Introducing Harmony: The New Voice AI Sales Platform from monday.com

Sean Nolan: Hi Vitali, thank you so much for joining me, and congratulations on the launch of Harmony. To start things off, can you quickly introduce the origins of this new platform? I believe it was created by a team at monday.com, but could you share more? Vitali Margolin: Harmony started as an acquisition that we made at monday.com a while back. At the time, I was leading the CRM product. We were thinking about what could come next in CRM, and voice is a meaningful part of that area. We bought a company called OneAI, which had amazing technology. We then decided that, instead of simply integrating it into the existing product, we would spin it off into a new product called Harmony. It can connect to the rest of the system later on. That is how it started. Sean Nolan: Great, thank you. Harmony is primarily a sales agent. Could you explain more about the problems facing sales teams that Harmony is looking to solve, and what makes Harmony unique in this space? Vitali Margolin: Harmony is a voice AI platform, but there are many platforms doing voice. What we found is that having a sales conversation is very different from having a support conversation. The conversation needs to feel much more fluent, and it is a much more difficult conversation to have. Selling something is not the same as simply having a conversation. There are a lot of things within the conversation that general AI does not do well enough. That is what we wanted to solve. We have smaller models within the conversation that know how to navigate it and do the things that general AI is not trained to do. For example, how do you negotiate or handle objections? Or even understand whether it is a voicemail or a regular conversation where somebody has picked up? When you are doing a simple demo, everything works amazingly well. But when you go out into the real world, the agent might need to call Sean, but Sean’s secretary picks up instead. The agent needs to navigate to Sean. You need to do all those things when you are handling a sales conversation, inbound or outbound. That is the world we wanted to solve for. This is our promise with the product. Sean Nolan: Absolutely. As you say, sales conversations are unique and also high-stakes. If a conversation does not go well, that is potentially a customer and revenue lost. It is extremely important to get these conversations right the first time. You mentioned that general AI can be limited in this space. I understand Harmony uses task-specific AI models. Could you share more about why those models add more value than a general LLM? Vitali Margolin: One important thing to remember is that it needs to be very fast. When you are having a conversation with a human, it does not take five seconds for them to reply. It needs to feel real. You cannot use large language models that are very slow in a real-time conversation. You cannot use Opus in a real-time conversation. It will feel weird. But smaller models are just not as good. When you think about it, during each stage of a conversation, if you break it down into steps, the steps are quite simple. In an introduction, you do not need an Opus-level LLM to say, “Hi, is this Sean?” Or if I am pitching you something and you say, “I am not interested,” or, “I work with a competitor,” you do not need such a large model. But it does need to be a trained model for that, or it needs to have specific context. The smaller ones are not as good at that. This is where we are investing: making it feel fast, under 200 milliseconds, so it feels like a real person, but smart enough that it can achieve the same results as a real person. Sean Nolan: Exactly. As you say, you are finding the balance between speed and intelligence. It does not need huge intelligence because that is not what it is being used for. It is about the specific use case and achieving speed. Taking a step back and looking at Harmony as a whole, how do you see the workflow of a traditional sales rep changing and evolving as AI continues to develop and join traditional workflows? Vitali Margolin: It has evolved a lot since we thought about the initial concept. When I am talking with a potential lead, what I say is: I am not here to replace any of your people. We can try, but that is not why I am here. I am here because your business has so much more potential. Your reps are answering phone calls from nine to five. Nobody is doing it 24 hours a day. I understand why, because it does not make financial sense. But with my product, you pay per minute, so it makes a lot of sense. What happens during peaks? What happens when there is Black Friday and you want to do something bigger? Think of marketing teams. If I am a company and I want to bring more traffic to my website, I go to Google Ads or Facebook Ads, increase my budget, and get a lot of traffic. But if I am a CRO and have a sales organisation, there is no dial. I cannot just click a button and get 50,000 more sales reps. But with AI, you can. I am not saying, “I am here to replace your team.” I am not. My AI will not close you a $50,000 deal. That is not what I am here for. But it will help you achieve your potential, the full potential of your business. It might replace some people, perhaps people you do not need anyway. But I think a lot of companies will reach a higher potential for their business using this kind of technology. Sean Nolan: It sounds like one of the benefits is flexibility. You do not need to use these agents all the time during quieter periods. But, as you say, during a Black Friday-type period, you can scale them up. You have the flexibility and capability to make sure you are not losing revenue. That seems to be one strong point. Vitali Margolin: Take monday.com, for example. We have more than 250,000 customers. There is no way we can call every one of them every month and ask, “How is it going? Do you have any issues? Do you have any problems?” But now it is very easy. You do it with the click of a button, gather all this important information, and route the information that is relevant to a real representative. They can then call the customer back and expand the account or solve churn. There is so much untapped potential in voice, and people still want to talk with someone. Even when they chat with support, many people eventually write: “Can I just speak with someone, please? I cannot write to you anymore. I just want to speak with someone.” You can tell them, “Okay, I am calling you right now.” It is that easy, and it is something you can do. Otherwise, you need to hire people, train them, and then they churn after three months. It is a very hard job. But if you can focus your team on just the most important things and use AI for the rest, it is a win-win. Sean Nolan: Absolutely. Thank you. Finally, looking ahead, what does the next 12 months look like for Harmony? Do you have any key priorities in mind? Vitali Margolin: Twelve months ahead in this ecosystem feels like millions of years ahead. It is so hard to imagine what will happen in 12 months. But, in general, we understand that we want to come to companies and solve problems for them. We do not just want to give them a product that does something specific. We want to solve an issue. We are expanding our product to more channels, including SMS, and allowing real agents to interact with our product. I want to create one system where, as a business owner, you can see all the conversations happening, AI and human, in one place. Then the AI part can learn from the human part. If we have an amazing sales rep doing an amazing job, I want the AI to learn from it without anyone having to say, “Jack just closed an amazing deal. This was his speech.” I want it to happen automatically. I want to give business owners those insights. We can improve the agents based on human interactions and create that loop, or even work the other way around. The goal is to provide a full solution where AI and humans are on the same platform, working together to allow the business to grow autonomously. If we can reach that, I think it is a game changer for us and for companies. That is where we are headed. In 12 months, we will see. Sean Nolan: Thank you so much. That sounds fascinating, and we will be keeping a close eye on Harmony and how the company continues to develop. Congratulations again on the launch, and thank you for your time. It has been great speaking with you. Vitali Margolin: Thank you so much, Sean.