Doug Camplejohn
(00:01)
Hi, this is Doug Camplejohn, your host of Revenue Renegades. In this week’s episode, I am very pleased to be joined by Dan Lee, the CEO and co-founder of Nooks. I’m going to let Dan tell us all about his founding story. Welcome to the show, Dan.
Dan
(00:16)
Yeah, thanks for having me here.
Doug Camplejohn
(00:18)
So lots of founders as listeners. We always love to hear the starting story. Tell us how the idea for Nooks came along and if there were any pivots along the way.
Dan
(00:30)
Yeah. So really quick background. Before starting Nooks, I studied computer science and AI at Stanford. That’s where I met my co-founders, Rohan and Nikhil. I had mostly done work in machine learning, both research and applied. Right before Nooks, I was working at Scale AI on their ML team pretty early on and ended up transitioning to remote because of the pandemic, and then started hacking on Nooks.
Nooks started very different from where it is today. It actually started as a virtual classroom for Stanford office hours, very different. All my friends were still in school and doing classes on Zoom.
Doug Camplejohn
(01:15)
That’s very different.
Dan
(01:26)
Nooks actually started as a project, not as a company. I was interested in learning, and I had mostly done ML work building models. Nooks is actually the first full-stack product that I built. My goal was, how do I have faster feedback loops? Models take days and weeks to train and months and quarters to figure out if they’re actually going to be useful, and products you can ship within minutes and hours, start getting immediate customer feedback, and make an impact.
So my first goal was to learn, and the problem right in front of me was virtual classes. That summer of 2020, two Stanford classes started using it, and then in the fall, a bunch of other Stanford classes started using it. It ended up being most of the Stanford computer science department plus several other departments using it for office hours during the pandemic.
We had a bunch of interesting features. You could do problem sets together. You could split into different groups, each working on different problems, and once someone finished, you switched groups. It was a nicer version of Zoom rooms. We had a way to watch lectures together while you’re sitting at a table with your friends, and you could talk to your friends while watching the lecture.
Dan
(02:51)
We were trying to innovate on the virtual education experience. Stanford actually wanted to pay us about $2 per student per quarter, and we were super excited. But then we realized that students wanted to go back in person, and we weren’t sure how much longevity there was in this business. So we instead switched to a virtual office. We figured remote work would stick around.
We had product, marketing, sales, and engineering teams all using Nooks to work together, kind of like Zoom but more for co-working than for meetings. We found that sales teams were actually our most engaged users, and that was really interesting. You’ve watched The Wolf of Wall Street? You know the scene where everyone huddles around Leonardo DiCaprio making calls and selling penny stocks?
Doug Camplejohn
(03:39)
Yep. Great movie.
Dan
(03:46)
He does really well, so they copy him and make a lot of money and party. For that use case, it’s exaggerated, but it’s about figuring out what works and replicating that across the team. It was a bunch of junior sales reps who were making calls in their bedrooms and then getting feedback once a week from their manager, who listened to their Gong calls from the last week.
They used Nooks to do this together like on a sales floor and to replicate that in-person experience. This was really interesting to us because Rohan, Akhil, and I all have engineering backgrounds. If you had asked me five years ago, I would have said, sales, that’s a dirty job. Shouldn’t you build a product that sells itself? But in spending time with the teams using Nooks, we realized that one, it’s not a dirty job. It’s actually a very human job. You’re solving people’s problems and creating customer value. It’s a great fit for collaboration because you want tight feedback loops to figure out what works. Also, the job is going to change.
It was early 2022 when we started focusing on the sales use case, moving from virtual classroom and virtual office to sales. ChatGPT hadn’t come out yet. We were surprised how quickly AI could do a lot of this work. Sales reps were writing emails, making calls, and doing research, and it was clear this was not going to look the same in the near future.
When starting Nooks, I was interested in building a better way for people to work together online. If you get that working, you have a lot of data on how they work. Can you use that data to make it smart? Can you automate manual work and automate best practices and winning behaviors? We found that sales was just this great use case for all of the above.
Doug Camplejohn
(05:47)
So you basically started in 2020. Were you doing that full-time at that point when you started, or was it still just a project on the side?
Dan
(05:55)
Yeah. I was class of 2021 at Stanford. I stopped in 2019 to work at Scale, and then started working on Nooks in mid-2020 as a project. We raised a pre-seed and a seed while we were still in the wilderness, figuring out if we were a virtual classroom or a virtual office. There was even a little bit of virtual events in there. Then we found the sales use case and started focusing on that in 2022.
That’s when we got out of the wilderness and started growing pretty quickly.
Doug Camplejohn
(06:28)
Got it. And so what does the product lineup look like today?
Dan
(06:31)
Yeah. At a high level, Nooks is the surface where sales reps source deals and close deals. We believe that the biggest opportunity in AI automation is on the sourcing side. You can’t 10x your win rate, you can’t 10x your ACV very easily, or cut deal cycles that much. You can’t 10x your ability to close.
But on the sourcing side, 90% of the work is probably able to be done by AI today: writing emails, making calls, doing research. Nooks helps with all of the above.
Doug Camplejohn
(07:12)
So let’s break those down. On the research side, obviously you’ve got the traditional list companies, whether it’s ZoomInfo, Apollo, and you’ve got this next generation of companies like Pocus, UserGems, Common Room, etc., that are all doing the signal stuff. How do you fit into that world?
Dan
(07:29)
I’d say of those two buckets, there’s always going to be a long tail of data going in. If you look at mature industries like finance, people buy datasets like security camera footage on grocery store parking lots to trade stocks. There’s always going to be a long tail of data inputs.
I think the second bucket you’re talking about is more synthesis on top, turning data into insight. We play more in that space. We integrate with ZoomInfo—ZoomInfo actually uses Nooks—and a bunch of other data providers. We compete more in that other category you described of signals and research.
Doug Camplejohn
(08:25)
Okay. So part of that is the “who should I go after” problem. Then there’s the “how to reach them” piece. You have dialers and things like that, competing with Aircall and folks in that realm?
Dan
(08:42)
Yeah. I’d say we own the end-to-end process. Think of a rep’s workflow. A rep has 500 accounts. If they’re an SDR, they make money when they book meetings in that set of accounts and generate opportunities. If they’re an AE, they make money when they close deals in that set of accounts.
We first help them research their book of business to identify priority accounts. You can’t call all 500 at the same time. Which ones have a competitive renewal coming up because we identified that from previous conversations we had? Which have a prior champion? Which are increasing headcount in the department you help with, so they’re investing in this area? Which have a relevant tech stack? All these different signals that a rep would be looking for.
We look at first-party data sources like your CRM, transcripts, calls, and Nooks data, as well as third-party sources like ZoomInfo, LinkedIn, 10-K reports, web research, etc. We help prioritize which accounts to go after. We also help you identify which people at those accounts you should go after. We have models trained both to help identify which accounts based on those signals, and then who at those accounts based on their persona and other interaction history we’ve had.
Then we’ll draft copy, both email messages and call scripts that you can use. We’ll help send those emails as well as place calls to them and basically automate that whole process, all the different pieces of generating pipeline.
Doug Camplejohn
(10:38)
So as I was prepping for this, I was looking at some of the posts that you’ve put out there. You talk about the broken state of go-to-market SaaS stacks. I would assume that what you’re describing here is consolidating a lot of those pieces. Where do you think it’s most broken today?
Dan
(11:02)
Yeah. I’d say consolidation is not the goal. Productivity is the goal. Consolidation is necessary, but not sufficient.
For example, in the past, the way to grow as a sales tech company was to bolt on adjacent products. Outreach was trying to build Gong, Gong is trying to build Clari, Clari is trying to build something else, and so on. That was the way to grow because there was a fixed amount of software spend. Software is enabling work, and there’s tens of billions of dollars spent on sales software.
But the opportunity today is orders of magnitude bigger. If you look at all the people using that software, that’s hundreds of billions of dollars, and AI can now read, write, speak, and reason like a rep. Consolidation is necessary because if you think of a fragmented stack today, which most people have, it requires a rep to be human glue between a bunch of disconnected workflows. You have human glue here, workflow here, human glue next, workflow next.
When writing an email is manual, it’s kind of okay if a person needs to find the email address in ZoomInfo, the deal notes in Salesforce, and the transcript in Gong in order to write that email, because it takes longer to write the email than it does to be the human glue. But when AI can write that email, that shrinks the workflow, and now the human glue takes more time than writing the email. So it doesn’t work anymore.
Consolidation is necessary to get the productivity gains that AI enables, but it is not sufficient, because if you just build the same stack you have today, it’s not actually built to do work for you.
Doug Camplejohn
(12:59)
You’re not taking advantage of it. So when you’re going into an account, what are you typically replacing or going up against in a competitive situation?
Dan
(13:08)
Yeah, it can vary.
Doug Camplejohn
(13:11)
You can leave out the competitor names if you want. I’m just thinking more about the replacement sale. For listeners who are thinking, “I’ve got a fragmented stack. What can Nooks come and solve for me or replace?”
Dan
(13:24)
Yeah. We play in several different pieces. Often you have a very fragmented stack for generating pipeline, and we can often replace most of that stack. We don’t compete in the data piece. You still probably need a data provider, but we integrate with all the data providers.
The rest of it is the action: making calls, writing and sending emails, doing the research and signals piece. We help with all of those areas. In calling, which is where we started, it’s actually pretty greenfield. Most people have some VoIP thing where there’s a phone number on their screen, they click it, and they call it. It’s a little better than their cell phone.
It’s great because we’ve run trials in our sales process where we come in against decades-old technology and we double or triple their pipeline per rep. We can show that in a week or two. That’s why we’re able to grow quickly. Calling automation today is kind of like email automation a decade ago, when everyone realized you don’t want to pay reps to manually bulk send emails.
Doug Camplejohn
(14:39)
That’s amazing.
Dan
(14:50)
People are realizing today you don’t want to pay reps to listen to ringing and answering machines, find phone numbers, take notes, and do all the manual stuff around calls.
Doug Camplejohn
(14:57)
Got it. I’m nerding out on this for a minute here. If you’re sourcing a list according to someone’s ideal customer profile or buyer persona, here’s a list of calls that you’re going to make. You’re doing the dialing for them. If you don’t have the right number, you’re falling over into data sources and retrying. All of that work is handled for the rep?
Dan
(15:25)
Exactly. Reps should talk to prospects and define the strategy. They should label edge cases once, not twice. AI should be doing more and more of that work over time.
The way I think about it, sales software used to enable work. Now AI can do work. There is a new type of data that you need in order to do the work. You can’t train a model on all the emails you’ve sent before and expect it to start sending emails as good as you. You can’t train a model on all the deals you’ve won and expect it to have as good a fit model as your best rep.
Your best rep is taking into account a highly dimensional combination of factors: what’s my interaction history with this person and with other people at their company? What are their relationships with those other people? What’s their tech stack? What phase of the buying or awareness cycle are they in? All these different factors to understand: are they a fit, what’s the best timing, what’s the best message that will resonate?
Being that surface that’s doing work for the rep, it’s our job to capture the rep’s mental model. This is data that doesn’t exist in the CRM. Reps are not naturally incentivized to label data. They’re incentivized to hit quota. So we have to be thoughtful about product principles and rep incentives—why are they labeling data for us—as well as the technical pieces. You need a transparent feedback mechanism. Reps should always be able to make progress. They need to trust the outputs we’re giving them. It’s basically building human-in-the-loop AI.
Doug Camplejohn
(17:30)
What are the kinds of things that they’re labeling for you? Having lived in the CRM world for a long time at Salesforce and LinkedIn, reps hate putting in any data. So how are you getting them to put in any data of value?
Dan
(17:42)
Yeah. As a rep, you’re incentivized to hit quota. Why do I need to update Salesforce? The opportunity today is that AI can now do more of that work for you, and it needs labeled data. The benefit needs to be immediate. There’s no “three months from now it will deliver some value.” Once you give some feedback, it should immediately make improvements. That’s how you drive rep incentives to label more data. You’re making them more likely to hit their quota when they label more.
Doug Camplejohn
(18:27)
Can you give an example of what a rep might label something as and what benefit they’d see right away?
Dan
(18:33)
Yeah. For example, when we do research and identify signals on whether an account is a good fit, we’ll cite the data sources and the reason why we believe this is the case. Reps can say yes or no, and we can learn over time which data sources we trust and how to interpret them. When it says they have a renewal coming up but this competitor actually always does two-year deals, that’s something you probably need to learn from the rep.
Another example is when we draft emails. We’ll give several different versions. The rep picks which version they want and can also give AI suggestions like, “make this part shorter” or “add information from this data source.” We have memory, so we learn over time. Reps shouldn’t feel like they’re training models or labeling data.
Doug Camplejohn
(19:42)
Right, that’s what I was going to say. That’s a great example with the email picking. If you’re saying, here’s a few choices, you can see which one they pick, you can see what they actually sent, and if they have different response rates and all that, so you’ve got this feedback loop.
Dan
(19:58)
Yeah. A lot of people listening probably watch Netflix. The Netflix algorithm learns faster when it suggests different movie options and then you pick among those, rather than just passively observing you search and pick yourself. The reason is you can test the edge of the decision boundary. If something is a close call, it might throw a few options that are close to the edge and see which one you actually pick. It’s called a bandit problem in information theory.
Doug Camplejohn
(20:33)
Very interesting. You’re already talking about AI automating a lot of the prospecting and dialing. How do you think the role of the seller is going to change over the next few years?
Dan
(20:49)
Yeah, it’s really interesting what happens when AI lets you do more with less. Sales is much more similar to recruiting than to customer support.
As a customer, ideally you can get an automatic answer to your question. That’s probably the best customer experience: full automation. As a candidate, you’re not going to take a job if you don’t meet someone at the company, because you can only take one job and you’re going to be picky about it. If you’re good, you have a bunch of people trying to hire you, so you have a decision to make, and you probably pick the one where you build trust and a relationship with someone.
Same thing with a buyer. When you’re spending a lot of money on something, solving an important problem, and you have a bunch of people trying to sell you things, you’re going to pick the one where you build trust and a relationship. Humans are going to be selling.
I think it also depends on the stage of the company. If you’re already tapping out your TAM, you can probably do the same with less. You might need fewer people. But if you’re in growth phase and you can make sellers more efficient, you might actually want more sellers.
Responsibility for generating pipeline will always exist. Someone will own that. The job can change, though. The reason why the SDR role has existed historically is because you don’t want your best enterprise sellers writing emails or making calls all day; it’s low leverage. But when you can make generating pipeline more efficient, you might want your best sellers having those conversations with prospects and defining the strategy. You’ll still probably need a talent pipeline, a place to start in the sales org, but there’s going to be more variance in how teams are structured because of the efficiency gains. More things can work.
So you might see different models, but humans will still be central.
Doug Camplejohn
(23:02)
What do you think about the whole AI SDR space?
Dan
(23:05)
I think, like I was saying, for a buyer, the ideal experience is probably not to get fully automated outreach. When you don’t have sales reps yet and you want to test selling, or if you have an ACV that doesn’t support using people, this can be a short-term boost. But I think the results speak loudest.
Doug Camplejohn
(23:42)
Yeah, I assumed that was your answer. I just thought I’d throw it out there based on your previous comments. Let’s switch gears a bit. You’re a first-time CEO, correct? Tell me about the lessons you’ve had going through these pivots and growing the team. The team is how big now?
Dan
(24:02)
We’re about 180.
Doug Camplejohn
(24:05)
Okay, that’s a lot of learning in a short period of time. What are some of the biggest lessons you’ve had and any hacks you’ve come up with for people management or hiring?
Dan
(24:19)
Nooks is the biggest company I’ve worked at. It’s also my first real job. I’ve interned at companies that were all less than 150 people. It’s been interesting with lots of learning.
I’m very fortunate to have mentors and advisors who are much more experienced than me that I can lean on. It’s extra important for me to understand when there’s something I don’t know and who to go to for help with it. It’s really helpful to have a support network.
Some of the advantages of being a first-time founder and having everything be new is that we get to approach things differently. The way companies are built today is probably different from how companies were built a decade ago. We have a core value called “ask why.” This means we should understand everything from first principles. If you’re doing something without knowing why, you should probably stop and understand why first, because there’s a chance you shouldn’t be doing it.
We’re building our own company. There’s no role model of “we want to look like X for sales.” Approaching every new problem with a fresh perspective is actually one of our competitive advantages.
Doug Camplejohn
(26:08)
There are people who say there are going to be sub-10-person or sub-50-person companies that will become billion-dollar companies. You’re on the other side of that in terms of employee headcount. Roughly, what’s the breakdown of product versus go-to-market? And then how do you think about using AI in terms of the efficiency of your own organization, besides your own products, of course?
Dan
(26:40)
Yeah. In terms of breakdown, roughly a third of headcount is on engineering, product, and design. Most of the rest is go-to-market, and then we have some G&A.
We use AI across the board. We use plenty of AI coding tools. Our engineering team is very efficient and productive. We were early beta testers of Cursor. Michael came to the office and worked with our engineering team. We use AI in every function. We actually test our sales reps for AI fluency and problem solving.
Doug Camplejohn
(27:38)
How do you do that?
Dan
(27:40)
We have this exercise where we have them learn a new tool, one we expect they haven’t seen before. We see how they do. Can they figure it out? How do they approach it? It’s hard to succeed these days and compete if you’re not AI-fluent and not good at problem solving, because you can get so much leverage from the right tools. So the profile across positions changes as AI gives you leverage.
Doug Camplejohn
(28:22)
What’s been the most surprising thing about being CEO? You go into a role with certain assumptions about what it’s going to be about. What’s been the biggest surprise?
Dan
(28:35)
There have been lots. It has evolved over different phases. I started building the product and wrote a lot of the code in the early days of Nooks. Over time I shifted more into being a recruiter and everything else in between. I was our head of finance and our head of people for a long time.
People say this all the time, but it’s really sunk in as the team has grown. As a leader—for myself and all the leaders on the team—I can do everything right but hire a bad team, and we’re not going to do very well. I can do everything else wrong and hire an amazing team, and we’re probably going to do pretty well.
This is true for all the leaders we have too. The importance of the team and thoughtfulness in building the team is probably one of the areas of biggest learning and what I think about the most today.
Doug Camplejohn
(30:02)
In your interviewing process, you mentioned you’re looking for AI fluency and problem-solving skills. Without giving too much away—although frankly, if a candidate learned about it on this podcast, they’ve done their research and that’s a good signal—what kind of questions do you ask in an interview to filter people out? Or after people have gone through an interview process, how do you decide as a group whether that person is a thumbs up or thumbs down?
Dan
(30:22)
Everyone come watch the podcast.
Dan
(30:47)
A lot ties to our values. I mentioned the problem-solving piece, which is related to the “ask why” value. You should be able to understand things deeply.
I’ll give another example. I ask every engineer: would you rather have really interesting, impactful customer and business problems but technically boring work, or really interesting hard technical problems but not understand the customer and business impact? You’ll see engineers in both buckets. For product-facing roles, we much prefer the former.
We have a very hard interview process and can screen for technical talent. But motivation should be aligned to customer and business goals. This is related to our value of “earn customer love.” Everyone, whether customer-facing or not, should be focused on earning customer love. On our engineering team, we have something like 13 international math, coding, and physics Olympiad medals, but they are all motivated by customer problems, and that’s something you don’t always find.
That combination of talent and customer motivation is key.
Doug Camplejohn
(32:32)
So “ask why,” the focus on the customer problem. What are some of the other values?
Dan
(32:38)
“Do more with less.” As a small startup, we have to prioritize ruthlessly. You can do anything but not everything. You should only do the most important things. If there’s something that’s not important, don’t do it. Be lazy about non-priorities.
Doug Camplejohn
(32:57)
How do you determine what’s the most important thing?
Dan
(33:00)
It’s related to customer and business goals. Goal setting is very important. Everyone should be clear and aligned on the goals.
In creating a high-performing team, there are maybe three ingredients. One, inputs should be correlated with outputs. If you put more in and don’t get more out, that’s a recipe for burnout. If you put more in and keep getting more out—more customer impact, business impact, personal growth, comp, etc.—that’s motivating.
Two, everyone around you should be doing the same. If you’re the only one working hard, you’re not going to be very motivated. But if everyone around you is working hard toward the same goals, of course you need to keep up.
Third, the goals need to be very clear. No one likes sprinting back and forth if the goals are changing underneath you. Everyone needs to be swimming in the same direction. When you get those ingredients and point toward the right goals, you can do great things with a small team.
Doug Camplejohn
(34:26)
I have to say, Dan, you strike me as someone who is wise beyond their years when it comes to being a CEO. So congratulations on all of that and all of the success. Nooks is an amazing story for something where you’ve been doing the sales thing for less than four years now. It’s pretty wild.
So let’s shift gears. I know we’re running out of time. Let’s find out a little bit more about Dan. Everybody in your company who listens to this is going to want to find this out. What’s one thing we’d be surprised to know about you?
Dan
(34:58)
One thing you’d be surprised to know about me: I grew up playing ice hockey.
Doug Camplejohn
(35:04)
Okay. Where did you grow up?
Dan
(35:04)
I grew up in New Jersey. My mom taught me to ice skate when I was little. My dad taught me to rollerblade when I was little. I naturally got into hockey and played from when I was young through the club team at Stanford.
Doug Camplejohn
(35:30)
You still play?
Dan
(35:29)
I still skate sometimes. I don’t play much. Ice hockey is probably one of the most inaccessible sports. You need Zamboni ice, 10 players, a ton of gear, and a referee.
Doug Camplejohn
(35:41)
Right. First thing, Northern California.
Dan
(35:53)
Not to mention goalies.
Doug Camplejohn
(35:48)
Yeah, it’s not a solo sport for sure. Separate from that—you can’t give that one as an answer—what do you like to do for fun?
Dan
(36:02)
I stopped playing ice hockey. I like the outdoors. I rock climb. I got a paddle board recently—paddle board, hike. I ski and snowboard a little in the winter. Anything outside. I got into slacklining over the pandemic.
Doug Camplejohn
(36:25)
Okay, interesting. What’s a product, besides your own, that you love and why?
Dan
(36:35)
I live in Slack. It’s probably my most used app, as I’m sure most people listening can relate to. It’s a huge part of our culture. Slack is the new social media in some ways, especially as more and more things happen digitally. It’s where a lot of interesting, “good to know” stuff happens.
Doug Camplejohn
(37:03)
We had Tamar Yehoshua, who was the CPO of Slack and is one of our investors, on the show recently. She’ll be very happy to hear that. I’ll pass it along. I agree.
Finally, how can our listeners stay in touch with you and help you on your journey and Nooks’ journey?
Dan
(37:20)
If you know amazing people, we are hiring across the board: engineering, product, design, sales, all of go-to-market. You can reach out to me, add me on LinkedIn, or apply on the careers page.
Also, if you know people or teams that should be using Nooks—any teams that are trying to generate more pipeline, which I imagine is a lot of you—come visit the website and submit a demo request.
Doug Camplejohn
(38:08)
Dan, it’s been great to have you on the show. I really enjoyed the conversation. Thank you for taking the time.
Dan
(38:13)
Yeah, thank you too.
