Have you ever felt frustrated while trying to complete a simple task through an automated customer service line? You're not alone. Many people share comedian Bill Burr's sentiment about wasting precious time dealing with robotic, unhelpful chatbots that seem to understand nothing about their needs.
The tedious experience of navigating traditional automated systems often leaves customers questioning the purpose of these technologies. However, a revolutionary shift is happening in the world of customer service automation, thanks to innovative companies like Posh.
Posh, an emerging technology startup, is pioneering a new generation of AI-powered chatbots designed to make digital conversations feel more natural and significantly less frustrating. Their groundbreaking system utilizes "conversational memory" – an advanced artificial intelligence capability that enables bots to remember previous parts of a conversation, much like humans do.
"Traditional chatbots typically take user inputs at face value, without connecting the dots of what was said earlier in the conversation," explains Karan Kashyap '17, SM '17, co-founder and CEO of Posh. "When you consider human interactions, especially in settings like banks or customer service departments, what was said previously is crucial. That's why we focused on equipping our AI chatbots with the ability to remember historical information within conversations, making them more human-like."
Currently, Posh's sophisticated AI chatbots are being implemented by more than a dozen credit unions across both voice and text-based communication channels. This focused customer base has enabled the company to train its AI systems using highly relevant data, significantly enhancing performance and user satisfaction.
The strategic roadmap includes gradually expanding partnerships with companies in various sectors to gather industry-specific data, allowing Posh to broaden its system's applications without compromising on quality or efficiency. Looking ahead, Kashyap and Matt McEachern '17, SM '18, co-founder and CTO of Posh, plan to offer their advanced chatbot platform as a foundation for developers to build upon.
The expansion strategy is likely to attract businesses across multiple sectors, especially considering the impressive results Posh has already achieved. According to Kashyap, some credit unions have successfully resolved over 90% of customer inquiries using Posh's AI-powered platform. This remarkable success rate suggests that Posh's technology could significantly reduce the frustration typically associated with traditional customer service calls.
"When we implement our telephone solution, there's no frustrating 'Press one or press two' menu system," Kashyap explains. "There's no tedious dial tone menu. Instead, we greet customers with a simple, 'Welcome to [credit union name], how can I help you today?' Users can describe their issues using natural speech rather than waiting for menu options to be read aloud."
Kashyap and McEachern's partnership began during their studies in MIT's Department of Electrical Engineering and Computer Science, where they also collaborated in the same research lab at the Computer Science and Artificial Intelligence Laboratory (CSAIL). Their professional relationship, however, quickly extended beyond the MIT campus.
In 2016, while still students, they launched a software consulting venture that included designing chatbots for companies across various industries, including medical devices, flight booking, and personal fitness. This consulting experience provided invaluable insights into the market need for more advanced bot platforms and better user experiences.
"That consulting period was an incredible learning opportunity," Kashyap reflects. "We gained hands-on experience in designing chatbots using existing tools, which helped us identify the market's need for a superior bot platform and more sophisticated bot interactions."
From the beginning, Posh's founders implemented a lean business strategy that demonstrated their long-term vision. Upon graduating, they used their savings from consulting to fund Posh's initial operations, paying themselves salaries and even hiring contacts from MIT.
Their participation in the delta v accelerator, run by the Martin Trust Center for MIT Entrepreneurship, provided crucial guidance and free office space for a summer. Following delta v, Posh was accepted into the DCU Fintech Innovation Center, establishing a connection with one of the country's largest credit unions and securing another year of free rent.
With DCU as their pilot customer, the founders received what Kashyap calls a "crash course" in the credit union industry. From there, they pursued a calculated expansion strategy, ensuring they didn't grow faster than Posh's revenue would support, which freed them from the need to raise venture capital.
This disciplined approach to growth sometimes required creative solutions. Last year, while looking to develop new features and expand their team, they secured approximately $1.5 million in prepayments from eight credit unions. In exchange, these credit unions received service discounts and a peer-driven profit-sharing incentive. Using this innovative strategy, the company has raised $2.5 million in total.
Now on more stable financial ground, the founders are positioned to accelerate Posh's growth and expand its market reach.
Even referring to many of today's automated messaging systems as "chatbots" seems overly generous. Most existing platforms are designed only for basic intent recognition – understanding what a user is asking for in that moment.
As a result, many virtual agents in our lives, from robotic telecom operators to Amazon's Alexa, can follow simple commands but struggle to maintain natural conversations. Posh's AI chatbots transcend basic intent recognition by employing what Kashyap calls "context understanding" – the ability to interpret user inputs based on the conversation's history. The founders have a patent pending for this innovative approach.
"Context understanding enables our AI chatbots to more intelligently interpret user inputs and handle topic changes without breaking the conversation flow," Kashyap explains. "One of our biggest frustrations with existing bots was that users often had to communicate in an unnatural way to be understood. We've eliminated that problem."
Kashyap notes that achieving context understanding is more manageable when designing AI chatbots for specific industries. That's why Posh's founders chose to focus initially on credit unions.
"Many existing platforms are spreading themselves too thin to make a significant impact in any particular vertical," Kashyap observes. "When banks, telecom companies, and healthcare providers all use the same generic chatbot service, it's like they're sharing the same customer service representative. It's impractical to expect one person to be expertly trained across all these diverse domains."
To onboard a new credit union, Posh utilizes the customer's conversational data to train its deep learning model. However, the learning process doesn't stop at deployment.
"Our AI chatbots continue to learn and improve even after they go live and begin having real conversations," Kashyap says. "We're constantly enhancing their capabilities; I don't think we'll ever deploy a bot and consider it 'finished'."
Customers can deploy Posh's AI chatbots across multiple channels, including online chats, voice calls, SMS messaging, and third-party platforms like Slack, WhatsApp, and Amazon Echo. Additionally, Posh offers an analytics platform to help customers understand what users are contacting them about, enabling continuous service improvement.
For now, Kashyap is focused on quadrupling the number of credit unions using Posh over the next year. However, the founders have never allowed short-term business objectives to obscure their larger vision for the company.
"We've always believed that AI assistants like Jarvis from 'Iron Man' and the AI from the movie 'Her' will become reality in the near future," Kashyap shares. "Someone needs to pioneer the development of bots with contextual awareness and persistent memory. While there's still much work to be done in advancing bot technology overall, we believe that by pushing the boundaries of what's possible, we'll succeed where other chatbots fail, and ultimately, people will prefer interacting with our AI systems over others."
