Around five years ago, the promise of chatbots and conversational interface (CI) vendors to significantly enhance operations and customer satisfaction led to a bubble in investment from financial institutions (FIs) and many observers — including us at Emerj — to slap vendors with the “hype” label.
The marketing departments at these FIs still sometimes parrot vendors in touting new “cutting-edge” chatbots, hoping to recuperate some of their sizable investment. In reality, many C-suiters were busy allocating money to less risky use cases, like compliance and fraud prevention.
On a recent episode of Emerj’s AI in Financial Services podcast, we spoke with Aigo.ai Chief Scientist Peter Voss about what differentiates chatbots from conversational AI technology with broader application across industries. In a dialogue with Emerj CEO Daniel Faggella on the podcast, Peter brings the chatbot banter back to reality and offers the audience three clear, even-handed insights about the much-discussed technology:
- Successful CAI adoption begins with “keeping the end in mind”: Successful and sustainable adoption of cognitive AI demands keeping the continuous function of processes in focus by planning the implementation of the next iteration of the technology as you’re strategizing the current iteration.
- Managing up to overcome silos for truly organization-wide transformation through CAI: C-suite leadership must be kept abreast of long-term strategizing from the beginning to integrate the organization-wide transformation that true long-term conversational AI adoption entails.
Listen to the full episode below:
Expertise: Cognitive AI, Startups, Natural Language Processing, SaaS, Business Strategy
Brief Recognition: Peter is a serial entrepreneur who has developed one of the leading conversational agents in the market, Aigo. Over the past two decades, Peter has founded several other AI-centric enterprises, serving as Chief Scientist or equivalent roles throughout nearly all.
Begin CAI Adoption Strategy with “The End in Mind”
Peter describes two critical features of truly conversational AI as a long-term transformative project for financial services firms :
- Close examination and planning of an integrated data structure, similar to a data warehouse. This way, the bots can service a customer across request types, divisions, and organizational silos.
- Gaining sponsorship from the highest level of the company for the organization-wide project of integrating such capabilities.
Customer-facing applications are still the most viable, promising use cases for chatbots in banking. Peter provides two main reasons for this:
- Better customer service that is more scalable and consistent and at a lower cost.
- More engaging technology that deepens customer relationships and increases customer “stickiness.”
Peter gives a caveat for those hoping to achieve the two points above: you need worthwhile technology and a strategic vision for where conversational AI can serve as an early project for transforming the fundamentals of your business.
Conversations between early AI adoption teams and FI leadership regarding chatbots often start with automating part of a call center or some other process along the customer journey. Among potentially viable conversational AI use cases cited by Peter that begin with automating repetitive processes:
- Renewing insurance policies
- Changing contact information (e.g., name and address)
- Order status updates
- Money transfers
- FAQ chatbot
Peter further states that these use cases represent how conversational AI and chatbots are usually the first step FIs can take to automate the vast majority of transactions — around 70 – 80% — given the right technology.
Plenty of opportunities exist for business leaders seeking to automate administrative workflows, particularly those of the customer-facing variety, through the use of chatbots. Peter recommends the following insights based on his experience with business leaders starting such a journey:
- Be sure to carefully analyze your agents’ workflows and determine which could be safely automated with a chatbot.
- When performing your workflow analysis, consider the sophistication needed to automate. What does leadership need to be made aware of for their expectations? To help bring in more help across departments to make implementing automation easier?
- Generally, such workflows are not ideal initial projects as they require deep data science expertise and considerable resources. Is the process a source of competitive advantage – e.g., proprietarily-held data that no other company possesses? Do you own intellectual property (e.g., patents) that will be affected?
As Peter describes these recommendations, he points to the importance of choosing an initial chatbot application that sets a tone for future iterations. In the process, he summarizes a helpful mantra for the successful long-term conversational AI adoption, or as he puts it, “begin with the end in mind.”
As an example, Peter describes a customer calling for a password reset. In this notional scenario, the agent is made privy to the reason for the password reset: to update the phone number on file. A proper chatbot should be able to handle both requests without routing the customer elsewhere.
Part of this integration is being omnichannel. “If somebody’s called in on a voice call and spoke to an agent, you want the bot to be able to utilize that information.”
“You want to be able to have data uniformity [and] growth path available, even if the initial implementation might be limited to one function,” Peter tells the Emerj audience. “You want to know that you are building the foundation so that the next chatbot … has fluid, seamless interaction with the customer.”
In other words, genuinely long-term conversational AI strategies must take into consideration the following:
- Integrating conversational interfaces with value-adding business processes.
- How functions will transition into future iterations of interfaces and other technologies.
- Forgoing short-term, “fix it quick” solutions for those more scalable and impactful, even if they take longer to implement.
Too often, chatbots and other conversational agents are viewed as an accessory that can help make a single process easier. Such a view is shortsighted and ultimately limits the lifespan of the application and the potential for meaningful business impact.
Part of this disciplined planning involves integrating disparate data systems into a comprehensive whole. Or, at least, having a repository wherein these data can be pulled in and manipulated according to the customer’s needs.
While such a system will likely require the expertise of a data scientist team, institutions can, to a great degree, streamline their agents’ workflow, reduce time to resolution, and decrease associated costs.
For more on how short-term chatbot projects fail to leverage into a transformation of customer experiences in banking, click on our discussion of the three phases of CX transformation with Kimberlee West of Uniphore.
Managing Up to Overcome Silos
A theme Peter returns to throughout the conversation is the importance of obtaining C-suite buy-in for AI initiatives in general and chatbot iterations in particular.
Conversational AI projects trend toward short-term “bandaid” fixes with negative ROI, especially when they are secluded to only one, likely siloed, business division. Ultimately if your CAI project isn’t helping bridge information and communications gaps across the organization, it will never achieve transformative capacities that define ROI for these technologies.
Peter emphasizes that bridging those communications and data gaps begins with keeping business leadership involved as a stakeholder from the very beginning of the project, even if the initial beachhead for the project is small in scope.
“Conversational AI needs to have sponsorship from the highest level in the company. Otherwise, it will be, ‘Oh, here’s this little problem we want to solve. Let’s get a chatbot, spin it up, and see what happens.’ [C-suite buy-in helps] to really understand where you’re going, the vision of what you are trying to do.”– Aigo.AI CEO Peter Voss
Most importantly, the larger cross-organizational use cases and business-transforming capabilities of conversational AI must be part of the leadership conversation. In turn, C-suites must have the awareness to realize how far conversational AI can extend beyond customer interfaces.
Here, noncoding professionals can be an asset in translating how these technologies can’t be driven by IT and software professionals alone – but require input from subject matter experts across the organization. Peter emphasizes how essential the C-suite’s role is in organizing potential new stakeholders across silos and departments, but only if they are adequately engaged from the very start of long-term strategizing.