Debunking the AI, Voice, and Chatbot Hype in Banking

The banking press (e.g., The Financial Brand, American Banker, Tearsheet) is full of stories about AI, voice, and chatbots these days. Authors of the articles cite studies regarding consumers' use of voice assistants like Siri and Alexa, and use those statistics to preach why banks need to: 1) Deploy chatbots; 2) Develop voice strategies; and/or 3) Implement artificial intelligence (AI).

People in general--and senior executives, specifically--don't like to admit publicly that they're confused about something, or that they don't understand something, especially when they're being told that that "something" is critical to the future of their business.

So let me say it for them: Many senior bank and credit union executives are confused about voice, chatbots, and AI. They're confused because much of what's being written about those topics is confusing.

AI Isn’t One Technology

Feeding the hype is the indiscriminate use of the term AI, which is not a single technology. Different types of technologies can be classified as AI, including:

  1. Computer vision, which can recognize objects;
  2. Speech recognition/synthesis, which can turn sounds into words;
  3. Natural language processing (NLP), which can extract meaning from language;
  4. Knowledge representation, which can sort information;
  5. Reasoning, which can combine data to reach conclusions; and
  6. Planning, which can sequence actions to achieve a goal.

At a broad level, these technologies can be slotted into two categories: 1) Input (vision, speech recognition, NLP), and 2) Output (knowledge representation, reasoning, planning).

One problem I have with all the breathless claims of the impact of AI, voice, and chatbots is that they focus on the input technologies, and not the output technologies. And I simply can't see how that will have a big impact on the economics and delivery of banking.

Let me put it another way: Improving your understanding of the "problem" is helpful, but it's insufficient if you can't improve the resolution of that problem.

The Chatbot Conundrum

Focusing more narrowly on chatbots, I see a conundrum brewing. To explain the conundrum, I'm going to make three statements or assertions (and I will back up some of those statements up with statistics).

Statement #1: Consumers' expected benefit from chatbots is convenience, not advice. 

Two-thirds of consumers expect 24-hour service to be the benefit of using chatbots, with a little more than half expecting instant response to simple questions. Only a little more than a third expect the benefit to be detailed/expert answers to complex questions.

To further support this view, check out the results of a study conducted by PointSource. They asked consumers when they would prefer to interact with a chatbot versus a human for a range of interactions (mostly insurance-related). The result of three types of interactions underscores my point. For "simple" things like looking for product information, consumers are pretty much ambivalent between chatbots and humans. But for more complex interactions like resolving problems and getting advice, the preference for humans is overwhelming.

I can hear my debaters' counter: "Ron, we're early in this evolution. The capabilities to provide detailed, expert answers to complex questions will be developed and will come in time."

Not necessarily.

Statement #2: Banks don't have the data needed to provide advice.

There's a growing chorus of folks writing articles about AI who rightfully point out that the key to success with AI isn't the technology, but the data.

Successful AI deployment is predicated on a feedback loop: Input technologies can be used to identify/understand the problem, output technologies can be used to provide advice, and the result of that advice is captured and fed back in to improve the process over time.

Two problems here: 1) If chatbots are predominantly used to provide answers to simple questions, how is a bank ever going to gather the data needed to address the more complex, advice-oriented questions? and 2) How is a bank going to capture the results of the advice provided?

I can hear the debaters' counter: "It's coming, we're just not there yet."

It might not matter.

Statement #3: Consumers don't care that much about advice.

Even if my debaters are correct on all fronts, there's a fact they're overlooking: Many (i.e., the majority of) consumers just don't care about getting advice about their financial lives--from their banks, or from anywhere. The advice impact is way overstated in the press.

Cornerstone recently surveyed 2,015 US consumers (between the ages of 21 and 72, with a checking account and a smartphone), and asked them: "If you were looking for a new checking account, which three features or factors would be most important to your decision?"

"Best capabilities/tools to help manage financial life" was mentioned by a small minority of consumers--just 16% of Millennials. So let's say you develop and deploy the greatest chatbot ever seen by mankind (or robotkind), and it provides the greatest advice ever provided. Big deal--few consumers will care.

If you were looking for a new checking account, which three features or factors would be most important to your decision? 

Young Millennial (1988-1996) (n=316)

Old Millennial (1980-1987) (n=469)Gen X (1964-1979) (n=709)

Boomer (1945-1963) (n=521)

Lowest monthly fee




Best overall value for the money




Best online and mobile banking tools




Best rewards program




Most convenient bank locations




Best combined debit and credit card rewards




Best capabilities/tools to help manage financial life




Best mobile payment tools and capabilities




Best in-branch experience




Best P2P payment tools




Source: Cornerstone Advisors survey of 2,015 US consumers, Q3 2017

Overcoming The Chatbot Conundrum

There's are two more things my debaters will get wrong: 1) I don't have my head in the sand, and 2) I am not dismissing chatbots, voice, and AI technologies out of hand.

My point is this: The AI/chatbot "field of dreams" is not true. You can't just "build it and they will change"--"they" being consumers, and "change" being their attitudes and behaviors.

For banks and credit unions to succeed with AI/chatbots/voice, a more fundamental change in the attitudes consumers have towards the banks and credit unions they do business with--and a fundamentally different type of relationship--is required. A relationship built on a value proposition of advice, not convenience. 

We are very far away from that in banking today.

Simply deploying AI technologies will not make that happen. The articles I read that proclaim the "transformative, disruptive, and game-changing" nature of AI do not address the conundrum. That's why the impact of these technologies--as it relates to advice, not routine customer service, fraud detection, or credit underwriting--is 10 to 20 years out, not two to three.

Ron Shevlin
Director of Research
Cornerstone Advisors