2023 Vertical Markets Spotlight: Speech Technology in Financial Services

When COVID-19 was in full swing, banking customers largely moved away from transactions in physical brick-and-mortar branch offices in favor of digital and virtual banking alternatives. Forrester Research noted in its 2022 “State of Digital Banking” report that 30 percent of U.S. adults with bank accounts checked their account balances using a smart speaker, such as Amazon Alexa or Google Home, last year.

And now, as the pandemic winds down, banks have not seen the huge return to in-person banking transactions that many had predicted.

Allied Market Research valued the worldwide market for phone-based banking technology at $984.6 million in 2021 and expects it to reach $3.7 billion by 2031, growing at a compound annual rate of 14.5 percent. The same firm found that the worldwide market for intelligent virtual assistant-based banking generated $1.3 billion in 2021 and is projected to reach $11.2 billion by 2031, growing at a compound annual rate of 24.5 percent.

The development of voice banking is predicted to move in several directions in the next few years. According to Intelligent Software Engineering, 18 million U.S. consumers have already tried paying by voice, and many banks and financial institutions offer this service to their clients.

Financial services firms today are being driven to speech technologies largely by the growing need to personalize services to stay competitive.

Financial services firms are also turning to speech to overcome contact center personnel shortages. They are staying true to interactive voice response systems and virtual assistants to authenticate customers and provide answers to basic customer questions, like account balance queries. Many more advanced uses could be coming in the next year as conversational and generative artificial intelligence technologies like ChatGPT start taking hold in the industry.

Already, bots are listening to both sides of customers’ conversations with their financial institutions to provide agents with real-time guidance, which is enhancing customer service, says Donna Fluss, founder and president of DMG Consulting.

But the real benefit is enabling customers to carry out some transactions without involving human agents at all. Technology vendor LumenVox reported recently that Security Service Federal Credit Union in San Antonio, Texas, reduced the number of calls being sent to agents from 26.2 percent to 24 percent after deploying a speech-enabled IVR. While that might not seem like much of a difference, for a company that receives 500,000 calls per month, that adds up to 11,000 calls each month or 132,000 per year that agents don’t have to handle.

The number of calls some banks, insurance companies, equities firms, and other financial services companies receive is nothing short of mind-boggling. Human agents are unable to handle the sheer volume, hence the need for speech-enabled and digital support, according to Daniel Ziv, vice president of global product strategy for speech and text analytics at Verint.

One Verint financial services customer, Poland-based BNP Paribas, receives an estimated 230,000 calls a month, according to Anastasia Novak, its speech analytics product owner and agile transformation leader in personal finance operations.

Beyond helping firms like BNP Paribas handle a large number of calls without human intervention, Verint solutions also provide automatic transcripts, summaries, and analyses of calls, according to Ziv, all of which are necessary to ensure regulatory compliance.

While agents can usually take notes themselves for regulatory record keeping and to provide context for interactions that need to be transferred to other agents or departments, it takes time for them to take good notes, and most are on the clock with other calls waiting, Ziv explains. “They will only have an abbreviated, short summary, which doesn’t give the other agent enough information.”

Summaries are indeed important for transferred interactions, but the regulatory nature of financial services also means complete transcripts are necessary to ensure that agents provided proper disclosures, asked the proper questions, obtained the necessary permissions, and followed all applicable data protection guidelines.

“Financial institutions have a compliance requirement; they might need to file a report, so they might need the full transcript,” Ziv explains. “But the summary is also important.”

The summary has to be accurate and complete enough to extract all of the key insights from the interaction, whether it’s the intent of the call, what was happening, what the agent promised, whether there is to be a follow-up call (or email or text), etc.

Another increasingly requested feature from financial services firms these days is automatic redaction of personally identifiable information to meet security standards, Ziv adds. Redaction typically relies on speech analytics to identify sensitive information, such as credit card or Social Security numbers, in recorded audio files or transcripts and then to either mask the data or remove it entirely.

Financial institutions cannot completely automate all transactions. Equities firms, for example, can offer online trading capabilities, but regulations don’t permit them to buy, sell, or trade assets based on instructions left via a voice recording.

Also helping financial services’ contact centers is voice biometrics. Not surprisingly, the financial services industry far outpaces other verticals in its adoption of voice biometrics, which is projected to increase from a $1.3 billion market worldwide in 2021 to $4.8 billion by 2028, a compound annual growth rate of 20.6 percent, according to the Insight Partners. The research firm Reports and Data expects the market to reach nearly $5.9 billion by 2028.

One of the main reasons for the recent voice biometrics surge is technological advances that have amped up voice biometrics’ accuracy rates, which average around 95 percent for most vendors.

Some of the world’s largest financial services providers, including HSBC, Wells Fargo, Barclays, and Santander, have been using speech-based biometrics for decades, but the technology has long been too expensive for smaller financial service firms to incorporate, according to Robert Wakefield-Carl, senior director of innovation architects at speech technology vendor TTEC Digital.

Costs are coming down, though, as vendors move their technologies to the cloud.

And amid sharp increases in incidents of fraud, consumers have been more comfortable with voice biometrics technology. In fact, roughly 80 percent of consumers have said that they trust modern biometrics to secure their financial accounts and would be happy for their bank to use it.

While voice biometrics, speech analytics, and other technologies are pretty advanced, some of the speech-enabled banking capabilities are still rudimentary because many firms still cling to the basic logic of a touchtone IVR rather than incorporating robust natural language understanding, Wakefield-Carl says.

So the voice prompts will ask what the customer wants to do, and can understand requests like “What’s my account balance?” or “I want to transfer money.” But for the latter one, there will need to be additional prompts. Most systems can’t yet handle a more complex request such as “I want to transfer $200 from checking to savings,” Wakefield-Carl explains.

However, AI is quickly advancing the types and numbers of interactions that financial services firms’ contact centers can handle without human interaction.

“Artificial intelligence has enabled rapid technical acceleration in the financial services industry, with many banks and credit unions onboarding virtual agents via chat or voice to take on and resolve customer traffic,” says Sam Danby, head of voice for Boost.ai. “With the kinds of furnishings and personalization end users have come to expect in 2023, institutions have begun implementing more complex and personalized service flows for virtual agents, using IVR and other speech technologies to allow customers to make transactions.”

Achieving the scale where every customer query can be addressed quickly and accurately might have been a challenge with legacy systems alone, but modern conversation AI integrates with multiple platforms to provide customers with a seamless experience as they navigate their queries, Danby adds. “The future of the contact center for financial services is one where customers have as much trust in a voice agent as they do their human counterparts. Much of the foundational work for this reality is already done and will be critical in an industry that sees increasing call volumes and spikes in traffic.”

Wakefield-Carl expects financial services firms to advance more contextually so that when a customer asks for a transfer of funds from one account to another, systems will understand the command without having to ask a series of follow-up questions.

ChatGPT is also starting to make its way into all speech verticals, according to Fluss. The financial services sector has one advantage in that firms in this vertical have massive amounts of data and extensive experience working with it. Though ChatGPT is primarily text-based, it can work with voice.

Fluss admits that “we are at the height of inflated expectations regarding generative AI, but the potential is phenomenal.”

With calls to the contact center showing no signs of slowing down, the growth of ChatGPT and the need for positive (sometimes as part of two-factor) verification continue to grow.

Fluss adds: “We’re going start to see more and more and better and better applications; the solutions are going to become more conversational, more able to handle more intense communications. This is a very, very exciting time within the industry.” 

Phillip Britt is a freelance writer based in the Chicago area. He can be reached at spenterprises1@comcast.net.