Please Press 1 to Find Out How Karan Kashyap is Making Your Conversations with AI Less Robotic and More Human

FinTech Sandbox
Published in
9 min readApr 21, 2020


While we may not know exactly how FinTech will impact our future, we have an idea as to who will be leading the charge. Our latest blog series, “The New Faces of FinTech”, spotlights some of the emerging leaders in the FinTech world to get their thoughts on what the future of the industry will look like. Their origin stories are different, their paths to entrepreneurship are unique, but their impacts on their respective industries are significant. No one truly knows what the future of FinTech holds, but these industry leaders may have an inkling as to what we can expect.

Up next in The New Faces of FinTech series is an interview with the CEO and Co-Founder of Posh Technologies, Karan Kashyap. Founded out of MIT, Posh is a team of passionate engineers and computer scientists determined to make natural language interfaces more humanlike and delightful.

Press 1 if you’d like to learn more about Karan’s background. Press 2 if…just kidding. Karan started Posh Technologies with the hopes of creating a world where the days of pressing buttons to communicate with robots will be a thing of the past. Having studied natural language processing and conversational AI at MIT, Karan was fascinated with the rise of products like Alexa and Google Home. Clearly, the likes of Amazon and Google have plenty of resources to invest in sophisticated, state of the art products, but what about smaller companies and startups? Posh Technologies is focused on making conversational AI tools more accessible to smaller businesses, and on helping them automate routine contact center workflows and provide a more human experience to their customers.

This is part 7 in the 2020 edition of our “The New Faces of FinTech” series created in partnership with growth marketing agency Ideometry. Tune in next week for another installment!

Tell us a bit about Posh Technologies?

Karan: Posh is a conversational AI company that came out of MIT’s AI lab in June of 2018. We work with financial institutions, primarily, which includes credit unions, banks, FinTech companies, and insurance companies. We work with these institutions to help them automate routine contact center workflows, and we do that in a couple of ways.

The first way is through a messaging-based interface, which is essentially a chat bot that we deploy on websites and digital banking experiences for users. A lot of bigger banks like Bank of America and Capital One are already using similar technology to help their customers, but we’re hoping to bring that technology to midmarket banks. These smaller banks can’t spend $30 million on building out a product like Bank of America can, so we’re hoping to make this technology more accessible to them.

The second way we’re automating routine contact center workflows is via the telephone. We still see that a lot of our customers have tremendous volumes coming in on their telephony systems, handling nearly 50,000 calls per week. There’s this notion of an IVR (Interactive Voice Response), which is the automated phone experience where a robot asks you to press one or press two and so on. A lot of our customers use these automated experiences, but that’s becoming a stone-age technology, so we’re replacing these traditional touchpad-based IVR experiences with what we call a conversational IVR. This technology lets you actually speak to our AI on the phone and navigates you to the right agent or provides you with an answer to your question without having to press buttons on your keypad.

What milestones has your company achieved so far?

Karan: We’re currently a team of 19 employees right now, all working in downtown Boston. We are working with roughly 20 customers as of today, and that number has been rising pretty quickly. We’re a bootstrapped company, so we haven’t raised a formal VC round yet. We came out of an accelerator program at MIT and when we first started Posh during the summer of 2018, MIT provided us with some initial funding.

What’s the origin story of the company?

Karan: My co-founder was an undergraduate in the same fraternity as me at MIT, so I’ve known him since 2013. We also did our masters program together at the same lab at MIT, which I graduated from in 2017 and he finished in 2018. In the year between the time I graduated and he graduated, we had a vision of starting a company together. We hadn’t necessarily finalized what the actual company would be, but we knew that it would be something related to natural language processing and conversational AI, just because that’s what we studied at MIT.

I was really interested in the trend of conversational AI being used by enterprises. We had heard about Bank of America’s Erica bot, and saw that products like Alexa and Google Home were getting a lot of traction in the market. It was apparent to us that this technology was going to revolutionize the industry, so we decided we wanted to start working with businesses and learn how they could benefit from digital assistants and conversational AI technologies.

In 2017, a few of my friends from MIT and I started working with businesses in a variety of industries, from healthcare companies to transportation companies. During this time, we were experimenting with pilots of conversational bots for these businesses, and these experiences taught us a lot about the improvements that could be made in this field. One thing that we really focused on was the idea of conversational bots not persisting memory or being able to keep track of contextual changes in conversations.

When you use your Alexa or Google Home, these products are very command-based. You’re not really having a conversation with it, you’re more so just giving it commands. But command-based AI bots don’t really work in customer service scenarios. In the scenarios that we were building Chatbot POCs for (Proof of Concepts), some of our early customers realized that a lot of these third party platforms don’t really work for bi-directional conversations. There wasn’t much of an opportunity for people to have a back and forth conversation with the AI, which is something our customers wanted.

That pain point spurred the idea of what we decided to do with Posh. We thought we could build a better platform so that conversational AI could be more human-like by persisting memory and being contextually intelligent.

Additionally, we’re seeing Bank of America and other big financial institutions launching similar digital assistants. We were reading about how much money they spent building these platforms out, and at the same time heavily marketing it. Since my co-founder and I are members of the MIT Federal Credit Union, we thought that a lot of people our age in similar situations would want something like Bank of America’s Erica bot. Unfortunately, we knew that MIT Federal Credit Union probably didn’t have the resources or the in-house expertise to build something like that. That made us want to focus on the mid market opportunities and deliver competitive solutions and conversational AI to them for an affordable price.

Ultimately, we want our technology to be relevant for even the biggest institutions down the road, but working with mid market companies is a good place to start since there’s a tremendous market opportunity there.

How has the conversational AI field evolved since you started Posh, and what do you think the future has in store for it?

Karan: The landscape of machine learning has changed dramatically since we started Posh. Exponential technology is called exponential for a reason, and AI is certainly an exponential technology. Especially in the field that we’re dealing with, since natural language processing is a subset within the AI umbrella. Over the past year, Google and OpenAI and other large research institutions have announced some really big breakthroughs when it comes to unsupervised learning.

These neural driven models are crushing previous baselines and state of the art metrics, which is really exciting for us. We were lucky to enter the space when we did because a few months after we built the foundation of our platform, all of these big breakthroughs started to come out. We’re able to capitalize on these breakthroughs since a lot of them are open source models and technologies, and implementing them into our products has really elevated the way we operate.

As an AI company, there’s two things that are really tough to come by. The first is good data. Having really good data is incredibly important for an AI company, and that’s where organizations like FinTech Sandbox help us tremendously in that sense. When you’re working with a forward-thinking customer, they’ll be open-minded to leveraging some of their data to train your models. We want to be able to leverage data across a wide variety of customers, because when you do it with just one customer, you risk having your models be biased towards them.

If we can convince a group of our customers to work with us collaboratively and give us access to their data, it allows us to create better models and, in turn, provide them with a better product. When our models are based off of a wide variety of data, it fills knowledge gaps and nuances that we otherwise wouldn’t be able to resolve.

The other aspect that’s really hard to come by is the computational power that goes into a lot of machine learning research. Fortunately, Posh is able to keep up with academia and industry knowledge by looking at what Google and OpenAI and other big institutions have done. These larger companies can spend a fortune on experimenting with different model infrastructures and training, whereas we don’t have the massive compute power to keep up with the major institutions.

That being said, by keeping up with the latest cutting edge research and thoroughly ingesting that internally, we’re still able to effectively leverage the latest technologies and implement them for our use cases. Having team members that come from universities like MIT and Harvard with Masters and PhD talent and are well-versed in machine learning research also helps us get access to the best people out there, which further helps us make the most of the latest AI breakthroughs.

How has the COVID-19 outbreak impacted both the short-term and long-term vision of Posh?

Karan: The Coronavirus pandemic has been something we’ve been talking about a lot internally recently. We’re fortunate in the sense that we’re a software company that provides digital self-service solutions to our customers, so our business isn’t being as badly hit as some other startups. With most companies working from home now and practicing social distancing, some of our customers are experiencing shockwaves as a result of this outbreak. For example, some of our customers have contact centers that are only 40% staffed since they’ve had to put measures in place to keep people from coming in. Only a fraction of their contact center representatives are able to work from home, so the COVID-19 outbreak has certainly brought on challenges for many of our customers.

On the flip side, this outbreak has made our customers realize the value of digital-first, self-service solutions like the ones we’re developing. Companies like Zoom are really succeeding in a time like this because they’re hitting on a big pain point. We’re in a similar position where financial institutions are realizing the importance of having conversational IVR automation implemented into their user experiences.

Most importantly, we’re doing what we can to make sure that our employees are safe. It’s great that a lot of our customers are actually speeding up their timelines to launch Posh because they’re seeing the value of a product like ours, but we’re mainly focused on keeping our employees healthy and happy during these difficult times.

What’s it been like being part of the FinTech Sandbox community?

Karan: One of the biggest benefits of partnering with FinTech Sandbox is being able to network with the other startups in the community. We’re still fairly young as a company, so being able to pick the brains of other founders and entrepreneurs that are ahead of where we are is incredibly helpful. Especially during times like this, it’s really valuable to have a network of other startup leaders that you can either vent to or seek advice from or just chat with. Another major benefit of the FinTech Sandbox community is the fact that the program democratizes data access, which has helped us a lot as we continue to grow.

We’re thrilled to feature some of the incredible companies that are currently participating in the FinTech Sandbox program. If you’d like to connect and hear from our interviewees in person, be sure to check them out at the upcoming Demo Day, which has been postponed until October. It’s a tremendous event that represents a diverse, global group of innovators in FinTech and gives others the chance to see our portfolio companies live and up close.

Ideometry is a Boston-based full-service marketing agency serving a global client base. With a full suite of creative, development, and strategic services, Ideometry helps growth-stage startups and Fortune 500 companies alike get the business results they’re looking for. If you’re doing something interesting, we’d love to hear from you. Get in touch with us at or email



FinTech Sandbox

Boston-based nonprofit accelerating innovation and collaboration in finance.