To date, the stories I’ve shared on Focused Chaos have been my own. But I always intended to share other people’s stories as well, and today that’s exactly what I’ll be doing.
Many of you may not know Scott Stevenson (yet!), but he has a great entrepreneurial story.
Scott is the co-founder and CEO of Spellbook, the first generative AI copilot for lawyers, launched last summer. Scott studied computer engineering, previously started a musical instrument business (which we’ll get into a bit) and was the Director of Engineering for a networking monitoring startup. Scott loves philosophy of mind, electronic music and mountain biking in Canada.
For me, Scott has an amazing journey to share.
He comes out of university and starts a company. He doesn’t know what he’s doing but figures things out. Learns. Adapts. He doesn’t accomplish his ultimate goals initially but then tries again, in a completely different industry and is now crushing it. And there are a ton of valuable insights in this interview worth digging into.
Before digging into this incredible interview, subscribe below. It’s free!
Q: We’ll get to Spellbook in a moment, but I’d like to hear your ‘origin story.’ You co-founded a musical instrument company after graduating university. The company, Mune, invented a new type of instrument. You don’t see a lot of new instruments being invented. Of all the things you could have done, why this? And what happened to the business?
I was really passionate about making music growing up, especially electronic music. Some of my first “engineering” experiences were programming and building synthesizers.
In university I met Dr. Andrew Staniland, a composer and music professor. He was doing some incredible electroacoustic composition combining electronics with modern instruments in a classical setting.
He noticed that classical audiences had trouble connecting with traditional mixing board style electronic instruments—people really couldn’t tell the difference between a DJ pushing “play” and more complex performances.
We designed the Mune to provide the audience a full view of electronic performance.
We didn’t hit our Kickstarter target, but I learned a lot:
I met one of my idols at our first trade show: Roger Linn. He was the inventor of the first digital drum machine. I realized how hard the electronic musical instrument business was when he said his company only had one or two employees besides himself! It was at that moment it struck me that maybe it wasn’t the right market for me–I was hungry to build something bigger. Our Kickstarter not making it was confirmation that the market was pretty small.
I learned how frustratingly expensive legal services were for an entrepreneur with little cash. I raised a small $20k angel cheque, and before I knew it half of that was spent on legal fees to set up the company, bring on our first employees, issue shares and protect our IP (we had a design patent).
That experience really planted the seed of wanting to make legal services more accessible & efficient.
Q: Two years later, you connect with Daniel Di Maria (lawyer) and Matt Mayers (designer) and start Spellbook. You said you were frustrated by the legal bills you paid at Mune, and then Daniel shared that lawyers were also frustrated by the inefficiencies of legal work. So you held a ~2 year grudge against lawyers, and then thought, “I want to make their lives better!”?
😆 Very early on we tried to build a legal automation system that small businesses could use to bypass lawyers, to stay on top of their records and get basic transactions done. However, we quickly realized:
People were still super nervous about doing legal work on their own, valuing the opinion of a trusted lawyer.
Lawyers were seriously hungry to end drudgery in their own work.
I met a few disillusioned lawyers, like Daniel, who had the feeling: “This is what I went to law school for?”—after spending countless hours staring at 10 Word documents strewn across the screen.
Then a law firm approached us after seeing our automation technology: “We really want something like this to better serve our clients. We know we need to adopt technology or our clients will adopt it themselves.”
After that, we were 100% on servicing lawyers, and that was a great decision. Everything we do now is focused on empowering lawyers while keeping them in the driver’s seat.
I’ve come to believe that lawyers have one of the hardest jobs in the world. A lot of folks (including myself in the past), do not see exactly what goes into closing a complex transaction.
For those that are curious, here’s a sneak peak into some of the new capabilities launching into Spellbook:
Q: You have a technical background in engineering. What made you believe you could be the CEO of a SaaS company for lawyers?
I was initially very technically focused and built the first versions of our products with Matt. Between working on Mune and Spellbook, I was fairly involved in smart contracts and smart legal contracts built on Ethereum. I guess this gave me time to get acquainted with contracts and their problems. It’s awesome to have a lawyer like Dan on the team, but I think when innovating, there are some benefits to being a bit naive to how things are done too. I could ask naive questions to Dan like: “Why don’t lawyers just do X?” and he would either tell me I was stupid or that I might be onto something.
Sidebar: Novice or Domain Expert Founders?
The topic of novice versus domain expert founders comes up a lot. I’ve seen this argument play out quite often. There’s no definitive answer. Some people prefer domain experts others prefer novices. When thinking about Highline Beta (the venture studio that I run) we tend to lean towards domain experts, who understand a vertical or industry but need co-founder like help to build a solution. I asked Scott for his thoughts on this topic, since he brought up his own inexperience in the legal industry.
Q: If you meet two founders, all things being equal, except one is a domain expert and the other is a novice (naive as you put it), which one are you investing in?
Scott: My top bet would be on a domain expert that has cultivated "Beginner's Mind." This is very rare though, so naive founders can often outperform domain experts.
This clip of mathematician Gregory Chaitin does an amazing job of explaining why, and it was really influential on me:
This Beginner's Mind attitude is deeply important to our culture, and I think it's crucial to any startup success: being outside the existing dogmas allows you to constantly outmanoeuvre competitors in unexpected ways.
It isn't just important to be open-minded about the domain in question, but about everything in the startup. 90% of startups fail, so doing what other startups do will probably just lead to failure. Naivety helps avoid the sort of cargo-culting that generates failure outcomes, and actually helps you be exceptional.
Naivety also helps avoid doing things for the sake of prestige. When we first launched Spellbook based on GPT3, people would ask: "Is that too easy? Where is your 10 person machine learning team?"
Q: In January 2018, ChatGPT wasn’t live and the AI craze was non-existent. Did you know AI was coming in such a big way or was the original version of the product significantly different and you’ve since “pivoted” into the AI phenomenon?
The original product (Rally) was built around document templating technology. It would essentially enable a lawyer to complete common transactions based on advanced legal templates + workflows.
Lawyers loved the promise, but in practice we would often get the same piece of feedback: “this is great for junior lawyers, but templates don’t capture the bespoke nuances of my drafting.”
At first we never took that feedback seriously. Then we heard it again, and again.
From trying GPT2, we had an inkling that something might be coming down the pipe to help our customers, but GitHub Copilot was the real “aha!” moment.
We tried GitHub Copilot when it launched and boom, inspiration struck. “Oh! This is what our customers are looking for!” GitHub Copilot was really the first generative AI copilot, aimed at programmers. You would just start writing code, and it was this super powerful auto-complete that would finish your code for you.
We were well-positioned to combine this new technology shift with the insights and technology we had developed from working on the “legal transaction streamlining” problem with 100 law firms. One unintuitive insight we had was that lawyers worked in Word all day, and really didn’t like leaving it. This made us bet on putting Spellbook in Word, just like GitHub Copilot was in the software developer’s environment (VS Code).
I think this approach of being a “wind at the back” of your users while they work through their normal habits and workflows is still underrated. People don’t really want to interrupt their workflow to go to ChatGPT, they would prefer if the AI was constantly thinking ahead of them and suggesting what they need before they think of it.
Q: Do you believe Spellbook has a competitive or unfair advantage? If so, what is it?
Yes, in more ways than one, but I’m going to keep those aces up our sleeves for now. 😂
I will say that VCs who think “proprietary data” is the moat alone, have an overly-simplistic view of how these products work. That’s just a nice Hollywood story that sounds good in a pitch deck.
[As an aside, I think pretty much any pitch that is “super legible” like a four act movie is suspect, due to what Venkatesh Rao talks about here. This is one of the most influential blog posts I’ve ever read!]
When you’re building something truly new, you realize that reality has a surprising amount of detail, and it’s actually not very legible. I think that’s a super important realization for founders building something that’s not just incremental. At Spellbook we talk about caring about these countless details as “Woodworking” instead of “Lego Brick Thinking.” I wrote about this here.
Q: How much bullshit is there in AI right now? And, do we really need a copilot for everything in our lives?
🤣 As a rule, I think 90% of what happens in startups and business is bullshit. That’s why so many businesses ultimately fail, or have terrible unit economics. Mostly people are just mimicking what others around them are doing, but that just leads to following the herd off a cliff. That’s why I think beginner’s mind and naivety are crucially important.
I like Luca Dellanna’s concept of “mimetic societies.” IMHO this is the main bullshit generating force in business:
No, we don’t need a copilot for everything in our lives, and companies coming to the party a year late and blindly dropping ChatGPT into their offering won’t get the results they want. This is just unthoughtful mimicry. But this is an easy kind of BS to spot!
The hard kind of BS to spot is the companies that are overly focused on prestigious R&D (eg. training their own LLMs from scratch with proprietary data sets) vs. taking cheap high-leverage moves that get you to delighting your customer 10x faster. If a 2010 SaaS company was building its own database software from scratch, that would not be a good thing. Similarly I think there are too many founders attracted to the prestige of keeping up with the firehose of AI papers, rather than racing to deliver great value to their customers leveraging the ridiculous new off-the-shelf tools we have. It’s still a ton of work to orchestrate these tools and tie them together in a compelling way, just like building a SaaS app is still a ton of work even if you don’t build your own database from scratch.
I think this speaks to one of the most common errors startups make (which I wrote about in my post “Is GPT-3 Too Easy?” referencing David Sirlin’s book “Playing to Win”:
Q: What kind of validation did you do with lawyers early on (and perhaps still today) to give you confidence in terms of product direction?
Through the course of the company, we’ve run over 200 growth/product experiments with over 100 landing pages. That’s how we often test message-market fit very inexpensively.
With Spellbook, we showed our first prototype to a handful of customers and got a kind of reaction we had never seen before. Pupils dilating, leaning in, a true physiological response that arose from lawyers realizing: “wow, my job is about to change forever.” NFX talks about what this reaction looks like here.
I think that’s what real product-market fit looks like, it is truly a neurochemical, drug-like response from customers that cuts through all the noise in their lives and puts adopting your product near the top of their priority list. It’s almost impossible to get there. Anything less and you end up with a poor CAC payback that never gets resolved, no matter how much you work on distribution.
Once we saw that in one-on-one meetings, we launched our landing page and quickly had thousands of lawyers signing up like we had never seen before.
Originally Spellbook was actually just going to be a fun little marketing lead magnet for us. It ended up becoming our main source of revenue in just a few short months. Startups are like that!
Clicking through Mixpanel product analytics is also one of my favourite weekend activities 😆. I think it’s really important to have analytics accessible at all times, so you always have this ambient sense of what is working and what is not.
Q: Looking back on the journey so far with Spellbook, what’s one big mistake you made? What’s one big lesson learned?
I feel like our strategy from day 1 was largely the right one: wake up every morning and ask, “what do we do today to move the boulder forward?” No high-minded strategy, no “being a multiplier”, no grand ideas—just nudging the boulder forward every day.
My biggest mistake was thinking that things “should” be easier.
It generates so much anxiety when you think that a startup is supposed to snap together like lego bricks. You can’t just implement a bunch of best practices, hire for key roles, delegate, and expect anything good to happen. It really does take most companies years of experimentation to find product-market fit.
The magic happens in the relentless massaging of details. It’s not efficient. If I were a VC, I would not invest in a company that seems overly structured and process driven. I just don’t believe you can create great things this way. It leads to “going through the motions.” These companies may have been able to be traded up through venture rounds during ZIRP, but I don’t think many will survive under real interest rates.
"If you’re efficient, you’re doing it the wrong way. The right way is the hard way. The show was successful because I micromanaged it—every word, every line, every take, every edit, every casting. That’s my way of life." - Jerry Seinfeld
It took me being involved with 3 startups to accept that things really are this hard. It’s not just your company. It’s the nature of the beast.
Once you realize this… Wow! The anxiety of being a founder completely melts away. You are no longer comparing yourself to some unachievable bar. You realize the “skill” you are learning is simply grit and patience. You just need to have a little more grit and patience than competitors.
Q: You raised USD $10.9M in May 2023. What was that experience like? Markets were (and still are) down, but given that you’re in a very hot space (broadly: AI), was the fundraise challenging, easy? What are your recommendations for early stage startups fundraising in this climate? Should they all embed AI into their stories, ASAP?
It was probably our most straightforward raise, because we hit a true product-market fit inflection point that was clearly visible in all our metrics. We quickly had a 30,000 lawyer waitlist and could not keep up with demand. I think no matter what you’re doing, if your metrics are hockeysticking like that, investors take notice.
The AI narrative helped (having a real landscape shift is helpful to explain why this product couldn’t have been built by a competitor 5 years ago), but it was the metrics themselves that really mattered.
I think most investors are savvy enough to see through founders who tack on AI to their narrative in an unthoughtful way. But I do think, if a founder doesn’t have evidence of PMF, there are still tons of opportunities to create it using generative AI.
If you think about it, it took so long for SaaS companies to find all the opportunities that the original SaaS CRUD stack enabled—a decade or two. It will take quite awhile for all the genAI use cases to be uncovered. I think it’s probably better to look in that fresh AI soil than to try to find an opportunity in the technology we have all explored over the past 20 years.
Q: What advice would you give to founders that are raising capital now?
It’s always a numbers game. Patrick Hankinson from Concrete Ventures told me early on that on average it takes 80 meetings to close a round. I always keep that in mind. IMHO for earlier rounds you should aim to meet with at least 80 investors over a 3 month period. That’s where you can really get deal momentum. The biggest mistake I see founders making is getting a couple “nos” and throwing in the towel. That really means nothing! You just need one investor out of 80 to agree to lead.
I think the best way to get this kind of volume is to have one person be the BDR, finding the meetings, and one person to be the AE, doing the pitch and trying to close.
For Seed and later, understand your unit economics inside out. Get a real empirical idea of your CAC Payback. If it’s not under 12 months you likely don’t have PMF, and you should probably conserve cash and experiment until you get it down. Investors care a lot about unit economics now–vision and story is not enough.
Q: You’re very transparent on Twitter/X (everyone, follow Scott!) so I’m curious: what advice do you commonly see promoted on Twitter/X that you think is total nonsense?
Well, there’s a lot of nonsense on Twitter and that’s part of the point for me. If you think of your goal as finding a good new idea, you can actually get there by traveling through a bunch of bad nonsense ideas. I find the sort of “stream of consciousness” content super valuable even if it’s totally wrong.
But I think the most dangerous nonsense is the sort of fortune cookie advice that makes the listener feel good and makes the speaker sound smart. These sentiments are super viral and people say them to get followers and engagement. They can stay around forever without being true.
For example, infinite variations of “work smarter not harder” make the speaker sound smart and make the listener feel good because now they don’t have to work hard. But making something great usually requires working smart andhard.
Q: You’re also on Twitter/X a lot? Why?
I’m addicted 😆
I really value finding ideas that generate “alpha.” The only way you can make money in markets is by knowing information and acting on it before others do. The whole game of investing and business comes down to that. Twitter was buzzing about GPT2 years before ChatGPT—being exposed to that was such an incredible advantage.
Actually, there’s one thing founders need to learn to do besides finding alpha: they need to learn how to resonate with others en masse. That’s marketing! Twitter is an awesome training ground for learning this skill. If you can’t make a tweet that resonates, how are you going to create a whole marketing funnel that resonates?
Q: You don’t live in Silicon Valley / SF? Why? Isn’t that the center of the AI universe?
Nope, I actually live the furthest away from SF in North America you could possibly drive, in St. John’s, Newfoundland! We also have an office in Toronto and folks across Canada and the US.
Honestly, I regret not spending time in the Bay Area earlier in my career. I think I would have learned a lot of lessons faster and met a lot of like minded hackers. I find it important to stay networked there, and to keep ingesting the zeitgeist via Twitter.
But sometimes I say that the spaciousness and beauty of the nature here in Newfoundland is one of our secret weapons. So many of my ideas (and tweets 😆) come from solitary walks in the woods, which I do almost every day. The cost of living is also great, which has helped us be a cockroach.
Q: How do you go about learning and improving yourself as a CEO? (Referencing this tweet)
I still have a ton of room to grow! But most of my learnings come from doing, and just trying to push the boulder with our team over and over.
In the referenced tweet, that was just hard learning from years of practice. I realized that weeks and weeks could go by where I was “juggling all the little balls” successfully, but no major levers were being pulled.
I am a super conscientious person, and I often feel like I am battling against that as CEO. I wrote about that in How to Finally Make Something. Over time I’ve just seen that by prioritizing the #1 thing and letting some small balls drop, we can perform much better. Blitzscaling also alerted me to the idea that any successful company will always have countless fires burning. If you let that hold you back from The Big Thing, you will get stuck.
I also find speaking with founders to be immensely helpful in growing, especially those who have made it to exit.
Q: What’s one thing you used to believe is true that you no longer believe? Why?
That if you trust the process/method, results will come. I absolutely do not believe this anymore. I tried it for years and nothing good happened. It’s not good when you transfer responsibility from yourself to some method. You need to force everything 😆.
Agree nearly 100% with it, especially: “It’s not a 1.5x change in efficiency, it’s a 10x change in efficiency.” This is exactly what we’re seeing on the ground.
As a side note, NFX are such great clear thinkers. It’s amazing to me that they have nailed their assessment so well from outside the trenches. To me that kind of thinking is the joy of business. It’s competitive philosophy where you get an objective score on whether you are thinking more clearly than your competition. Building a business is an amazing way to develop the ability to think clearly, and that to me is a sort of spiritual pursuit. That’s what makes it worth it for me.
I hope you enjoyed this interview! I learned a ton. If you did too, please subscribe below for more great content:
