Brian Smith, Sr Product Manager, Intuit, shares his research into the top challenges faced by conversational AI teams in 2022.
Last year, I wrote my first story on Medium focused on the fundamentals and principles in conversation design. This emerging field has seen tremendous growth over the last couple of years. Many tools for creating and designing digital assistants have become accessible with low barriers of entry. We saw hundreds of digital assistants emerge when industries were impacted by COVID-19 in 2020.
But one aspect of conversation design that many do not talk about are the challenges faced today. Not just from conversation designers but product managers, data scientists and machine learning engineers. Maybe there is limited access to the right data for personalization or building models, or there’s one person tasked with executing all aspects – from strategy to new capabilities and design. Or maybe the idea of conversational AI is still in early stages of getting senior leadership bought in.
Whatever it is, I was curious to know if it was just me, or if others were experiencing similar challenges. So I asked people in the field. Here’s what I learned.
Not having standards or adapting quickly will drive slow growth
Regardless of size, vertical or channel, a company must recognize the importance in creating standards across processes, principles and teams – especially in design. Without this, a digital assistant could have many tones of voice, features being used in the wrong way or disconnects across digital platforms.
“Most [companies] are starting to realize the opportunity of combining IVR, chat and assistant teams together (let alone standardizing the experiences), but are struggling to do so because they’ve historically been in silos.” [Teddie Straughn, Voiceflow]
There’s not only a need to combine teams and create shared goals that are centralized, but also create design processes and patterns to help with identifying the right use case. Oftentimes, teams know WHAT customers need help with but they don’t know HOW to break it down for implementation. The experience should dictate the tech. Not the other way around. And we should always make data-informed decisions.
Design is not a priority and leaders often don’t understand the process
With chat and voice design, there are additional steps in making sure we build out an experience that guides the customer down a happy path – or at least 80% of our customers down a happy path. If a user gets off track, teams must prioritize how they will design fallback experiences or ways to get them back in.
Many companies today have not prioritized conversation design. And if they have, the process for designing these interactions is not very mature. When words and text are the interface, we must prioritize new practices and design roles to support the scale with these platforms.
“I’m the only person working on this effort.”
I can resonate with this one. When I first started in conversational AI, I was responsible for setting the strategy, understanding how we would integrate into existing products, build new capabilities, design conversations and drive technology integrations.
“Making big changes is a big undertaking that can feel like a SUPER slow process as a design team of one.” [Hillary Black, Mav] If one person is responsible for multiple pieces of the process, or all of it, it can make it difficult to know where to focus time and effort.
Companies must realize the importance of identifying the right types of roles to drive diverse perspectives, even if it’s done cross-functionally. There is a need to build teams with backgrounds in data science, linguistics, content writing and product management to build scalable experiences. The more you invest, the better you will become.
Organization maturity plays a key role in CUI
The maturity of an organization will drive a company’s ability to scale chat and voice interactions. If the organization is less mature, they may focus more on gaining leadership buy-in. Leaders may also underestimate the complexity of these experiences. In larger, more mature organizations, the existing tech stack or tools may not enable advanced transactional use cases and competing priorities can cause delays in execution. Oftentimes, larger organizations may not have reusable components or APIs, and data is not easily accessible.
“The more AI becomes the face of the business, the more you’ll need to have technology that can enable it to do its job.” [Kane Simms, VUX World] Helping sell the concept is one thing, but keeping it top of mind with senior leadership and driving modernization across technology is the other
Regardless of organization maturity, teams must take small steps to outline a path that includes MVP. Then you can start to scale from there. If a company is pushing for task completion or transactional experiences but their tech stack causes limitations, how can we effectively scale CUI in a way that still shows impact. For example, capturing some of the information and then passing a customer off to a human to complete the task.
Conversation design is rapidly emerging and what we see today is only the beginning. While we’re still learning, growing and adapting to maturing technology and changes in customer expectations or behavior, we must address these challenges head on. If we work together in identifying commonalities across industries, companies and experiences, we can start to pave a clearer path to deliver more positive customer interactions with CUI.
This article was written by Brian Smith. Brian leads the conversational AI strategy and roadmap for digital assistants across all Intuit products – TurboTax, QuickBooks, Credit Karma, Mint and Mailchimp. Over the past decade, Brian has led teams in executing customer experiences for e-commerce, digital products, insurance and software. You can read more from Brian on his Medium blog and learn more about his experience at brnsmth.com.