Conversational AI for Enterprise Applications

A few months ago, we wrote a piece on Some of the Latest Trends in Artificial Intelligence in which we noted that AI will have the greatest impact on “job functions where a majority of time is spent collecting and making sense of data.” At the top of that list are finance and insurance. When it comes to which applications of artificial intelligence are being used most by firms in 2019, survey says that chatbots top the list with 26% of firms saying they use this technology which we’re now calling “conversational artificial intelligence” or “conversational AI.”

In earlier days, we questioned the maturity of chatbot technology when the best technology on display at Facebook was a witty weather chatbot called Poncho. Today, that’s all changing. Conversational AI is now being used to decrease the need for John-in-Mumbai customer support agents with up to 90% of customer support inquiries being handled by chatbots in some cases. If the algorithms run into a problem, they just throw a human in the loop. Implementing conversational AI is as simple as choosing one of the many platforms available, and soon the term “chatbot” will go away because everyone will just use natural speech to engage with technology without even thinking about it. That ability to use natural speech is what one startup called Clinc is focusing on.

Clinc’s Academic Roots

Click for company websiteFounded in 2015, Michigan startup Clinc has taken in $59.8 million in funding, the vast majority of which was raised last month in the form of a $52 million Series B round. All that money is being spent on developing conversational AI algorithms that take a factor-based approach to understanding what people are saying. Taking into account wording, sentiment, intent, tone of voice, time of day, location, and relationships, can all provide insights as to what people are actually saying. It all started back in 2015 when the four founders published a paper which proposed a conversational AI platform that could compete with the likes of Google and Amazon.

Click to read the paper
The conversational AI paper that started it all

After the paper was published, there was “an explosion of interest” from all kinds of industries who wanted the technology commercialized for various applications. The four individuals highlighted above then founded Clinc with the mission to “use conversational AI to enable the access to knowledge across all types of commercial, industrial, business, and financial applications.” Fast forward to today and their clients include names like USAA, US Bank, and Barclays. An article by Forbes a few months ago talks about how successful Clinc has been so far:

Jason Mars, co-founder and CEO of Clinc, said conversational AI is just at the frontier of its potential. The company has 30 million people using its technology, including six million customers of Isbank in Turkey who ask about balances and transfer funds with Clinc.

Clinc’s AI is also being leveraged by one of the largest banks in the UK. The bank is powering its Facebook Messenger bot with Clinc’s AI and successfully handles over 25,000 queries per week at a 92 to 96 percent completion rate, which, if handled in the call center, would cost an average of $10 per query. This amounts to an estimated $10 million dollars in savings for the bank since implementing Clinc’s AI.

Voice can also be very useful in fraud cases, Mars said.

What stands out about Clinc is the academic foundation that underlies their technology, something that’s evident in the number of papers the founders have published and continue to publish. Last year, they published a paper on how to optimally use crowdsourcing for obtaining better training data for their algorithms. In the below example, you can see how Clinc might be able to easily obtain thousands of potential queries from fictional banking customers just by using crowdsourcing.

Crowdsourcing big data for training

The ability to crowdsource training data is something we talked about in our article on how Crowdsourced Big Data is Big Business for Appen. In exercises like this one, Clinc looks for outliers which they’ve defined in a paper as errors: mistakes in the data that are counterproductive, or unique: informative samples that improve model robustness. It’s the latter category which makes the system more robust. Once this approach is optimized, then it can be used to develop applications in other industries beyond finance and banking.

Conversational AI for Enterprise Applications

It’s no surprise to see Clinc focused on finance and banking applications considering they’re industries where AI applications are most pervasive. However, Clinc is also targeting other applications such as the following:

  • Call Centers – the system can handle basic customer requests like updating an address, paying a bill, looking for a location, or reporting an outage. Getting started is a matter of days, not weeks. Clinc has built a “chatbot builder” platform similar to the one we looked at being built by the charismatic Indonesians over at Kata.ai.
  • Auto – Advanced Driver ASsistance (ADAS) is on the critical path towards full autonomy, yet an estimated 73% of drivers with ADAS-enabled vehicles have not even attempted to use these features. Clinc built Adasa, the first speech-based conversational interface for ADAS, after analyzing 9,000 conversations between drivers and Ford’s customer service division.
  • Drive Thru – Do you want fries with that? The purpose of that question is “suggestive selling”, and it’s something that can be used to increase sales on the front lines. AI can not only help suggest more sales but also potentially increase accuracy and speed, something that many fast food operators could stand to improve upon.
    Fast food drive thru key metrics
    Fast food drive thru key metrics

    The above chart was taken from a great article by Medill Reports Chicago which talks about how Clinc “is currently in late-stage talks with a number of top quick-service restaurant companies in North America to augment the voice control capabilities of drive-thru windows.” Imagine being able to recognize a customer by their voice so that you can then apply personalized marketing.

  • Gaming – If you’re a gamer, just think about how difficult it is to figure out how to change spells while you’re in the middle of a fight while playing one of today’s increasingly complex role-playing games. Being able to tell your game “give me a healing potion” while in the middle of combat is just one perk of being able to talk to games while you’re playing them using natural conversation.
    Role-playing games you can speak to
    Role-playing games you can speak to – Source: Clinc

    It’s coming.

  • Healthcare – If nomenclature is a problem in finance, just think about how bad things can get in healthcare. Clinc’s first use case is in the area of transcription, but they’re facing some serious competition from the likes of publicly traded speech solutions provider Nuance Communications (NUAN).

The last time we looked at the world’s best transcription technology from Nuance, we were less than impressed. Three years have passed since then, and things have changed. Nuance is now pivoting into healthcare with plans to dominate this space with the recurring revenues to match.

Dragon Medical Cloud ARR
Source: Nuance Communications

We’ll look at Nuance’s pivot in a coming article, but the takeaway here is that there are plenty of competitors out there trying to do the same thing Clinc is in these various industry verticals. Even when it comes to Clinc’s core focus in banking, other companies like Kasisto are also having success in conversational AI for banking with backing from Wells Fargo, DBS Bank, Mastercard, and BBVA to name a few.


Clinc stands out because they’re addressing the elephant in the room which is why we were initially so skeptical about the viability of chatbots. A conversational AI interface is useless if it provides a frustrating customer experience. Unless you can interact with technology in a normal manner without speaking to it like a third grader, then there’s no point in implementing it. In industries like finance where the nomenclature can be confusing, sometimes you need to understand what people are trying to say as opposed to what they are actually saying. These nuances are what Clinc trains their algorithms to handle, something they detail in their paper on Outlier Detection for Improved Data Quality and Diversity in Dialog Systems. If Clinc can succeed in building more natural conversational AI than others, they’ll become a preferred technology, and then maybe we’ll see some consolidation in this space.


Leave a Reply

Your email address will not be published.