A Conversation With Viraj Mody
A Conversation With Viraj Mody
What the Co-founder & CTO of Common Room has to say about the future of AI
Please subscribe, and share the love with your friends and colleagues who want to stay up to date on the latest in artificial intelligence and generative apps!🙏🤖
Last week we had the chance to sit down with Common Room Co-Founder and CTO, Viraj Mody. Common Room seamlessly integrates with the modern buyer journey, offering a cohesive perspective across social platforms, communities, products, CRM, and beyond. Fast-growing companies use Common Room to gain valuable insights into what their customers are conversing about and drive increased adoption. With Common Room, companies can expedite their pipeline growth more efficiently than ever before.
Viraj is a second-time founder, having started his first business Audiogalaxy, which was later acquired by Dropbox. At Dropbox, Viraj was one of the early pioneers in developing the PLG framework, and was instrumental to Dropbox’s incredible growth from 2012-2018. After Dropbox, Viraj spent some time at Convoy as the Technical Advisor to the CEO before joining co-founder Linda Lian to found Common Room.
Today, Common Room has an impressive roster of customers spanning B2B SaaS, developer tools, and now AI, counting users from marquee names like OpenAI, Figma, Atlassian, Snowflake, Github, Notion, Pulumi, Webflow, co:here, and more. Since its founding in 2020, Common Room has raised >$50M from investors like Greylock, Index Ventures, and Madrona.
In the conversation, we dig into a variety of topics from how understanding the voice of the consumer is critical to how we’re still in the early stages of developing GTM motions and business models for AI-native companies.
Here are a few of our key takeaways from the podcast - let’s dig in!
1. In any vertical, but AI especially, understanding the voice of the consumer is critical. Community building is important!
There is currently a sprawl of places where consumers are talking about product features (Reddit, Discord, X, GitHub Repos, Slack). As a company, understanding what is being discussed in these different communities paired with other core signals (product usage, chats per session interacting, etc.) can be critical to adoption and growth.
Listening to user sentiment and product feedback before commercializing the product can be powerful. First, connect directly with your customers, then connect in your different data sources (CRM, Snowflake, BI Tool) to understand a full picture of what’s happening across the organization.
Its also important to figure out who the user is and ascribe the appropriate significance to them. “A feature request from a college student holds very different weight than a feature request from an engineer at an ML tooling company”
2. OpenAI opened the floodgates for early adopters of AI, but AI tools need to change workflows and lead to increased output & outcomes.
It’s very easy to build a sexy proof of concept or prototype, but the moment you have to scale, it can become very challenging to show value.
There are several opportunities for AI-native companies, including everything from automating mundane workflows to building out mundane data connectors to complete tasks.
This year we will see enterprises moving from “let me just try this AI thing out” to “what is AI actually doing for me? What’s the ROI? How do I see the results in my product"?”
3. GTM motions for AI companies are still nascent.
Nobody has figured out the correct GTM motion for this new wave of AI companies. Every company is still building as fast as they can, but in time that will change and they will need to find ways to sustainably scale.
There is a lot of basic to paid conversion happening among B2B companies, but many of these companies have not connected the dots of figuring out WHO to sell to.
Things will evolve quickly. You have Microsoft knocking on the door for B2B enterprise selling in AI; same with Anthropic and a bunch of others. Startups will have to figure out how to charge (freemium, prosumer, enterprise licenses, etc.). “Someone’s going to have to pay for the H100s…”
4. Ease of adoption is critical, but build a product people want and love.
Build something your customers want and love! If you ignore that and try to get into the mechanics of running a GTM motion too fast and too deeply, you will lose the plot.
There is a lot of value in tinkering! Our entire industry exists because the pioneers tinkered, but finding ways to get companies to pay and adopt your product in the long term is critical.
The days of spamming people with no context are over. I get two dozen emails a day from vendors I don’t look at. But if an engineer on my team tells me “hey Viraj we should go explore this product”, I will proactively set up time with the sales rep. Product matters above everything.
5. 2024 will be the year for commercialization.
2023 was the year for fun side projects and tinkering, but it’s not cheap to do AI production at scale in a commercialized way. Most of these projects bleed money.
2024 will require creativity around optimizing, fine-tuning and commercializing the model to make it work economically.
Things will only get more exciting from here on out.
Special thanks to Viraj for joining us on this week’s Aspiring for Intelligence post. We’re excited to continue following Common Room’s success in the years to come! You can read the full transcript of the conversation below!
Note: VR = Vivek Ramaswami and VM = Viraj Mody. We made some edits for clarity.
VR: Viraj, Co founder CTO of Common Room, we'd love to start with your background and the journey to founding Common Room. Maybe you can just take us through that. I know you're early at Dropbox and had a great career in tech even before that.
VM: Yeah. Thanks for having me. So Common Room is about four years old. And, we have a team of founders that have experiences circling towards what ended up becoming Common Room. So I was at a startup that was acquired by Dropbox and Tom, one of other founders, him and I had started this company.
So Dropbox, we were there pretty early and this was the beginning of the PLG. This was one of the PLG pioneers in some sense before I think PLG was even a term. And so at Dropbox, we built a bunch of internal tooling that just helped the business grow. In unique and interesting ways. And then I did a detour at Convoy for a bit just experience what building a marketplace is, which is nothing like building, consumer or business software.
But then reunited with Tom and then joined up with Linda, our CEO and Francis, one of other founders was an early designer at Facebook and created. Facebook groups to circle around this problem of helping companies grow. Why are they champions, right? It's basically a thing that I guess Microsoft did exceedingly well in the early days with that mvp program And then many companies tried to replicate using spreadsheets and pen and paper AWS heroes.
Linda started that and was involved with that for a while at Amazon. Dropbox, a lot of our PLG motions were based on identifying champions and mapping them to companies and calling a CIO. Yeah, that's how the genesis. Of the company and the idea came together. And since then we've been able to get help companies grow their business via the champions, either champions in the community or champions who use their product.
And it's seen some pretty strong adoption from companies in the B2B verticals for productivity. We have, company like Figma, Atlassian, Asana Notion, and also in the developer tooling space, Temporal, Pulumi budget companies, Vercel. And what's unique about and then we have, increasingly large companies in the AI verticals starting to use this too.
What's interesting, what's common to all of these is champions exist everywhere. Identifying them, who they are, what they're doing and customizing how you reach out to them and help them, help cultivate them is like the common thread across all of these.
VR: And now, Common Room has this incredible set of logos. If you go to the website it's every high growth tech community, tech company with a big community, Figma, Airtable, Notion, just the who's who, which is really impressive and, temporal and SaaS companies and developer centric and infrastructure companies.
So pretty much anybody with a community and a champion can use Common Room. But what's been interesting is seeing, as you mentioned, now you're seeing this bubbling up of more AI specific companies using Common Room, right? You have co:here, Databricks, replit. Both model AI companies, AI apps, these LLM native generative AI companies that we're starting to see it pop up more and more.
Maybe you can just take us through that. What have you been seeing in terms of these AI specific companies and the value that they're finding from Common Room? What are some of the similarities between them and this sort of previous gen or, of classic PLG companies like Dropbox and Figma and air table and others in terms of how they're using like this.
VM: Yeah, for sure. Open AI was like one of our first. Yeah, customers back before even GPT four. This is in the early GPT three days. And yeah, since then we've got on boarded a bunch more. I think what's unique and interesting about them obviously is the rate of growth, right? That's pretty unprecedented.
And that's common to all of these. The PLG motions again are not too different from traditional PLG companies where, you go to open AI, you start playing with the free version and then, you use it enough and you want to graduate to the bells and whistles attached to the paid version.
Same thing for a lot of the other AI companies. What's interesting and where something like Common Room gives them leverage is the sprawl of places where people discuss these. Is huge, you have entire subreddits dedicated to talking about specific features Of these companies, or dedicated subreddits for these companies.
Obviously you have Twitter and X, where there's like a bunch of conversation about it. All of these companies have Discord forums where Discord servers where people talk about stuff, Brainstorm ideas, give feedback. For a company that's growing so quickly, consolidating all of this into Signal.
And separating out the noise is just like a lot of work. Imagine it was your job at open AI to say, Hey, what features are customers asking for this week? You wouldn't even know where to start. Should you start at Twitter? Should you monitor the Discord server? Should you jump into Reddit?
Should you look at official feedback coming in? And so that's where there's you visit your Common Room dashboard. We auto categorize things so you can see, what bugs are bubbling up, what product requests are bubbling up. We help de anonymize people. A feature request from a college student holds very different weight than a feature request from an engineer at an ML tooling company, right?
Without that contention on a Discord server, everyone looks the same. But we're able to map together a bunch of these person 360 views that help you identify, Okay, look, this person from this company said this thing, and then four other people had a similar comment on Reddit. And then this Twitter thread happened all in the context of this company that helps you, focus where to pay attention and we're not to pay attention.
And then just core signals around product usage, right? If I can identify that, hey, in the last week. Some people from Acme Corp signed up for a free version and have created, 15 unique chats per person, tried our, image based generation models. With so many credits, et cetera, then I can have my outbound team or sales team or whatever your GTM motion is, reach out to them in a way that mirrors a lot of like traditional
VR: Yeah that's interesting. And you mentioned that, one of the, one of the interesting things that you can do is you can go into the discord servers or at least track the users and the folks who are saying good things about your product and talking about your product in discord and reddit, basically any forum in which they exist.
Are you seeing a different motion with AI companies and AI users than you did with the previous gen or these non AI companies or the sort of classic SAS in terms of where they exist or how you find them or how they even talk about products. And, one of the things Sabrina and I read a lot about is like this move to open source, right?
You have a number of open source products and models and actually figuring out what's working. What's not is this constant continual like research in motion type thing. And so what are you seeing that's maybe different or differentiated about AI companies than you were seeing about other customers that you have?
VM: Yeah. I think the thing that stands out to me is many of them are still figuring out what they go to market motion looks like. I suspect no one really has nailed it just because, between trying to build as fast as they can, hire as fast as they can and support all the throngs of users coming in, it's really hard for them.
It's been really hard for them to identify exactly how to do it. But I see a few sort of threads forming there. There are a lot of these open source based companies that are really zeroing in on community, right? They have these really rich Discord servers where people bounce around ideas.
They have open GitHub repos, a lot of activity and PRs and comments and everything is very like community driven. And then once they get, any commercial entity backing that will likely go the route of, a temporal which again is like open source software commercialized because ultimately someone's going to have to pay for those H100s, right?
So there's that. And then there's the more like. I feel like OpenAI is a good example of a Notion like go to market Notion, right? Or like Figma like, where you have a lot of early adopter champions mirroring consumer behavior that come in and play with it. I'm using it for vacation planning and, I don't know, help with my kids homework or something.
And then, at the same time, I'm also using it at work for co pilot. And then that makes me A really great candidate to sell as an enterprise customer, right? They should be reaching out to me as long as they know I'm a CPO at some startup saying, Hey, there are, don't you want a license for your entire team to go to XYZ?
And I think they're still running through the motions of connecting these dots, right? This is there's so much just basic to paid conversion happening that they likely haven't gotten to the stage of connecting these like B2B enterprise dots, but it's going to happen and it's going to happen soon, right?
Because Microsoft's knocking on that door, obviously Anthropic, and a bunch of companies are knocking on that door. And all of these, we're going to have to figure out how to make some of these prosumer to enterprise plays happen as part of the GPM stack. And obviously, tools like
VR: How early can a Company or should a company start using Common Room then, do they, it sounds like they, this is part of helping figure out their go-to market, right? And so maybe they don't have any pricing plan in place, or they don't have, it's all free open source. So what, when do you think a, when do you think a company can start finding value out of com?
VM: We're bucking the trend of all these B2B SaaS companies that have a, call for demo and bespoke pricing and essentially no free product. We have a forever free product. You can have sign up, some company you funded yesterday with three engineers sitting in your office can sign up for Common Room, hook up their LinkedIn and Twitter or connect to some subreddit or stack overflow or the github repos.
All of that is free. Everything I said is part of a free offering, right? It never hurts to start early. And that's why we made this free. Look, I've built enough startups where I know if I have to go a shell out. 20, 000 on day one. I'm not going to do it, but if I can start seeing value and over time, make this part of my stack, that's great.
And so that's our approach to how we think about these like emerging verticals. Anybody can use it. The ROI, I want money for my product when I'm making you money. For a while, the world didn't work that way. I think that's how the world should work. If I'm, if you're early and this is a way to just even listen to product feedback we have, in the early days, that was like the big use case for OpenAI, right?
Just listening to user sentiment and product feedback before they were thinking about commercialization and all that. So this is just a great way to keep your eyes and ears on the ground because as an engineer, I build best. When I know exactly what my customers are talking about as companies grow the distance between you know The builders and the customers increases even if you ignore all the go to market side of things and purely focus on that bit Something like Common Room is beautiful because Every engineer can directly connect exactly to what customers are saying, whatever they're saying it.
And then as your data stack matures you bring on a snowflake, you bring on a CRM, start connecting those, and that's when the true unlock happens. Because now you're taking all this power of signal, identity. Categorization and connecting it to business systems, right? So you can say, Hey, look, I have three AEs and here's their book of business.
And oh, instead of cold calling this C level exec, here's an engineer who's been contributing in my GitHub repo and who's been asking questions in my discord server. I should just reach out to them. If they're happy, they'll champion my product to their company instead of me, just bouncing off cold calls.
It's just like such a natural thing that all these software companies that charge tons of dollars and pay one just make it I don't know. We get a bad rap because of a lot of their behavior. So I'm hoping to buck that trend and say, Hey, look, just go to our website, sign up, start
VR: Having been around for four years with Common Room, but obviously in the early days of Dropbox, that when tools like this didn't exist for PLG companies, they had to do their own digging, of figuring out how do I figure out consumer sentiment? How do I transition that and use all that momentum and use that to help me figure out a go to market and our pricing plan. And then how do I go and, create upsell and expansion opportunities? Inside of these companies. So they had to lay the groundwork for all of this.
Now, Common Room is an easy way for you to go and get that all in one place. From the very beginning, I'm curious, given all that experience, but, personally, having seen these companies and also the ones that have done really well. on Common Room. Are there any, pitfalls or obstacles that, this new generation of LLM companies can avoid or be cognizant of when it comes to community building, champion building?
Seeing around corners when it comes to, Hey, we may not have a go to market today, but we're going from PLG or a product market fit now to a scaling company. How can they avoid, maybe they get another five years that they can pull ahead in terms of the market, but then how do we build appropriately.
VM: Yeah. I think the traditional advice that I still stick to is build something your customers want and love like that's gotta be step one. I feel if you ignore that and try to get into the mechanics of running a go to market motion too much, you lose the plot. So that has to be step one.
It's Hey, are you building something that people. Love and want. And then, very quickly you get to the will they pay for and I feel so there's like a ton of value in tinkering our entire industry exists because, the pioneers tinkered around with stuff and then find a way, found a way to make money off of it.
So there's obviously those use cases, but I'm assuming your questions more in the context of like commercial. Yeah, I companies, right? And I feel like community building is critical to helping understand which direction to take your product and get that signal of what customers do and don't love. The thing I would suggest the company starting up here is have in mind the end goal of what your business is going to look like.
Is it going to be mostly professionals paying out of their pockets for your product? Is it going to be like enterprise licenses? Is it a product for organizations within a company? Are you targeting the sales and marketing team using your AI tool? Is it like broad play infrastructure that's targeted more at engineers?
And then focus both your community efforts on that segment. But also the way you then leverage that into outcomes for business, right? The days of just like spamming people with no context are over, I get two dozen emails a day I don't even look at. But if my engineer is someone on my team tells me, Hey, Viraj, we should go explore this product all years, right?
Like I will go proactively set up time with the sales rep. So think about how that dynamic translates to your business. And I think what's interesting is a lot of AI companies are still figuring it out and it's expected, right? It's early enough where not, there's no like simple table or standard formula for this, but I'd say.
Building your community and your motion with intention and then adapting if it's working or not, right? Trying to do something that's clearly not working just to check the box. Doesn't make sense.
VR: It's good advice. We, we've talked a lot about the external use cases with customers. I'm curious about the internal use cases in terms of how Common Room you and your team are using AI whether that's using co pilot or whether you're, building models and how it's like how have you found that Common Room and the team and the technical team are using AI and AI products.
VM: So obviously it's making its way into our product itself for the benefit of our customers. We started with, I think like GPT 3. Even maybe like fine tuning some of those models to do much better sentiment extraction than default at the time to, categorize things automatically to summarize things automatically.
These all have evolved as technology has evolved. Things like, hey, if I need to send, I don't know, if I need to reach out to Vivek on Slack, what's the best way to craft a message that's fully customized to all of my interactions with Vivek across, social and product, etc., in a way that is like super genuine, but also like scalable to thousands of customers.
So that's, those are examples of how gen AI showing up for our customers. And we'll continue to push that. The one principle that I hold dear is I want to make sure it adds value, right? Adding a little gen AI thingamajig in your product that really is just a Oh, hey, look, this is sexy, but it doesn't actually impact my workflows or doesn't actually lead to outcomes.
I'm not a big fan of that. Those are, side projects to mess with. And we have plenty of those. But if it makes it into the product, it's got to deliver value to the customer. And then internally, obviously, we use we use via VS code. It's the default ID for our team. So co pilot.
GitHub Copilot's a big deal. It's helpful. It's not yet at a point where it's gonna threaten anybody's job, I think. It definitely makes a lot of the sort of loops faster, right? If you're stuck somewhere, you have some, something to bounce ideas off. It may not answer the question for you, but it may point you to a resource you didn't have, or Ignite a path that you hadn't thought about.
That's where we see a lot of value across our team. Someone's stuck doing something, they'll just stick it into Copilot or ChatGPT and you'll get some idea that you know, will lead you to a solution that would have otherwise taken 3, 4, 5 hours. So like pretty high leverage there. The support bots is another good internal use case where like we're training some bots to make sure that as we get Questions from our customers.
We can answer them more efficiently.
VR: The last part, the the support bots, is that using some external third party customer support product? Or is that something that that Common Room is building and training themselves?
VM: Do we did a hack week, maybe two months ago or something, where one of our teammates built something like that. And then we're in the process of actually evaluating a couple of external products as well. What's really interesting is the training data. If you keep it purely to your external public facing documentation It's useful, the really hard questions that do come in that requires an engineer to jump on, require a bunch of internal context.
And so part of what we're exploring right now is how do we train it so that it is able to answer some of these customer facing questions? Correctly without a bunch of internal context that may not make sense to customers, right? Imagine some customer asked, How do I do X? And I said, Oh, in this configuration on the server, go change this toggle.
It's that makes no sense to a customer. But internally, there's a lot of those conversations. Part of that conversation is actually valuable to the customer. And I haven't found a product that's nailed it like there's a bunch of like products out there. There's you know Crawl your public documentation into a fantastic job, which is like answering basic questions or they will crawl your entire stack history and just give all this internal gibberish that makes no sense to a customer.
So I suspect we're going to end up with some sort of tuned model that, again, it's probably a bigger focus on how we source the training data or clean the training data, but
VR: It's interesting to hear you how you talk about there's usefulness to these products, but they are not all the way there yet. Like it's not something that you can just either set it and forget it, or you can roll it out in production to all of your customers and it's just working.
And it's fascinating because I think we hear every day about. Some new AI product and, and the McKinsey report of 4 trillion of added benefit to the economy. There's all these, there's hype about AI rightfully in many cases, but in terms of the actual day to day usage, there's still a lot of work that needs to be done.
And a lot of this, it sounds the core products you might be using chat, GPT and co pilot some of these bigger ones from these incumbents. But there's work that needs to be done. For all these new apps that are coming to the floor for you to actually be able to use that with your customers and roll that out and not feel like you need an engineer or a customer support. You have to intervene. Yes.
VM: I think that's where some of the risk with this is right. The build a prototype with Excel or air table as your backing database is a great analogy. You can show something really sexy as a proof of concept, maybe even get a few users to use it. But the minute you have to really scale.
Look, AirTable is great and, Google Sheets is great, but it's not going to scale to Postgres at scale, or like real databases, etc. I feel like it's similar with a lot of these AI tools and AI bots, where a lot of the basic stuff is just so much easier and nicer. But man, if you cannot realize where the sharp edges are You can lead your customers down some like pretty rough paths, or, you can, and everybody understands like the hallucination piece.
I had a crazy example the other day of GPT gas lighting, I would. Others would say hallucinating, I would say gaslighting. Cause it's it was something clearly wrong. And, as part of the, so my kid was like curious about the strongest currency in the world. And so we asked JPT, and it's the Kuwaiti dinar.
It's 1 is equal to 3. 26 Kuwaiti dinars. And, my kid's like smart enough. He's nah, that makes no sense. A dinar stronger than a dollar if 1 is 3. 6 dinar, it should be the other way around. So he asked ChatGPT, no, that sounds wrong. It should be the other way around. And ChatGPT says no, actually you're mistaken.
This is true. And I'm like, oh my God. There's like a difference between hallucination and gaslighting. Make it understand. No, dude, your math's right. And so like it had the fact, but it had the logic inverted. And when you try to correct it, it was like, no, you're wrong.
It's what? Again, obviously I understand how this works. I know why this happens, but would just like something like this. I'd say 90 percent of the world would just fall for it. And, this would show up in many different ways. As an industry, we're going to have to find a way to deal with it because I don't think it's acceptable
VR: like the hallucination versus gaslighting. Hallucination is you make a mistake once. Gaslighting is, I'm going to convince you that I'm right. You're wrong.
VM: Exactly. And I was like, no, stop.
VR: Especially when you have, kids using this too. And so just I'm curious on a personal level, What are you most excited about in AI?
And, maybe we can take this question as more, what do you think is really overhyped in AI today? And what do you think is real?
VM: I am personally excited of just like the opportunity for mundane work to get automated. I think it's, everyone's really excited about all the great sexy things, the complex things that AI can do for you. I'm not convinced that's true. I feel like it like falls down pretty badly as like the level of complexity increases, but man, for the men, for the mundane stuff, it's like, Hey, I'm visiting Vietnam.
What are the five places I should absolutely not miss? That would take me 20 minutes to source via traditional Google searches or whatever, with like very high degree of confidence, most. LLMs will spit out a good answer to stuff like that, or I think some of the things that folks are working on to connect, it's like a LLM based Zapier that does a lot of these mundane connect this thing to that thing and then do this other thing and hit this button.
That stuff I'm really excited about. I feel like Apple on their iOS had the shortcuts app that was obviously very rule based. It was not AI at all. But I feel like it was under under leveraged in the ecosystem. But that's a really good example of something like that as a AI driven Application would just be amazing.
If this happens, then and this happens and insert some like complex logic in there or complex like scenarios in there, then go do this thing. Man, like those kind of things I'm really excited about. I think with a bunch of that in my personal life a lot. But I do think some of the hype around, oh, we build this model that can do this complex thing really effectively.
It's it can in the happy pants, life is not all full of happy pants.
VR: It's interesting. I feel like simplicity can. Really win here, right? Just do the simple things right. And that's 90 percent of The tasks people end up doing, and you mentioned LLM based Zapier, like that's a concept we've thought a lot about as well. And, Zapier is creating Zapier AI, but that's different from an LLM based Zapier because one is like stitching together different and then you're running it and the other, is the flip side of that.
And so that's that's interesting. And we're starting to see. So more products come up in that direction. This is great for us. Is there anything that, we didn't touch on that you think is worth covering?
VM: Yeah we could talk about this stuff for hours. I think 2024 is going to be really interesting because I suspect this is a year most people are going to need to commercialize. A lot of fun projects or like side quests and it's not cheap doing a lot of this at scale in products in a way that doesn't bleed money.
It requires a bunch of creativity and effort and I suspect 2024 is going to be when a lot of people figure that out. It's take all these models, fine tune them or create a very purpose built. Mini model, then you can commercialize it or, then you can actually make it work economically.
I feel like that journey is going to happen pretty quickly in 24. So I'm curious to
VR: We've talked about last year was just all the experimental budgets going into AI and this year it's going to be, what's the ROI? What kind of return am I getting?
VM: Exactly. It's LLMs are exciting and necessary for certain problems. You don't need anything nearly as large for the bulk of problems. What's your small scale model that solve this one problem really well.
VR: Well, thank you so much for Viraj for joining is. This has been terrific and we're really excited to see what Common Room room goes too!
VM: Of course, thanks so much for having me.