Article

CUSO Posh Poised to Produce Better Bots for Credit Unions

customer service ai chatbot concept
Contributing Writer
member of Bellco Credit Union

8 minutes

DCU-incubated fintech brings MIT intellect and a commitment to helping smaller financial institutions keep pace.

If you think it’s time for someone to put serious brain power and artificial intelligence tools into producing better chatbots and robotic interactive voice response for credit unions, there’s good news.

Two bright graduate students, Karan Kashyap and Matt McEachern, were working in the MIT AI Lab in 2017 with a professor who helped to create virtual assistants like Apple’s Siri and Google’s Alexa. They graduated from the lab in 2018 ready to launch a commercial fintech that would specialize in conversational AI—that is, helping humans have productive conversations with robots.

From the many possibilities, they chose to focus on banking, credit unions in particular, organizing their company, Posh Technologies, as a credit union service organization with 16 credit union investors/owners. With that focus, Posh bots key on banking situations and recognize and use banking terms, including such common acronyms as ACH, NSF, ATM and ITM.

The company’s first three products, described on its website, are:

  • Its Website Bot sits on a credit union’s website and automates answers to simple text questions such as branch hours, ATM locations and routing numbers. It is focused on frequently asked questions, not integrated into back-end systems, and easy to install, with plug-ins for live chat handoff and scheduling appointments.
  • Its Banking Bot sits on online/mobile banking applications and integrates with application programming interfaces and core or back-end systems to enable workflows for doing such things as checking balances, making payments, transferring funds, blocking and unblocking cards and providing travel notice. The conversation is in text, but a graphic or recorded video can be attached. Banking Bot can also personalize responses to authenticated members and offer targeted marketing messages.
  • Its Phone Bot channels incoming calls away from clunky “press or say 1” scripts to spoken conversations with an Alexa-like voice, all without queues or wait times. Posh’s Phone Bot authenticates callers, answers questions, executes banking workflows and intelligently escalates to human contact center reps when necessary, the company says. Posh Banking Bot activities allow phone/voice conversations, it says. The bot can converse in English and Spanish now and will soon handle French and German.

Posh had a credit union focus even in the MIT lab days because Kashyap and McEachern thought small and mid-sized CUs were falling behind the megabanks in this critical tech. Through a personal connection, they pitched it to a D.C.-area credit union and got an encouraging response. They pitched it to Cambridge neighbor DCU and got an invitation to get more support through the DCU FinTech incubator.

The name, “Posh,” came early in the development, partly because they intended to produce an elite product and partly because, in the 1800s, it was an acronym for “Port Out, Starboard Home,” the best seats on steamers travelling between England and India.

The Posh team spent the first year of the company’s two-and-a-half-year existence building the technology. Then it started signing up customers in 2019, selling automation of level 1 and level 2 contact centers requests, Kashyap reports. Now it is focused, he says, on personal financial management and broader use cases for conversational experiences throughout the member’s lifecycle. The company’s staff of 30 includes four salespeople; the rest focus on products, he says.

The arrival of Posh is a significant event in the CU technology world, suggests Vasilios Roussos, managing director of the Boston-based DCU FinTech Innovation Center. Using AI to train robots that can interact with people is a hot market with lots of vendors backed with lots of capital providing lots of innovation and competition, he observes. What makes Posh stand out is its platform and its focus. It has the advantage of the intellect of its organizers and its MIT AI Lab and DCU FinTech pedigree. “They have the most compelling tech team out here,” he notes, “and the reputation to recruit a powerful workforce.”
           
Plus, Posh has a keen focus on credit unions. CUs are providing the data for the AI to process, the domain expertise, the capital through a CUSO, and most of the client list, Roussos notes. AI learns what it focuses on, he summarizes, and Posh focuses on CUs like no other provider. “They’re hitting it out of the park,” he says.

Ben Bauer, VP/marketing for $350 million Simplicity Credit Union, Marshfield, Wisconsin, first heard of Posh at a marketing event and liked the demo. But his CU only decided to work with after going through a formal selection process that reviewed multiple vendors.

“We have a vendor management process that relies on due diligence and reference checks,” Bauer observes. “It can be tempting to fall in love with shiny new solutions, but the research and vetting are very important.”

Posh won because it had the best functionality, because it was CU-owned, because its pricing was competitive, and because its people were vibrant and efficient. “They cut out the fluff in their presentations. They were very respectful of our time,” he recalls. “They earned our business for sure.”

$3.4 billion Truliant Federal Credit Union, Winston-Salem, North Carolina, is one of the 16 credit unions that invested in the CUSO. It expects to implement Posh’s AI-based technology solutions to enhance its member experience, says Sandeep Uthra, chief information officer. Uthra heard the buzz about Posh through his CIO network and liked what he was hearing. The rigor of the MIT AI lab was a plus, he says.
 
So was the company’s CU focus. “They’ve proven that they can work well with credit unions,” Uthra points out. “They understand our culture and that we are grounded in member service, which they can help us achieve faster. Working with a partner that has the AI expertise makes sense.”

At another large credit union where the Posh technology essentially runs the telephone banking service, the bot has up to a 97% containment rate, reports Posh CEO Kashyap. This means only about 3% of interactions are escalated to a live rep for members reaching out to this phone system. A touchpad is still offered, but roughly four out of five callers go with an entirely voice-driven experience, he adds.

Even if there is escalation, Kashyap points out, it’s a handoff, not a restart. “We can pass on the call intelligently to the best prepared available contact center rep,” he says, through the credit union’s telephony system for voice or its live chat service for text. On escalated calls, the bot can save up to a minute and a half of an agent’s time by authenticating a member, he says, noting that the bot can use voice recognition as part of two-factor authentication.

The containment numbers Posh reports are high because most of the calls are for simple, predictable requests. No bot is ready to offer mortgage refinancing activity or investment guidance today. With AI, however, the potential to learn and broaden the conversation is real.

“Where AI fails, we analyze the reasons,” Kashyap explains, “find the pain points and knowledge gaps and keep making improvements and increasing resolution rates.” The “we” here is mostly Posh, but credit unions can train the bot to some extent without getting into code, he says.

Credit unions are valuable to Posh beyond being customers and investors; they are a major source of fodder upon which to advance the AI.

“They supply us with tons of data,” Kashyap says, “in the form of chat and call transcripts that can be used to train and build our AI ground up, using comprehensive credit-union-specific data. That domain expertise helps us anticipate the questions credit union members will ask and program the AI to find useful answers.”

When escalation isn’t available, Posh can create a support ticket within its chatbot and integrate with any ticketing systems a credit union uses, Kashyap explains. The handoff comes with context, so a human picks up and continues the conversation and members don’t have to repeat themselves, he notes.

Bots can’t do everything, Kashyap concedes, but the list of what they can do is growing.  

“We do balances and check orders, travel notices, card activation and deactivation,” he says. “A member can say, ‘Pay my bill on the 19th’ and the bot will do it. It can be told to pay that bill on that date every month. It can set up a transfer between accounts on a set date. It can accept ACH transfers from other banks or credit unions.” Transfers out of the CU are feasible, he notes, but most CUs want a person to see any transaction involving money leaving the CU.

When members want to refinance a mortgage or reinvest the proceeds of a maturing certificate of deposit, the bot will refer them to a human, Kashyap explains. “We support the financial institution’s logic. Whatever they want to allow, we will execute, up to the limits of what a bot can do.” Financial counseling could soon be within those limits.

The chatbot is a radical agent of change that will revolutionize routine communication, Kashyap asserts. More importantly, it will revolutionize marketing outreach once the bot has had time to be nourished with a data-rich AI diet. Robotic marketing sounds pushy, but it doesn’t have to be, he insists.

“Our engineering team and our CU clients are always looking at where to draw the line and how often to send messages. The bot can learn what works and what doesn’t with each individual.”

Richard H. Gamble writes from Grand Junction, Colorado.

CUES Learning Portal