Karapet Gyumjibashyan is Senior Director of Product at Krisp, an AI meeting assistant company. He began his career at Synopsys, where he engineered new interface solutions and led trainings for R&D teams. From there, Karapet transitioned to a product innovation role at Ginosi Hospitality Management before joining Krisp.
In our conversation, Karapet talks about how going above and beyond to exceed customer expectations can make their experience “magical.” He discusses how Krisp’s product evolved from serving individual users to large organizations such as call centers and talks about the future outlook of the product’s features.
Krisp is an AI-based meeting assistant. It started out as a noise cancellation app — joining meetings and removing noise from the background so that you can host meetings from anywhere, even with dogs barking, children crying, etc. I was the first product manager to join the company.
We had the right product at the right time — when COVID hit, we saw huge growth as people were suddenly making a lot of video calls. Since then, we’ve expanded the product line. We’ve built a full-fledged meeting assistant that transcribes meetings and creates an AI-based summary in addition to providing top audio quality. It works without bots, which you see in a majority of similar products, and it works with all the apps, so we differ in that way too.
We also built a product line specifically for call centers, where voice quality is especially important. For this product, we’ve developed a new feature called Accent Localization where we localize the accent of the call center agents in real time to match that of the customer’s locale to provide a more personalized customer experience.
Krisp is solving problems for pretty much everyone. It doesn’t matter what kind of call you’re taking — it should be noise-free. In a business context especially, it’s also important to have good documentation like notes, follow-ups, and action items.
From that perspective, our technology is widely applicable across businesses. We found that individual employees within a company would get Krisp, install it, and then their coworkers would get it too. Eventually, they’d purchase a team plan or an enterprise plan. That’s one of the interesting aspects of how Krisp grew to its current size serving both small teams (like sales, engineering, etc.) that need it to be more productive in meetings and big enterprises that invest in the quality of voice at scale, like large banks, insurance companies or BPOs (business outsourcing companies).
We’ve developed specific product offerings for both customer types and organized around those different use cases, from self-serve flows to high-touch enterprise sales processes, because we realize full well that those folks require a different set of processes and motions to complete a deal.
Yes. In general, we’ve found that individual consumers want to self-serve — they want an easy credit card checkout and simple experiences without the human touch. Meanwhile, enterprises want a different set of features, and they want a sales team and customer success to be involved. They want to pay with an invoice and they always have higher security requirements.
With that said, we built a set of core technologies based on AI, and then found the different markets and motions by which we’re able to reach and sell to the right customers and market.
“Magical” is the most frequently used word used to describe what Krisp does. From our understanding, magical means unexpected easiness. When you accomplish something easily but in an unexpected way — something that fully exceeds your expectations — you usually call that magic. The interesting aspect of that is that customer expectations change. We have to keep exceeding expectations to stay magical, so that’s what we do.
We’re trying to build a user experience and overall flow in a way that’s very easy for users. Specifically, we’re building for busy professionals who have lots of meetings and don’t have time to mess around with audio settings — they just want things to work. When they’re able to accomplish that in an unexpectedly easy way, they describe the experience as being truly magical.
It’s a complex thing to understand and measure. That’s why we’re adopted a complex, compound approach based on different signals. First, there are business metrics like revenue or revenue growth. Then there are product metrics like retention, usage, and engagement with features.
We also look at user sentiment. For example, if users take it upon themselves to share the product on social media and refer their friends, that’s always a great indicator that the product is well-received.
We also look at product feedback from reviewers. Within the product experience, there are different points where people can rate the quality of the transcript, the summary that we’ve created, etc., and write a quick comment explaining their rating. This feedback gets sent to our dedicated Slack channels, where the entire product team is subscribed so we can all see it. This also creates an additional motivation for teams to be closer to the users and act on the feedback.
There are several ways to approach this from both a technical and product perspective. We took an approach with AI because we saw the potential to shape it in a way that brings a lot of value to our users. For a company of our size, we have a disproportionately large in-house team of researchers and scientists developing AI technologies. As a result, we’re able to develop in-house, proprietary AI solutions that address user needs.
On our product side, whenever we are faced with a tradeoff, we always try to use technology to minimize user effort. Initially, for example, the engineering team wanted to add an additional control so users could choose options like low-power mode. After a lot of deliberation, instead of adding it to the main screen, we decided to build in the functionality to automatically choose the optimal mode for the user.
This was extra effort on our side to develop, but it was worth it in the long run. We prioritize keeping our solution simple and use the analogy of a DJ deck. We say if we’re not careful with our UX, our app will end up like a DJ deck with a lot of controls. We continually remind the team “We’re not building for DJs.”
On the surface, it looks quite simple, but the cost at which we get to that simplicity takes a lot of work. Structurally, we have engineering, product, design, and AI teams. We’re building technology in a way that’s never been done before, and when you’re going into this uncharted territory, there is a lot that you can do. That’s often the scary part.
Building the product involves a lot of discussion with different teams to find the right balance. It’s a delicate dance, but the thing that unites us is the values of Krisp that everybody tries to follow. One of them is to keep it simple — we only want to add complexity where and when it’s needed. The other value that we have is to ship fast. We’re trying to ship and then iterate, iterate, iterate until we get to the polished version.
From the very start, we thought of customer support as an extension of the product experience. At the time, whatever we did wrong on the product side would end up as a support ticket. It wasn’t something that existed separately and so it made sense to build that team within the product org. Now, it’s a separate customer experience organization, but in earlier days, setting it that way was more logical.
Also, it was a great way to stay close to the users while we were trying to validate product-market fit. Now, it’s obvious that our product is valuable to many users and is needed in things like call centers or meetings, but we needed validation back early on.
The best way for us to continuously receive that validation was to stay close to users. I was answering the tickets along with our co-founder and CEO. We were constantly trying to understand what was happening. That helped a lot. Then, as we expanded, we built separate functions. Now, we have a separate support team, but we have set up the system with Slack, Zendesk, etc. to keep close ties and always stay up-to-date.
I come from an engineering background. In my previous life, I worked as a chip design engineer, but I felt that that field was too narrow for my interests. That’s when I started to transition into what I do now.
I’ve found that my engineering background and understanding of how technology works help a lot when dealing with technical stakeholders. It’s certainly helped with stakeholder management because I work with tons of AI researchers, engineers, and technical people. Being able to understand and empathize with them helps a lot in building credibility, coming up with the right ideas, or bringing fresh perspectives.
Also, the majority of our engineering, AI research, product, and marketing teams are located in Armenia. When working with customers who are mostly based in the US, it helps us to have an outsider perspective. That has some challenges, naturally, but it also helps with broadening the horizon and bringing fresh ideas when dealing with customers, clients, partners, etc.
We’re really looking forward to the future, especially regarding the product directions that we’re building for. Meeting Notes is very exciting, and we see good growth there. Users love it, and we think that we are in the right place to deliver something unique and valuable for the market.
Right now, for example, we have Wikis or Notions where the written documents live, but there is a lot of information that’s currently buried inside the meetings.
There is no good, cheap, and scalable way to extract this information from the meeting and make it useful. We think that we are at the right place to do that, do it well, at that scale. We’ve enabled a lot of company workspace-level features that enable collaboration and make the knowledge from within the meetings useful and collaborative. That’s one part that we’re very excited about.
Secondly, we have innovations in the call center space with noise cancellation, accent localization, and other cool technologies. They’ll be released soon. We’re constantly thinking about where we can add more AI magic to the experience of our customers, and once we find the right place, we do our best to execute on that. That’s where we’re heading.
LogRocket identifies friction points in the user experience so you can make informed decisions about product and design changes that must happen to hit your goals.
With LogRocket, you can understand the scope of the issues affecting your product and prioritize the changes that need to be made. LogRocket simplifies workflows by allowing Engineering, Product, UX, and Design teams to work from the same data as you, eliminating any confusion about what needs to be done.
Get your teams on the same page — try LogRocket today.
Want to get sent new PM Leadership Spotlights when they come out?
Cristina Fuser, VP, Product at Buzzfeed, talks about how being tech-led and data-led can be a strategic advantage.
Jann Curtis talks about understanding your audience’s purpose and what they hope to get from the conversation.
Scaled agile is an approach that allows you to extend agile principles across multiple teams, projects, or business units.
“Disagree and commit” is a management principle that encourages team members to voice their opinions during the decision-making process.
Learn more →
