Enterprises customize Clinc’s conversational AI to connect with their customers


Conversational AI platform company Clinc has experienced significant product pivots since it was founded by a group of University of Michigan computer scientists in 2015. It has also undergone leadership changes and investigations into its former CEO’s alleged pattern of sexual harassment. Now the company is carving out a niche within the financial services market, where its customizable conversational AI platform streamlines customer queries and collects related data for enterprises.

Clinc used to work across multiple verticals, with early clients including Wendy’s, Chick-fil-A, Stryker, and Ford. Last year, the company made a more deliberate pivot toward the financial services market with its virtual voice assistant, Finie, and concurrent developer platform. Its personalized conversational AI uses one model for multiple channels, helping enterprises like banks answer customer questions and understand customer data in less time.

In an interview with VentureBeat, Clinc product lead Matthew Taylor said the startup had limited bandwidth and decided to focus on fintech following early success with the industry. According to Taylor, the pandemic also accelerated Clinc’s development. “COVID really increased the urgency of being able to allow people to get answers to the retail banking questions without having to visit the branch,” he said.

He said Clinc allowed people to more easily cancel credit cards or go through their spending histories without in-person contact. And Clinc’s enterprise clients, like the Bank of America and U.S. Bank, were able to drive more revenue by decreasing call volume to human agents, providing agents with the AI-collected data for background, and using the mobile app marketplace to upsell and cross-sell different financial products based on this personalized data.

Clinc’s out-of-the-box financial assistant Finie — short for “Financial Genie” — answers customer questions like, “Can you cancel my Chase Sapphire card?” or “How much have I spent at restaurants in the past six months?” For the former request, Finie would hit the bank’s API to lock the card and then send the user a response in a payload to confirm the change.

Queries are coded and addressed according to an enterprise’s particular business logic. For example, a customer’s voice-suggested address change might present in lowercase unless an enterprise adds intent to capitalize it. Enterprises also custom-define limits, like caps for the amount of funds a user can withdraw at once.

Taylor claims Clinc differentiates itself by training its models on semantic structures of data. “The way we built our tech is not to look for keywords or to look at the grammar or syntax,” he said. “We have an integration into our crowdsourcing platform that allows us to collect a lot of data very quickly that can give this AI model the ways people would talk in everyday life.” The platform maps slots based on semantics by embedding space for open-ended slots.

For example, Taylor said users can ask “How much dough do I have” instead of “What’s my balance?” Clinc could then use NLU to process the colloquial language as it would a more formal version based on a user’s sentence structure. But it’s uncertain whether context clues could capture phrases with more significant deviations from dictionary-defined words.

Finie is low-code, and Clinc’s corresponding SaaS platform, architected with AWS, customizes its responses. Enterprise IT teams either prototype the AI solution themselves or ask Clinc to do so based on their specifications.

The Clinc platform starts with a state graph that operates like the AI model’s “brain.” The graph branches out into purple lines detailing the model’s particular competencies, intents, and parameters for transactional data. Developers might add loops for context retention between the nodes here, like a transition connecting different questions about spending so that variables can carry from one conversation to another.

According to Taylor, the platform’s SVP, or entity extraction, enables better decision trees with crowdsourced data from end-user interactions and Amazon Mechanical Turk. Its analytics suite checks the number of queries per user per day and the competency breakdown for out-of-scope responses and latency with RESTful APIs. Clinc also collects detailed information on users, tracking queries to their individual IDs.

Next, Clinc is looking to expand its scale of deployment with companies such as Pfizer, which owns over a thousand financial institutions and has introduced Finie’s prebuilt capabilities into its mobile applications. Clinc also launched Finie with U.S. Bank last July and is looking to grow with non-U.S. banks in Turkey and Singapore. “The financial services industry has been known for adopting new technology the quickest amongst most other industries,” Taylor said.


VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform
  • networking features, and more

Become a member

Products You May Like

Leave a Reply

Your email address will not be published. Required fields are marked *