In the digital era, chatbots are ubiquitous. They’re found in every corner of the internet, from customer service webpages to social media platforms. California-based startup Gleen is one such player in this saturated market. However, they are not just any ordinary chatbot service. Founded by veterans from Microsoft and LinkedIn, Gleen offers a unique chatbot specifically tailored for the most demanding segment of the market – technical communities.
Gleen has a focused approach that propels it to address the most pressing problem in large language models (LLMs) today, which is hallucination. In the context of artificial intelligence (AI), hallucination refers to the phenomenon where AI fabricates false information, yet responds with unwavering confidence. The risk is particularly high in discussions that revolve around complex and esoteric topics. A false, yet confident answer can easily misguide participants.
Gleen has taken up the challenge to rectify this issue before it can be deemed a serious contender in the market. Founder and CEO, Ashu Dubey, explains that the potential implications of hallucination can be massive, especially in areas like cryptocurrency. For instance, if someone falsely claims that the price of a particular crypto token is going to rise to $200, it can lead to a massive market manipulation.
Gleen’s unique vision and ambition have charmed investors from both the established software world and the burgeoning crypto industry. The company recently closed an oversubscribed funding round, raising $4.9 million. The list of institutional investors includes renowned names like Slow Ventures, 6th Man Ventures, South Park Commons, Spartan Group, and CoinShares.
In addition, Gleen has attracted the attention of various angel investors. Notably, Solana co-founder Anatoly Yakovenko, former COO of blockchain data provider Chainlink Mike Derezin, and ISM Angels are among them.
Dubey posits that the majority of chatbots in the market are simply “wrappers” of ChatGPT and other LLMs, meaning that their responses are likely identical to those generated by calling the OpenAI API. Such an approach doesn’t solve the hallucination issue.
Instead, Gleen has developed its proprietary machine learning layer. It sources data from enterprise knowledge, which is then used to cross-verify the responses from LLMs, thereby preventing hallucination. In Gleen’s technology stack, LLMs make up less than 20%. The majority of the work lies in how Gleen stores data and how its proprietary system retrieves data to generate the most accurate response based on domain knowledge.
Gleen’s system, once confident of an answer, dispatches it to different LLMs, such as OpenAI, Anthropic, or a fine-tuned Llama, to create a response. To date, Gleen’s model has been trained on 100,000 pairs of questions and answers.
“Search is our own proprietary algorithm, and that’s where our secret sauce is,” says Dubey. Technical communities and companies where the quality of the answer matters greatly appreciate Gleen’s technology. A good response versus a bad one can make all the difference in these contexts.
With every new user onboarded, Gleen needs to learn its domain knowledge by extracting data from the user’s knowledge base, forums, or Slack or Discord discussions. Dubey believes that Gleen’s ability to abstract information from these sources is one of the startup’s strengths. The company “[doesn’t] need very clean documentation,” he adds.
Initially, Gleen provided its Discord chatbot to a web3 customer. Now, however, it generates more revenue from non-crypto users. With a team of eight employees, Gleen currently serves over ten customers who pay per conversation generated by the bot.
“Customer support is easily a $10 billion market,” Dubey claims. Gleen is venturing into the medium-sized enterprise market, with a long-term goal to “solve customer service for everyone, from mom-and-pop stores to very large corporations.” The early user base is highly technical, but Gleen aims to be “industry agnostic” in the future, which will be a real test of its AI system’s adaptability.
Gleen plans to use its fresh funding to build, drive sales and marketing. A significant part of its go-to-market strategy will involve educating users on issues around hallucination, security, and compliance in the field of generative AI.
“Inbound will continue to be our biggest channel. That’s also very defensible going forward,” Dubey states.
Gleen’s direction is influenced not just by the rapidly evolving AI technologies it employs but also by its customers—companies dealing with cutting-edge technologies like itself. The unpredictable nature of these variables presents one of the startup’s biggest challenges.
“As a CEO, you’re investing in a particular technology or taking a product to the market. But what if the underlying technology completely changes one year from now? You have to restart from scratch,” Dubey muses.
Despite the challenges, Gleen remains committed to solving customer support problems for emerging companies in the best possible way. “We are not married to the technology. We are married to the customer problem,” Dubey concludes.