Founder Interview
How Zeff Built an Enterprise Deep Learning Platform with 3 Customers and $30K Monthly Burn (Interview with Co-Founder and Chief AI Officer Ben Taylor)
- Interview Date
- July 9, 2019
- Interviewee
- Ben TaylorCo-Founder and Chief AI Officer
Company Metrics at Interview Time
Customers (2019)
3
Monthly Burn (2019)
$30,000
Team Size (2019)
5
Historical Snapshot
These numbers were reported by Ben Taylor during the interview recorded in July 2019 and are a historical snapshot, not current figures. See Zeff’s current numbers.
Key Takeaways
- 01Zeff had 3 paying enterprise customers at the time of the interview in July 2019
- 02The company was burning approximately $30,000 per month, funded by the founders
- 03Team of 5 people, with 2 co-founders and 4 considered core leadership
- 04Entry-level pricing was an annual subscription with transaction-based overage charges
- 05Zeff previously scaled to 5 or 6 customers before pulling back to focus on insurance and assessment verticals
- 06The company had received 4 or 5 acquisition offers due to AI sector hype
- 07Zeff had term sheets in hand and was hoping to raise $2,000,000 at the time of the interview
- 08The platform focused on unstructured data: image, audio, video, and text
- 09Founders had been bootstrapped for approximately two years before this interview
- 10Ben Taylor holds a background in chemical engineering and previously worked at Intel, Micron, a hedge fund, and HireVue
Company Metrics at Time of Interview
| Metric | Value | Source |
|---|---|---|
| Customers (2019) | 3 | Founder interview, Jul 2019 |
| Monthly Burn (2019) | $30,000 | Founder interview, Jul 2019 |
| Team Size (2019) | 5 | Founder interview, Jul 2019 |
| Co-Founders | 2 | Founder interview, Jul 2019 |
| Acquisition Offers Received (2019) | 4 to 5 | Founder interview, Jul 2019 |
Growth Breakdown
Revenue
Zeff had paying enterprise customers from day one, though early customers received discounts. At the time of the interview, the company had 3 active customers, and Ben Taylor said that until Zeff crossed a million dollars in revenue, all of that early revenue was "just market research".
Customers
Zeff scaled up to 5 or 6 customers and then scaled back to 3. Ben Taylor said the team "definitely made some mistakes chasing a few of those alpha accounts", big vanity brands that a company without "a team of 10 plus" cannot yet support, and that some earlier customers had been professional services work. The team refocused on insurance and audio or video assessment as its core go-to-market verticals.
Team
The company had 5 people at the time of the interview, with 2 co-founders and 4 members considered core leadership who had worked together for about a year. The sales lead, who is not a founder, brought 28 years of enterprise sales experience, including time at Oracle.
Funding
Zeff was bootstrapped at the time of the interview, with founders personally covering the $30,000 monthly burn. It had term sheets in hand, said it would "probably move forward with one of these", and was hoping to raise $2,000,000.
Growth Strategy
Vertical Focus Over Breadth
After chasing a range of enterprise accounts including large vanity brands, Zeff narrowed its go-to-market strategy to two verticals: insurance and audio or video assessment. Ben Taylor argued that AI is a complicated sell and that understanding a customer's domain deeply is essential to closing deals.
Customer-Driven Product Development
Ben Taylor credited the company's most competitive innovations to customer interactions rather than internal ideation. Two years of close customer feedback shaped the platform's direction, ensuring it was not built in a vacuum.
Annual Subscription with Consumption-Based Overages
Zeff structured its pricing as an annual subscription with a generous usage ceiling, billing for overages only above it. Ben Taylor said he could not remember the exact numbers but put the ceiling at "maybe 10,000,000 images per month" or "maybe 10,000 video hours per month". This model was designed to reward experimentation and not penalize customers for running R and D workloads.
Senior Enterprise Sales Talent
Zeff's sales lead, who is not one of the founders, brought 28 years of enterprise sales experience, including time at Oracle. He was paid a mix of commission and equity, getting more equity than a typical VP of sales might.
Best Quotes
“So we wanted to make deep learning easier for the enterprise. And so we we chased AutoML. A lot of people are doing that, but we focused on unstructured data. So image, audio, video, and text. So those are the datasets that we're focusing on. The revenue model is a subscription based platform. So people pay an annual subscription to get access to the platform to build these models.”
“We scaled up on on our customer counts. I think we hit five or six, and then we've scaled back to three. And one of the interesting things with AI... So so two quick points. One point with a SaaS startup, and this is true not just for us, for other people, you'll land into vanity brands. So think about like a Facebook or some some top Fortune 50 company that's sexy. They might come across your plate, and you'll be tempted to entertain them as a customer.”
“Until you have a customer... Because our customers right now, they're kind of scattered, we have a product customer, we have a marketing customer. Until you have repeat customers in a vertical, I would argue you have not found product market fit. And so that's what I'm hoping to have in the next six months is I can say, look, have two insurance customers, two assessment customers. These are real verticals that are growing.”
“We're we're being aggressive with burn because we really need engineers and the engineers we have to hire are not... They're all principal level engineers. We can't hire kids out of school and expect them to ramp on enterprise level deep learning, which requires big data needs, high performance computing needs. Yeah. So we're... Every new contract we close, we hire.”
“There have been other AutoML companies that have developed in a vacuum. And that's why I really value the revenue model that we started with, that we've had paying customers from day one. We don't have all those customers today because some of them have been professional services. So we haven't developed our platform in a vacuum. The last two years has been very intimate customer feedback.”
“When you're an entrepreneur, you worry about the payroll. Like, can I make payroll this month? And if I'm talking to you and you say, man, it's a hard month, I'm gonna go without pay this month. My my... Or my reaction is like, wow. Well, like, bravo. Like, wait until you do that for, you know, months and months and months.”
What Happened Next
At the time of this interview in July 2019, Zeff was a bootstrapped, 5-person enterprise AI startup with 3 customers and had term sheets in hand for the $2,000,000 it was hoping to raise. The figures and plans described here reflect the company as it stood during that conversation and should be read as a point-in-time snapshot. Visit the Zeff company profile for the most current information available.
View Zeff’s current profile and metricsRun a software or AI company? Get interviewed by Nathan Latka and tell the story behind your numbers.
Full Transcript
Chapters
- 0:00Introduction and Background
- 0:38What Zeff Does and the Revenue Model
- 0:59Pricing Structure and Transaction-Based Overages
- 3:21Customer Count and Pulling Back from Vanity Brands
- 4:19Vertical Focus: Insurance and Assessment
- 5:22Product Market Fit and the Path Forward
- 5:48Team Size and Seniority
- 7:15Burn Rate and Hiring Philosophy
- 7:47Bootstrapping and Founder Sacrifice
- 10:21Raising Capital and Valuation Target
- 11:29Customer-Driven Innovation and Product Development
- 12:26Famous Five: Books, Tools and Sleep Habits
Introduction and Background
Nathan Latka
00:00Hello, everybody. My guest today is Ben Taylor. He's the co founder and chief AI officer at a company called Zeff dot ai, z e f f dot a I. Previously worked at Intel and Micron at a hedge fund and a Sequoia startup called HireVue and then founded a deep learning, auto machine learning platform back in 2017. Ben, you're ready to take us to the top?
Ben Taylor
00:18Yeah.
Nathan Latka
00:19Was that company... Was that auto machine learning platform in 2017
00:22that you founded? That is Zeff. Correct?
Ben Taylor
00:26Yeah. It was called Zeff. Now it's Zeff. Yeah. That's a sad story.
Nathan Latka
00:29Yeah. No. We'll just stick to Zeff. We'll jump into that stuff maybe later. But tell us on what the company does and and kinda what the revenue model is. How do you plan to make money?
What Zeff Does and the Revenue Model
Ben Taylor
00:38So we wanted to make deep learning easier for the enterprise. And so we we chased AutoML. A lot of people are doing that, but we focused on unstructured data. So image, audio, video, and text. So those are the datasets that we're focusing on. The revenue model is a subscription based platform. So people pay an annual subscription to get access to the platform to build these models.
Pricing Structure and Transaction-Based Overages
Nathan Latka
00:59And do have any idea of kinda what you're gonna price against? Is it seat based or number of images scanned per month? Or what's the pricing look like?
Ben Taylor
01:06So the entry level pricing, we started at $250,000 a year, and then we have a transaction based scaler that goes up from that. Our pricing could come down in the future. So we do want to grow with our customers. And as we find more more opportunities inside the business, we want to be able to grow those contracts.
Nathan Latka
01:26What's a transaction based scalar though? Like transactions of what? Give me an example.
Ben Taylor
01:31So that would be... Normally the way we build out the SaaS model is we give them a very healthy ceiling where we don't track transactions. But if they, for example, if they exceed... I can't remember off the top of my head the numbers, but if they exceeded maybe 10,000,000 images per month or a million video,
01:52maybe 10,000 video hours per month, then we would start tracking that and billing them for overages,
02:01which is pretty interesting. So you never want to tax innovation. So that's one of the challenges with AI is you want to reward discovery. And so we don't want to charge them for training and R and D experiments. We want to charge them for consumption.
Nathan Latka
02:18Real consumption or there's real utility output?
Ben Taylor
02:22Yeah. We did look at charging them per model early on. The issue with that is the first model might be high value. It's easy. But what if you wanna do a second model that's mediocre value? It causes issues.
Nathan Latka
02:36Because they don't wanna experiment because you to charge them.
Ben Taylor
02:40Yeah. And I'd rather just have you have all models in production that you can think of, even if some of them aren't very useful
Nathan Latka
02:40Yep.
Ben Taylor
02:40Or valuable.
Nathan Latka
02:47So when did you launch... So when did you guys write the first line of code for the company?
Ben Taylor
02:52Two and a half years ago.
Nathan Latka
02:53Okay.
Ben Taylor
02:53So call me Yeah.
02:54'20
Nathan Latka
02:56So
Ben Taylor
02:58so three years ago, I would put us into the class of what you'd call a wantrepreneur, where they're employed full time, but they think they're an entrepreneur. And then it's not until you cut the strings, you burn the ships, where your personal paycheck's dependent on your ability to hustle. So we did that two years ago. We've been we've been bootstrapped.
Nathan Latka
03:17That's great. So still bootstrapped today, and how many customers have you scaled to?
Customer Count and Pulling Back from Vanity Brands
Ben Taylor
03:21So it's interesting. We scaled up on on our customer counts. I think we hit five or six, and then we've scaled back to three. And one of the interesting things with AI... So so two quick points. One point with a SaaS startup, and this is true not just for us, for other people, you'll land into vanity brands. So think about like a Facebook or some some top Fortune 50 company that's sexy. They might come across your
03:48plate, and you'll be tempted to entertain them as a customer. The issue is you're always one feature away from closing them. And honestly, if you don't have a team of 10 plus, you're not at a stage where you can support like a monster customer like that. So we definitely made some mistakes chasing a few of those alpha accounts. So we scaled back and really focused on the verticals that we're chasing right now, they're insurance and assessment.
04:14When we're kind of
Nathan Latka
04:14Risk assessment.
Ben Taylor
04:14It's almost like a restart.
Vertical Focus: Insurance and Assessment
Ben Taylor
04:19Insurance and any audio or video assessment. So there's other things we support with our platform, but it's very important to have a go to market strategy because AI is such a complicated sell. You don't want to start the conversation over with every new prospect. You want to kind of under... You want to understand their domain. If you're insurance, I wanna come talk to you about loss or fraud.
Nathan Latka
04:38So, I mean, can I take the three customers you have and they're $250,000, your accounts are $20,000 a month, you're doing about $60,000 a month in revenue right now?
Ben Taylor
04:47So the early customers, they definitely had some discounts. The customers we have in the pipeline right now, they're the ones that were kind of going through procurement and final contracts. They will come in close to the full amount.
Nathan Latka
04:59Yeah. What the... Just the... But if you took it just revenue last month on an MRR basis, are you more like 20 or $30,000? What... What's that at?
Ben Taylor
05:06Yeah. Be more like 20 or $30,000. And honestly, anything
05:11to have some revenue is great, but until we cross a million revenue, I would say all of that early revenue is just market research.
Nathan Latka
05:11Mhmm.
Ben Taylor
05:11Like...
Product Market Fit and the Path Forward
Ben Taylor
05:22And until you have a customer... Because our customers right now, they're kind of scattered, we have a product customer, we have a marketing customer. Until you have repeat customers in a vertical, I would argue you have not found product market fit. And so that's what I'm hoping to have in the next six months is I can say, look, have two insurance customers, two assessment customers. These are real verticals that are growing.
Nathan Latka
05:43Yep. What's the team look like today? How many people?
Team Size and Seniority
Ben Taylor
05:48We've got five people. We're pretty proud of the seniority of the team, but this, compared to most of the startups I think that you interview, I would say we we just found product market fit.
Nathan Latka
05:59Mhmm.
Ben Taylor
05:59It's two years of market research.
Nathan Latka
06:01How do you... I would argue you haven't found that yet, just because there's not enough customer volume. I would say you're you're still very much exploring, but that's also a very back of the napkin kind of thing.
Ben Taylor
06:10Yeah. So the funny thing with AI, you know, AI hype is insane right now. So we're doing deep learning as a service and the exact count doesn't matter, but whether it's four or five acquisition offers, we've got a lot of acquisition offers. And you might say, yes, they're acquihire offers, but it's crazy. Like just the potential upside that people are teased with doing AI SaaS startups right now is just mind blowing. Just the hype that's out
06:41there for these companies that...
Nathan Latka
06:43Well, yeah. So I mean, no, I mean, no, you're right. The problem is, is it doesn't matter if you have the best team. If you don't know to communicate that you have the best team, it almost doesn't matter because the people who you think are the best team will then, like, become the best team because of... You know, they'll get more customers and they'll use that money to go buy real engineers.
Ben Taylor
07:01Yeah. Yes. Yeah. You definitely have to ship value, but owning the mic has been very helpful.
Nathan Latka
07:07Mhmm. What is the... So I mean, how aggressive are you being right now in terms of like burn? Are you operating at cash flow positive or or how much money are you guys burning each month?
Burn Rate and Hiring Philosophy
Ben Taylor
07:15We're we're being aggressive with burn because we really need engineers and the engineers we have to hire are not... They're all principal level engineers. We can't hire kids out of school and expect them to ramp on enterprise level deep learning, which requires big data needs, high performance computing needs. Yeah. So we're... Every new contract we close, we hire.
Nathan Latka
07:39Yeah. So mean, how much are you burning per month right now? Are we talking like $30,000 a month, a $100,000 a month?
Ben Taylor
07:46No. Like $30,000 a month.
Bootstrapping and Founder Sacrifice
Nathan Latka
07:47And who's filling that
Ben Taylor
07:47So
07:51the the founders have definitely, you know, paid, you know, we... We've kind of paid our our share.
Nathan Latka
07:57You just keep putting in more money?
Ben Taylor
08:00Yeah. But it's interesting because you talk to... I know you know some of the Utah startups pretty well, but talking to companies like Weave and Degreed that are doing really, really well right now. You... If you had talked to them three years ago, their VCs were telling them to close-up shop and founders were maxing out credit cards. So like, you know, we know what we signed up for.
Nathan Latka
08:20Yeah. By the way, I I I think I don't think there's anything wrong with that. I like founders that keep their companies and don't raise any VC. I think that's a smarter way to do it if you can manage burn.
Ben Taylor
08:29Yeah. Something that's funny when you talk to entrepreneurs is when you're an employee, you're a pet. When you're an entrepreneur, you worry about the payroll. Like, can I make payroll this month? And if I'm talking to you and you say, man, it's a hard month, I'm gonna go without pay this month. My my... Or my reaction is like, wow. Well, like, bravo. Like, wait until you do that for, you know, months and months and months. Like,
08:53wait wait until you do that a year before it means anything. Who cares about you? Like who cares about this?
Nathan Latka
08:58Do you have any non founders? I think all five are founders, right?
Ben Taylor
09:02No, we have two co founders.
Nathan Latka
09:04Okay.
Ben Taylor
09:04But
09:05we kind of see, four of us are seen as kind of the core leadership team. We've been together for... Four of us for a year now. And I think I talked to you briefly about some of the seniority, like our sales guy, he did 13 at Oracle, twenty eight years sales experience enterprise, our creative director came from Cellstack.
Nathan Latka
09:28Some would say that's a good thing and some would say, oh my god, it's a bad thing, have to unteach them all the bullshit Oracle stuff.
Ben Taylor
09:34Yeah.
09:37We've been happy, but
09:41it's interesting, like, VCs definitely have an opinion about everything.
Nathan Latka
09:45Is he one of the founders?
Ben Taylor
09:47He's not.
Nathan Latka
09:48Okay, do you guys pay him a commission on stuff or is he all equity at this point?
Ben Taylor
09:54It's a mix of both.
Nathan Latka
09:55Okay. So you do have him structure like a salesperson where it was kind of on target earnings. There's a kind of a commission plus base.
Ben Taylor
10:03Yeah. But he's he's getting more equity than maybe your typical VP of sales would would be.
Nathan Latka
10:11He's, you know, team member, what, four or five?
Ben Taylor
10:14Yeah.
Nathan Latka
10:15Yeah. Very cool. So I mean, what's the next step? Are you raising capital right now?
Raising Capital and Valuation Target
Ben Taylor
10:21So a few options. We do have term sheets that we're...
10:28Are gonna probably move forward with one of these. So yes, we are gonna raise capital right now.
Nathan Latka
10:31How much are you gonna raise money?
Ben Taylor
10:34We're hoping to raise 2,000,000.
Nathan Latka
10:35And what do you hope to get in terms of valuation?
Ben Taylor
10:3930 post.
Nathan Latka
10:40Okay. And how do you validate that besides just touting the resumes of the founding team, which it sounds like you guys have quite the resume?
Ben Taylor
10:49I... So, yeah, we've got the experience and the resume. I would argue that if people really understand our business, they they realize that we found we found something unique.
Nathan Latka
11:01You mean, we've talked a couple of times in speaker green rooms. My impression always when I leave a conversation with you is brilliant guy, but I don't know that they know how to build a mouse. Like, I don't know that they know how to be a slimy marketer and actually get people signed up for this thing. And I think you probably see that. And I almost... I feel like you even thinking about that to you would
11:22feel like, why do have to worry about this bullshit? If I build the best product, people will find it. That's what I worry about with your kind of company.
Customer-Driven Innovation and Product Development
Ben Taylor
11:29And there have been other AutoML companies that have developed in a vacuum. And that's why I really value the revenue model that we started with, that we've had paying customers from day one. We don't have all those customers today because some of them have been professional services. So we haven't developed our platform in a vacuum. The last two years has been very intimate customer feedback. And I would argue that our biggest innovations right now that
11:55make us competitive, those good ideas did not come from us. They came from customer interactions. They came from engineers. They came from people that don't know anything about AI, but they have an opinion
Nathan Latka
11:55Mhmm.
Ben Taylor
11:55About how they think it should behave.
Nathan Latka
12:08Yeah. Well, look, I I hope it works for you. Raising two on a 30 pre is basically selling 6% of your company, you know, basically pre revenue. Yeah. So I mean I
12:16mean
Ben Taylor
12:16It's 28 pre, 30 post. Not that that matters.
Nathan Latka
12:22Oh, okay. You're selling 7% of the company, right? Whatever.
Ben Taylor
12:25Yeah.
Famous Five: Books, Tools and Sleep Habits
Nathan Latka
12:26Very good. Well, listen, let's wrap up here, Ben, with the famous five. Number one, what's your favorite business book?
Ben Taylor
12:33The Lean Startup from Eric Ries.
Nathan Latka
12:35Number two, is there a CEO you're following or studying?
Ben Taylor
12:40My co founder.
Nathan Latka
12:41Number three, what's your favorite online tool for building your company?
Ben Taylor
12:47LinkedIn.
Nathan Latka
12:48And number four, how many hours of sleep you get every night?
Ben Taylor
12:52Used to be three or four, polyphasic sleep. And I think it's more now. I'm trying to sleep more.
Nathan Latka
12:59Yeah. Yeah. Obviously So sick. You don't survive on three hours of sleep every night without at some point catching up.
Ben Taylor
13:06I I've I was doing polyphasic sleep for two years solid, but
Nathan Latka
13:12What is polyphasic sleep? Have no idea what that is.
Ben Taylor
13:14Oh, so I go to bed at eleven, wake up at two to 3AM depending on how stressed I am for work to be done. And then I need to take two twenty minute naps during the day.
Nathan Latka
13:24Oh, got it. It's like... It's almost like what they call like superhuman or whatever, or it's like... But that's extreme.
Ben Taylor
13:31And they have one more extreme. Called Uberman. You abandon all core sleep. You only sleep twenty minute naps for the rest of your... Like every three hours basically. Right? Yeah. It's super disruptive. The one I did where I had two naps, I could get away with it without impacting my family. Because I take a nap before they wake up and I take a nap before I go home. And so honestly, they may not even know that
13:54I'm
Nathan Latka
13:54So
Ben Taylor
13:54doing
Nathan Latka
13:55So okay, so what's your situation? Sounds like married, how many kids?
Ben Taylor
14:01Married, three kids.
Nathan Latka
14:02Okay, and how old are you?
Ben Taylor
14:04I am 36, but I've got a lot of gray in my beard, and that's all from the hedge fund.
Nathan Latka
14:10There you go. Alright. And last question. What do you wish your 20 year old self knew?
Ben Taylor
14:18Oh, sorry. My my Internet broke up.
Nathan Latka
14:20That's okay, Ben. Do you
14:21your year
14:22old self knew?
Ben Taylor
14:25That what you teach yourself is probably more important than what you learn in school.
Nathan Latka
14:32Guys,
Ben Taylor
14:32what you
14:32teach I'm a chemical engineer. I'm a chemical engineer. That's what I went to school for. And now I'm an AI.
14:40You know, I live and breathe AI. It has nothing to do with chemical engineering.
Nathan Latka
14:43Guys, there you have it. AI machine learning for insurance companies and anything where there's an audio video kind of risk assessment. Is there a target market three customers paying $10,000 a month right now, $30,000 a month in revenue? Really ramp... You know, focus on scaling here, trying to find a product market fit. They're bootstrapped burning about $30,000 a month. Founders are obviously funding that loss. Team of five right now, again, as they look to scale, hoping
15:03to raise $2,000,000 on a 28,000,000 pre money valuation here in the next couple of quarters. Ben, thanks for taking us to the top.
Ben Taylor
15:09Okay. Thanks, Nathan.