual user data, collected in real-time. Success with conversational AI is hugely dependent on speed—on how quickly you can create and iterate. While pre-built models might get out across the starting line, there are feasible ways to accelerate much faster while creating solutions that are exactly what you need. 

Your best shot at putting conversational AI to work for your organization is to find a flexible platform that lets you integrate any tool, NLP/NLU engine, and third-party service you need—basically letting you pick whatever puzzle piece best suits your needs from a highly commoditized marketplace. Having control over the tools and technology at your disposal lets you prioritize speed and flexibility.

Speed and flexibility allows you to orchestrate the basic building blocks—think of them as skills—that you’ve polished internally in new ways to benefit your customers. The skills that you’ve created internally represent your organization and its unique people, and because your team learned how to create new skills and iterate on existing skills a user-centric culture has been established.

Conclusion

Pre-built use cases can’t compete with organizations that are leveraging flexible tools to create custom automations—both internally and externally—on a daily basis. Your best hope with pre-built use cases is that you can somehow integrate them into an ecosystem built on a flexible platform. The much faster path is to start building your own automations and skills in a way that engenders design thinking across your organization. Ultimately, it’s the orchestration of skills into personalized experiences for customers and employees that drives value. The sooner you can start doing that, the better.

The only way to exploit the massive advantages conversational AI presents is by using speed and flexibility to create meaningful, personalized experiences. If your organization isn’t doing that, you’re not really in the race. And across all industries, this is a race that will be run and won in a flash.

 

Download the PDF here.

Comments are closed

ukraine flag

We stand with Ukraine and our team members from Ukraine.

Here are ways you can help