Latest / AI for Founders with Ryan Estes / From blind vision to billion-dollar AI: Alberto Rizzoli’s startup journey

Transcript

From blind vision to billion-dollar AI: Alberto Rizzoli’s startup journey · AI for Founders with Ryan Estes

• 0:02
  Welcome back to AI for Founders. I'm your host Ryan Estes and
• 0:05
  today's guest is the real deal. Alberto Rizzoli is the Co
• 0:09
  founder of the Seven Labs, a platform building next Gen.
• 0:13
  AI agents that tackle the soul sucking admin work clogging up
• 0:17
  modern companies so you can actually get back to building.
• 0:20
  We talk about what it means to really automate your back
• 0:23
  office, the surprising struggles of European founders breaking
• 0:26
  into U.S. markets and why most successful companies are
• 0:29
  secretly built on pivots from a viral AI for good project
• 0:33
  helping the blind to a $33 million raise during one of
• 0:37
  ventures coldest winters. AL's is a story that's part Ted
• 0:41
  talk, part gladiator training, and if you've ever wondered how
• 0:44
  to future proof your startup Al Shay, Al shares the single
• 0:49
  question every founder should ask.
• 0:51
  What product would kill my business?
• 0:53
  Then go build that. All that more.
• 0:55
  Let's go. But first, some business.
• 0:58
  This episode is brought to you by Kick Caster, the podcast
• 1:00
  booking agency trusted by over 800 top founders.
• 1:04
  If you're ready to build trust at scale, unlock new revenue,
• 1:07
  and turn your story into your best performing marketing asset,
• 1:11
  podcast interviews are the fastest way to get there.
• 1:13
  Kick Caster gets you booked on shows that matter to your
• 1:16
  audience. You bring the inside.
• 1:18
  They handle the strategy, the outreach and the placement.
• 1:21
  Whether you're raising capital, whether you're raising capital,
• 1:24
  hiring top talent OR expanding your brand, your voice needs a
• 1:28
  stage and podcasting. That's the stage with no
• 1:31
  distractions, no scrolling, just your message loud and clear.
• 1:35
  Visit kickcaster.com to apply Kickcaster podcast interviews
• 1:39
  done for you. That's Kickcaster by Moberst.
• 1:44
  OK, let's be honest, startup life is basically building a
• 1:47
  plane while it's flying during a thunderstorm and half your team
• 1:50
  is remote in three time zones. Does that sound familiar?
• 1:53
  Then you need Code Story. It's the podcast where tech
• 1:56
  founders get real about what actually takes to build a
• 1:59
  product, survive launch day, and scale without losing their minds
• 2:03
  or their engineers. Hosted by NOAA Lab Heart, Code
• 2:06
  Story gives you the behind the scenes of tech startups, the
• 2:09
  sleepless sprints, critical bugs, breakthroughs you didn't
• 2:12
  see coming, fires you definitely didn't see coming.
• 2:15
  If you're a founder or builder, Code Story isn't just relatable,
• 2:19
  it's practically mandatory listening.
• 2:21
  Search for Code Story wherever you get your podcasts or head to
• 2:24
  codestory.co. Because building great tech
• 2:27
  isn't a fairy tale, it's a survival sport.
• 2:30
  And these are the stories that made it to the other side.
• 2:32
  Code story.co. What are
• 2:54
  you doing? I'm doing great.
• 2:56
  Thank you. Great to be here.
• 2:57
  How are you? 100% we, we, we caught up a
• 3:00
  little bit before the, the interview.
• 3:01
  And so I know you're in the sweltering heat of London right
• 3:04
  now. It's 100 plus degrees
• 3:06
  Fahrenheit. So that's what you said, 31
• 3:08
  Celsius. 31 Celsius, it's a record over here.
• 3:10
  London empties out whenever it gets hot, so just outside of the
• 3:15
  window there's probably 100 people at the pub getting a
• 3:19
  refreshing beer. So the EU KG DP is about to get
• 3:22
  a major hit. It's great, you know, nothing.
• 3:27
  Nothing knocks down the heat like a cold pint, you know.
• 3:31
  Yeah, yeah, the, the British know this very well.
• 3:34
  It's it's actually hard to to do business if you if you don't
• 3:38
  drink in the UK, which isn't exactly fantastic, but but it
• 3:43
  does lead to some fun outcomes every now and again.
• 3:46
  Amazing, what got you to London? Are you?
• 3:50
  Are you HQ in London? We are yes.
• 3:53
  So we're in London and New York. The sevens HQ is an Oxford
• 3:56
  Circus in London. And we started the company here
• 4:00
  because we wanted to create a European AI company that would
• 4:03
  rival the Silicon Valley ones. And we know there's a lot of
• 4:07
  incredibly high talent in Europe that we could leverage.
• 4:10
  And my Co founder also was about to have his first child in in
• 4:14
  London. And so we, we thought it'd be
• 4:15
  the best place to, to, to get started.
• 4:19
  On the contrary to that though, most of our growth happens in
• 4:22
  the US And I, I think maybe for founders listening, my best
• 4:26
  advice seems to be start your company anywhere, but consider
• 4:30
  your market to be the US no matter where you are, especially
• 4:32
  if you're a software company. So we we see about 70% of our
• 4:37
  revenue coming from the US and the average sales cycle for US
• 4:41
  account is generally close to half of what it takes for us to
• 4:45
  close in Europe for the same amount of revenue.
• 4:49
  Amazing. Give us a quick overview of your
• 4:52
  company V7 Labs. Yeah.
• 4:54
  V7 is an AI company that helps companies build AI agents that
• 5:00
  automate the most tedious and repetitive work that we see
• 5:03
  every day. This is all the administrative
• 5:06
  office work that large businesses and small businesses
• 5:10
  have to deal with everyday rather than the interpersonal
• 5:13
  and creative work that we really want to do.
• 5:15
  And these agents are able to ingest any form of unstructured
• 5:19
  data, whether it's documents, leases in the real estate world,
• 5:23
  or investment, investment opportunities and private
• 5:27
  equity. And then through reasoning
• 5:29
  steps, through step by step agent steps, they're able to
• 5:32
  turn them into structured data. On the other side, a
• 5:34
  spreadsheet, a finalized IC report, a status of every
• 5:41
  property that a company might own.
• 5:43
  And effectively, we're eliminating the concept of a
• 5:46
  back office that keeps people in front of Excel or Word for hours
• 5:50
  on end, and allowing people to delegate this work to AI so they
• 5:53
  can focus on things that are actually more fun and productive
• 5:56
  for the business. Cool.
• 5:58
  So when you're talking about agents or you mean kind of
• 6:00
  specifically organizing data in spreadsheets and then creating
• 6:03
  customized dashboards, is that kind of the gist of it or?
• 6:07
  Yeah, the way you see an agent is an AI that also has the the
• 6:10
  the arrow of time. So whilst working with ChatGPT
• 6:14
  you might give it a document then get a response right away.
• 6:17
  And the AI agent is used for something to requires more than
• 6:20
  one thinking step in order to complete a task.
• 6:23
  If you give, for example, a financial model to Chachi PT or
• 6:27
  rent roll and you ask it to a single prompt, it might not
• 6:31
  always give you either a consistent answer or an accurate
• 6:33
  one. And an agent effectively looks
• 6:37
  at the document and says, hey, OK, I need to evaluate this
• 6:42
  investment opportunity or I need to evaluate this, this legal
• 6:45
  claim or the, this tax claim. Step one, I'm going to 1st
• 6:50
  extract all of these values. Step 2, I'm going to look at my
• 6:53
  company's guidelines to see whether any of them go against
• 6:57
  the, maybe the, the tax claims urgency requirements.
• 7:01
  Step three, I'm going to see what I need to do with this tax
• 7:04
  claim and who I need to notify. And by following these steps,
• 7:07
  one by one, AI becomes a lot more accurate in performing
• 7:10
  them. And effectively, it acts like an
• 7:12
  internal employee. And little by little, companies
• 7:15
  can create many different agents for many different tasks so that
• 7:19
  everything that is considered administrative can be delegated
• 7:21
  to AI. And the way we do it is we we
• 7:25
  have an additional agent called Concierge, where users can
• 7:28
  effectively route any task. They can e-mail it, or they can
• 7:31
  drag and drop it in their screen or or mention them on Slack, and
• 7:35
  they will effectively delegate the this type of work to the
• 7:39
  correct agent based on the nature of the request.
• 7:41
  So it's effectively like a chief of staff that works for all of
• 7:45
  the employees, and every employee has a network of agents
• 7:49
  below them that they can delegate work to.
• 7:53
  Amazing. It seems like N8 N is kind of
• 7:56
  building traction in this space and I'm using it for different
• 7:59
  workflows of like coming from let's say a slack trigger that
• 8:04
  pulls it from air table that sends to GPT for a prompt that
• 8:08
  publishes through lovable things like that.
• 8:10
  How does the tool kind of differentiate from N8 N?
• 8:14
  Yeah, there's a lot of really great workflow automation tools
• 8:17
  like Zapier and Aden that allow you to do effectively API to API
• 8:21
  connection and automate entire processes of creation.
• 8:24
  B7 Goes specifically focuses on document understanding a lot
• 8:29
  better than than other tools. And we deal with documents that
• 8:33
  can be pretty complex. Sometimes there are hundreds of
• 8:35
  pages long, sometimes they contain financial information
• 8:38
  that needs to be recalculated and thought through, sometimes
• 8:42
  they are excel sheets with multiple tabs.
• 8:44
  And so the companies that use V7 tend to be tax giants like KPMG
• 8:50
  or private equity firms or banks, Oregon fintech companies
• 8:53
  that are processing lots and lots of documents for their own
• 8:56
  customers and user API to understand them at scale.
• 9:00
  The user experience is familiar to what you might see with an 8
• 9:03
  N, but the elements of the the tool are a lot more focused on
• 9:07
  extraction of data from multiple documents and then retrieval of
• 9:12
  rule sets and and evaluation guidelines and then completion
• 9:16
  of that work in the form usually of a structured output, so
• 9:19
  another document or sometimes ACSV output or adjacent output.
• 9:24
  Cool. And, you know, people uploading
• 9:25
  their financial information, what's the guardrails and kind
• 9:28
  of like privacy for the tool? Yeah, we spent a lot of time
• 9:32
  trying to remove hallucinations entirely from the the the
• 9:36
  product. And they're prevalent because AI
• 9:39
  is very keen on giving you an answer, even if it doesn't know
• 9:42
  the actual answer to the question.
• 9:45
  And so we built a system called Visual Grounding that allows any
• 9:50
  model that you use with MV7 go to always look for the exact
• 9:55
  location of an answer that it's giving you in a source document,
• 9:59
  whether it's from the web or it's from your internal
• 10:01
  knowledge base. And it's actually pretty complex
• 10:04
  because every document needs to be scanned with computer vision
• 10:07
  models to understand where all the text is present.
• 10:10
  And then the model takes a little bit more effort to go and
• 10:12
  find exactly that source. But it works a lot better than
• 10:15
  just using RAG, which will give you sometimes a document
• 10:18
  reference, sometimes a page reference.
• 10:20
  But even RAG can sometimes hallucinate an answer.
• 10:23
  And so we wanted it to be super precise because very few of the
• 10:27
  work flows that we actually operate within are only end to
• 10:32
  end AI related. There's always there, there's
• 10:35
  usually a small element of human review, for example, in
• 10:39
  evaluating a term sheet for an investment, pretty high stakes
• 10:43
  operation. So you want it to be super
• 10:45
  accurate and you want to review the output of BI really easily.
• 10:48
  And so clicking on an answer by the AI and usually an agent will
• 10:52
  give many answers to its response.
• 10:54
  You know, it's, it's extracting a bunch of information.
• 10:56
  So you're looking at maybe 12 different data points you need
• 10:59
  to verify. You want to click on any of
• 11:01
  these data points and see immediately where in the
• 11:04
  document that's been looked at and that's been sourced from, so
• 11:07
  you can verify that the the information is actually correct.
• 11:11
  Amazing. I'm just imagining all the
• 11:15
  $500.00 an hour attorneys that you're like hey, hey, hey, use
• 11:18
  V7. I want to take that 6th hour
• 11:20
  bill down to one hour. Let's go.
• 11:23
  Yeah, cool. I, I, this is great context for
• 11:27
  your new project, doing a lot of research on you and I, I want to
• 11:31
  thank you because you really take kind of founder LED
• 11:33
  marketing to the next level. There is so much great
• 11:37
  information about you online. I watch your Ted talk, read your
• 11:41
  blog. You know, there's, there's tons
• 11:43
  of stuff all over the place. Makes my job a lot easier.
• 11:46
  One thing I wanted to start with is you received the National
• 11:50
  Genteel Prize for Science and Innovation from the President of
• 11:54
  Italy. Tell me about that.
• 11:56
  Yes. So my previous startup was
• 11:59
  called ipoly and this was 2015, early days of AI and we were
• 12:03
  building an engine that could run convolutional neural
• 12:07
  networks. So the computer vision models
• 12:09
  that were being used before LLMS or LLVMS directly on the CPU of
• 12:13
  smartphones. And at the time it was the
• 12:15
  iPhone 5, which was very underpowered compared to today's
• 12:18
  phones. So I have to quantize these
• 12:20
  models and build a custom engine that could be very, very
• 12:22
  efficient. And her goal was to help the
• 12:25
  blind and visually impaired identify the world around them.
• 12:28
  I, I grew up with a family friend who became totally blind
• 12:31
  due to a hunting accident. And I remember, you know,
• 12:35
  people, especially people that become blind later in their
• 12:37
  life, they still want to build a mental model of the places that
• 12:40
  they've been to, to build memories and to recall them.
• 12:44
  And so he wanted to know when we're rocking around, what
• 12:48
  objects are around them, what the scenery looked like.
• 12:51
  And that's not something that a guide dog can provide to you.
• 12:53
  It's only something that an intelligent communicating
• 12:56
  creature can do. And so we thought, what if AI
• 12:58
  could actually do that? And we built a, an AI model with
• 13:03
  trained on 5000 individual classes.
• 13:06
  It was a really big effort to collect that data and then
• 13:10
  eventually was able to to understand all of these these
• 13:13
  objects to to various degrees of accuracy.
• 13:17
  The challenges there is that the accessibility is not a very easy
• 13:21
  market to turn to create a venture backed company and that
• 13:25
  that eventually went to IPO. So that was challenge number
• 13:28
  one. Challenge number 2 was that we
• 13:31
  were really loved by our availability of data and
• 13:33
  producing training data is actually quite expensive.
• 13:36
  It requires humans to label images and explain what's inside
• 13:40
  the image. And this was this, this is still
• 13:44
  a problem today. And actually, when when we were
• 13:47
  building I Poly and and it had this enormous viral appeal and
• 13:50
  was a a true example of the I forbid.
• 13:52
  So it won a lot of awards. Part of the inspiration of the
• 13:57
  seven also came from that. So, but before we launched the
• 13:59
  seven go the a agents platform, our first product it it was
• 14:04
  called V7 Darwin and it's used by a researchers to build and
• 14:08
  teach new things to AI models and it effectively is used to
• 14:13
  label data reader really efficiently.
• 14:14
  So rather than having agents that understand documents at
• 14:19
  scale and read information within documents, it would read
• 14:22
  images, documents and video and pre complete some of the
• 14:25
  teaching steps that human need to effectively build AI.
• 14:30
  Now we're in an era, we're in a world in which AI is already
• 14:33
  capable of no longer needing to learn a lot of things, but just
• 14:36
  getting on with the job and just being told what instructions and
• 14:38
  needs to follow. And that led us to creating an
• 14:41
  agent platform that launched about a year ago.
• 14:44
  Amazing. Now so did was was V7A pivot
• 14:48
  from AI opoly or was that did you shelf that?
• 14:52
  It, it, it, it's an entirely different company.
• 14:55
  So I probably was based in in San Francisco.
• 14:58
  V7 is in London. We did have some folks, some
• 15:03
  excited Polly people come in to come in to B7.
• 15:05
  Our, our very first hire is actually from from the previous
• 15:08
  company Andre Tzini, who's our founding engineer.
• 15:12
  But it was a different market and different product type and
• 15:15
  we went from what was basically AB to C application to AB to B
• 15:19
  model. And, and I think that if you
• 15:23
  look at our product, there's a strong UX focus to everything
• 15:28
  and a strong presentation focus on everything.
• 15:30
  And I think it's really in our DNA.
• 15:32
  Simon Edwardson and myself are B to C founders.
• 15:36
  Do we? We like people that love our
• 15:39
  love using our product. And so we, we, we still put a
• 15:42
  lot of effort into that. Very cool.
• 15:46
  It's it's very hard for entrepreneurs and founders.
• 15:49
  Maybe people listen to the show to know when it's time to change
• 15:53
  and then what that process is like.
• 15:56
  So walk me back to that time where you're like, you know
• 15:58
  what, I'm going to take a new company in a new direction.
• 16:02
  How did you know when it was time to end it?
• 16:04
  And then what was the transition like as you were gearing up for
• 16:06
  V7 I? Think almost every company
• 16:09
  before they hit, well, this successful product market fit.
• 16:14
  We'll have dozens of pivots, you just don't hear about them.
• 16:18
  But almost every gate company has seen a lot of pivots before
• 16:23
  they actually hit PMF. Even some of the success stories
• 16:25
  that that are Live Today and and really rocking it and it usually
• 16:31
  there, there's a couple of forces against you.
• 16:33
  One is if we invest, if we try to get more traction, if we
• 16:37
  invest more time into this project, will we actually get
• 16:40
  there. And there's many products that
• 16:42
  actually give up before they hit PMF, but they're really on the
• 16:46
  route to that. There is just some point of
• 16:48
  friction, some usability issue, some accuracy issue that doesn't
• 16:52
  get them there. And then there's the other side
• 16:55
  of things which is actually you're operating in in the wrong
• 16:58
  market. And that was the case for Ipauli
• 17:00
  where even though the even if the app would have been perfect,
• 17:03
  accessibility would have never created the correct returns for
• 17:09
  what we're actually building from an entity perspective,
• 17:11
  which was a venture backed company that eventually was
• 17:13
  going to have to multiply its revenue very, very fast.
• 17:16
  And so it would have probably been better off as an open
• 17:19
  source project that others would have contributed to.
• 17:22
  And I think the the difficult choice that needs to be made is
• 17:28
  can be mitigated in a couple of ways.
• 17:30
  One is making sure that your market is large, but another one
• 17:33
  is making sure that your market is growing really, really fast.
• 17:37
  An example that's relevant to V7 is that the market for data
• 17:44
  labeling is growing fairly fast, but the market for computer
• 17:48
  vision data labeling isn't growing really as fast as the
• 17:51
  development of AI agents, this part of RND, of AI.
• 17:55
  And so every now and then, especially for AI companies, you
• 17:58
  need to rethink what you're building.
• 18:02
  And a good way to do that is asking yourselves every six
• 18:06
  months. And this is something that we do
• 18:08
  with ND 7 as well. What is the product that if it
• 18:11
  were to come out today would kill my company and you scope it
• 18:15
  out and then you build that product.
• 18:18
  And in the case of of V7 Go, we had a lot of customers that were
• 18:23
  developing computer vision models for document
• 18:26
  understanding for the insurance market for banking.
• 18:29
  And they were spending months developing bespoke models that
• 18:32
  could understand the document template and a document type for
• 18:35
  logistics or for banking. And when GPD 3 came out, we
• 18:40
  realized, hey, I think LLMS plus OCR at the time LLMS couldn't
• 18:45
  actually see things are are going to replace a lot of this
• 18:50
  labeling work that needs to be done for documents.
• 18:53
  And so we sacrificed a double digit percentage of revenue to
• 18:57
  create something that eventually was going to go straight to
• 19:01
  either the customers or newer companies that were going to
• 19:04
  just start using these, these LLM focused workflows to, to
• 19:08
  solve document tasks. And it's always super painful,
• 19:13
  but in the long run, it's, it's almost always the best decision
• 19:18
  ever. And in a way we, when we were
• 19:21
  developing Go, which again launched a year ago and, and D7
• 19:24
  is 6 years old. So almost five years into the
• 19:27
  company, we had to create almost a startup within the startup to,
• 19:32
  to, to create something brand new, almost like a lot of things
• 19:35
  were, were being developed from scratch.
• 19:38
  And this is something that like a lot of startups don't say to
• 19:41
  investors, but AI requires a lot of rebuilding from the ground
• 19:45
  up. Most self driving car systems
• 19:48
  have been rebuilt almost from the ground up as since based on
• 19:52
  what they were five years ago. And the the difficult thing is
• 19:57
  that when you when you're a five person startup or ten person
• 20:00
  startup, you have this ungodly amount of energy and sacrifice
• 20:04
  that goes into it when you're a 50 or 60 person startup.
• 20:08
  People maybe might not work the weekends as as often as as it
• 20:13
  happens, or they might not. Stay up late or the work days.
• 20:16
  Yeah, or over the work days. And so we're we're definitely in
• 20:21
  a world that is seeing companies with a shorter lifespan than
• 20:25
  than 10 years ago or 20 years ago.
• 20:28
  And I think CE OS and founders need to think of ways of like re
• 20:32
  injecting that energy back into the company.
• 20:35
  And a good metric for that is if you're not building something
• 20:38
  super disruptive today, if you're just like adding features
• 20:41
  that are incremental to whatever you're building today, you might
• 20:46
  be at risk. You might it might be time for
• 20:47
  you to start thinking what is the product that will kill my
• 20:51
  product if it were to come out today and and start building it.
• 20:55
  Love that. That's amazing advice.
• 20:58
  I want to take you a couple back a couple years.
• 21:01
  So 2022, I saw some numbers you were reporting that ARR tripled.
• 21:06
  Not only that, but you did a $33 million raise.
• 21:09
  And if people don't remember 2022, man, that you couldn't pry
• 21:14
  a wallet open with a crowbar from VC, how did you do that?
• 21:18
  What was the presentation like? It might be the only like larger
• 21:22
  raise that I've seen from a company in that time period.
• 21:27
  Yeah, it was a really tough time.
• 21:29
  The 2021 was was incredible for for fundraising.
• 21:34
  And then ironically 2023 was also easy 22 was was pretty
• 21:38
  hard, brutal. The we, we had really good
• 21:43
  traction and we had really good retention and we had a product
• 21:47
  that was very differentiated and we had a lot of example stories
• 21:52
  of customers that were turning their internal platform or some
• 21:56
  competitors to be switching to us.
• 21:58
  And on top of that, this was right before maybe the AI craze
• 22:04
  that is that has been happening now.
• 22:06
  But our investors did believe that there was going to be
• 22:10
  something around the corner that was going to make AI avoid a
• 22:14
  winter. AI has had a few winters in the
• 22:16
  past. And right now I think it's
• 22:18
  clearly and it's spring and it's, it's finally making its
• 22:20
  way outside of the research lab and into everyone's phone and
• 22:24
  into dinner conversations that you hear on a daily basis.
• 22:28
  How do we actually do it? I think there is mirrors of
• 22:32
  different things that that we focus on.
• 22:33
  We, we, I wrote a blog post on this.
• 22:35
  We kept a very well maintained CRM for investors.
• 22:41
  At the end of the day, fundraising is sales.
• 22:44
  There's a Sankey diagram of all the rejections that we received
• 22:47
  and all the offers that we received.
• 22:49
  And there's obviously a lot of rejections, a lot of funds stop
• 22:54
  deploying money. Also some very big names at the
• 22:57
  time, we're just not deploying anything.
• 22:59
  A lot of partners getting fired and, and head count productions
• 23:02
  even on the on the BC side. So it was difficult and the more
• 23:08
  systematic you can get with fundraising and not an OV CS
• 23:12
  wanted to build relationships with the with them and, and
• 23:15
  invest time into the relationship.
• 23:16
  And that is really important. But any great salesperson will
• 23:20
  tell you that you still have to create a lot of pipeline and you
• 23:24
  have to do it very systematically if you want to be
• 23:25
  successful. So the founder, you should just
• 23:28
  create your pipeline, do it systematically, book some time
• 23:31
  in your calendar to do so, catch up with investors very briefly,
• 23:35
  and then create a killer deck that shows a very opinionated
• 23:39
  point of view. Not just another workflow
• 23:44
  automation tool, but one that has a a a very specific and
• 23:48
  unique focus and a unique angle. Because they will probably have
• 23:51
  seen 3 versions of your company before, even if it's really
• 23:55
  creative. So like, be very opinionated
• 23:59
  about what makes you different. Love it.
• 24:02
  What percentage of your time now is kind of investor relations?
• 24:06
  Still very small. I think when, when you're not
• 24:09
  fundraising, you should give us actually as little time as
• 24:12
  possible to investors, be friendly and maintain a
• 24:15
  relationship. But I would say I, I, I block
• 24:19
  off an hour and a half every Friday for what I call investor
• 24:23
  CRM and communication. And sometimes it's angels,
• 24:28
  sometimes it's it's investors. And it it, it can feel like a
• 24:34
  slog because you are effectively investing time into something
• 24:38
  that has uncertain returns. But it kind of has to be done.
• 24:43
  You bet. Are you comfortable with the
• 24:46
  sharing and economics about the business as far as like what
• 24:48
  your ARR is? We don't share these outside of
• 24:52
  kind of fundraising milestones because otherwise we get on on
• 24:57
  on the tracker. But I'm happy to share non
• 24:58
  financial numbers if that's helpful.
• 25:01
  Cool. Yeah, team size, customer
• 25:03
  numbers, that kind of thing would be great.
• 25:05
  We are around 80 people and we out of which 10% of them are in
• 25:12
  our newer New York office and then the rest are in Europe and
• 25:17
  we have about 350 customers at the moment 3. 150 customers,
• 25:23
  Amazing. And you said most of those are
• 25:25
  in the states now. Was that intentional?
• 25:27
  We should go to market to the States first.
• 25:30
  Yeah, actually this it, it was not we, we founded V7 to make it
• 25:35
  sort of a a strong company in Europe.
• 25:37
  But we find that European companies move a lot slower and
• 25:42
  they tend to have smaller budgets for AI project product
• 25:46
  projects and even some really big names that we have that are
• 25:50
  technically UK headquarters. The sale happened to AUS team
• 25:54
  that is in charge of ML And so it's usually best to consider
• 25:59
  the US to be your market anywhere you are.
• 26:02
  And yeah, so it, it, it was not expected and it does require a
• 26:07
  few later Zoom calls with customers, but it's, it's
• 26:10
  definitely the better way to grow.
• 26:13
  Yeah, 100%. Eighty folks to manage that's
• 26:16
  that's a pretty big number for startup.
• 26:18
  Do you feel good about that? I, I do and I think that one
• 26:23
  trend that's happening is that companies are a lot more
• 26:25
  efficient with fewer people than they were a few years back.
• 26:28
  Middle management seems to be disappearing as a, as a concept,
• 26:31
  because there are more tools to just manage people at scale and
• 26:35
  manage work at scale. There is the, you know, we, we
• 26:41
  switched from web flow to framer recently just for, for our, our
• 26:45
  marketing website. And that's one example of how a
• 26:48
  tool has reduced the amount of time that our marketing team
• 26:51
  needs to spend on just managing one pretty important component
• 26:55
  of the business and freeing up time for other things.
• 26:58
  So I, I think the, in 2021 we were seeing head count
• 27:03
  explosions within companies and in part because of what interest
• 27:07
  rates, in part because it was the best thing for business or
• 27:10
  assumed to be the best thing for business.
• 27:12
  Today we're seeing the opposite. We're seeing the venture
• 27:15
  capitalists also being attracted to AI companies that have like a
• 27:19
  cap on the people that they have until they hit a specific
• 27:22
  milestone. And we're seeing some incredibly
• 27:24
  revenue efficient businesses that have really only a couple
• 27:28
  of handful of people and are making the same amount of money
• 27:31
  that 100 or 200 person startup was making in 2020.
• 27:35
  So that landscape has changed a lot.
• 27:38
  Notwithstanding though, there's still a big ghost within the AI
• 27:42
  world, which is churn and retention.
• 27:45
  And there's a lot of AI successes that are getting
• 27:48
  traction initially, but then people stop using the tools or
• 27:52
  they haven't successfully deployed anything in production,
• 27:55
  and then they turn after that. And the only really good way to
• 27:59
  do that is actually having people helping out customers.
• 28:03
  So in the form of solution engineering or customer success
• 28:07
  for actual series work being done with AI and even though
• 28:11
  there's there's leaps and leaps being done on user experience
• 28:14
  with AI, like you mentioned lovable, it's kind of easy to
• 28:17
  use. Getting that to become a tool
• 28:21
  used every day within an organization takes a little bit
• 28:24
  of effort, and unfortunately that has to be human effort.
• 28:28
  Yeah, 100%. To your point, you mentioned
• 28:33
  that AI is in a massive hype cycle right now, which is great.
• 28:37
  Begin it's it's well deserved. It's so cool.
• 28:39
  How do you see that resolving? I think we're at the start of
• 28:45
  its impact. A lot of people are using it for
• 28:50
  like the reality. If you look at the world
• 28:52
  economy, there's, there's a lot of attention to where I can be
• 28:58
  used and there's a lot of great prototyping being done with AI,
• 29:02
  but there's not that much AI being run inside hospitals or
• 29:07
  government today. And so we're, we're yet to see
• 29:11
  the economic impact be realized pretty soon.
• 29:14
  And, and so for, for that, we're at the start, there's some
• 29:18
  plateauing performance that we're seeing on, on new models
• 29:21
  being released. But it's not that big of a worry
• 29:24
  because even just with the breakthroughs that we have today
• 29:26
  with sort of the transformer model being king, there's still
• 29:30
  enormous gains to be had in all white collar work and, and
• 29:35
  coding that are still unrealized.
• 29:38
  So there's a lot of growth to be done.
• 29:40
  But there is one scary, scary issue within startups that is
• 29:45
  sort of like what YC might have asked you 10 years ago, which is
• 29:49
  why, why wouldn't Google just crush you?
• 29:52
  And nowadays it's like, why wouldn't any of the model
• 29:55
  providers just build this functionality in house?
• 29:59
  And the it's happening with media generation where four O
• 30:02
  can create some pretty good images.
• 30:04
  And even though it's not as good as say a major or something
• 30:07
  specialized, the majority of the consumer base will just say,
• 30:11
  Hey, this is good enough for me. And that might happen in voice
• 30:14
  where we, we have some great voice whistles being developed,
• 30:19
  but the bulk of people need good enough.
• 30:23
  They don't need excellent in some cases.
• 30:26
  And, and so it's important to focus on the people that need,
• 30:30
  if you're building a startup for now in AI, focus on the, they
• 30:33
  need the top 5% of, of that, that sort of quality of work
• 30:38
  being done rather than something that potentially a model
• 30:42
  out-of-the-box could do like retrieving a file from your,
• 30:45
  your drive, which is now becoming embedded within cloud
• 30:49
  and, and Chachi PT. Yeah, it's speaking to that.
• 30:53
  If if founders are listening right now, what AI tool would
• 30:57
  you build that would get you fastest to $1,000,000 annual run
• 31:01
  rate? You know, YC put out this call
• 31:06
  of building not software as a service, but basically service
• 31:10
  as a software. And I think that there's a very
• 31:14
  unserved opportunity right now in replacing old businesses that
• 31:19
  are doing work manually by doing that work, but using AI tools
• 31:25
  that already exist. An example of that is that we,
• 31:29
  we have some some great customers in, in processing tax
• 31:32
  claims, for example. But because they are large,
• 31:35
  they're still taking quite a significant amount of time for
• 31:38
  digital transformation side of things.
• 31:39
  And there's a lot of internal friction points in actually
• 31:42
  doing this work on a by by implementing AI into it.
• 31:48
  In the UK, most companies need to file an R&D tax claim every
• 31:53
  year and it's a big write up that is very manual.
• 31:58
  And just doing that by utilizing AI software and setting up the
• 32:02
  company to provide a service with a very little human effort
• 32:06
  and a lot of AI implementation is actually, I think an
• 32:09
  underserved slice of the market. Considering that almost everyone
• 32:13
  is building a SAS tool that is doing something with AI and
• 32:16
  there's only so many things that AI can do today, actually going
• 32:20
  to fulfilling that service end to end might be a good idea.
• 32:24
  And just using the great AI tools that are already being
• 32:26
  developed to do so. Love that services or software
• 32:33
  V7 sells tomorrow for a billion dollars.
• 32:36
  You just wipe your hands out of you're out.
• 32:38
  What do you do next? Probably take a one month
• 32:41
  holiday and then start another startup.
• 32:45
  You got your on anything particular?
• 32:47
  I don't know. I would, you know, I, I would
• 32:50
  love to, to, to work with BCI one day with bring computer
• 32:53
  interfaces. I would love to work on the next
• 32:57
  frontier. Really, I've always been
• 32:58
  attracted to the next frontier of technology.
• 33:01
  I think PCI is really interesting and I'm a novice in
• 33:04
  it, which excites me. And there's probably some
• 33:07
  Dunning Kruger effect in there, but but I would probably jump
• 33:09
  into it. There's the, you know, I'm
• 33:13
  probably not smart enough for the the protein side of, of AI
• 33:16
  discovery, even though I think it has truly enormous impact.
• 33:21
  And on the longevity side, it is a topic that I'm passionate
• 33:25
  about, but I don't take enough care of myself to actually be a
• 33:28
  good champion for it. So it'll probably be something
• 33:30
  that has to do with the human brain if I were to start
• 33:32
  something new. Well, maybe vision that seems to
• 33:35
  be a reoccurring theme in in your career in your.
• 33:39
  Business straight into the visual cortex, that would be.
• 33:42
  That would be nice. Yeah, the BC is, I'm absolutely
• 33:46
  fascinated with that. I've kind of put out a beacon
• 33:48
  through this podcast to like reach out to people because I'm
• 33:53
  very interested in this, particularly around, you know,
• 33:55
  triggering AI functions, AI tools, AI agents with kind of
• 33:59
  like underutilized body movements.
• 34:03
  You know? So like if I'm thinking about
• 34:05
  vision, it's like, you know, if I blinked 4 times in a row, like
• 34:09
  that should have an assigned task.
• 34:11
  Or if I wiggle my ears, you know, like I think there's going
• 34:14
  to be this really interesting interface that the fact that
• 34:18
  we're kind of doing this his never the ergonomics had never
• 34:21
  really fit human design, you know. 100%, yeah.
• 34:27
  My dream is that you could translate what's being processed
• 34:30
  in your visual cortex directly into pixels, so your stream of
• 34:34
  thought could actually be visualized.
• 34:35
  And I think that that is sort of the next unlocking of creativity
• 34:38
  after AI, because I'm a very visual learner and a visual
• 34:42
  person. And if I imagine something in my
• 34:44
  head, we only have two arms to turn it into reality.
• 34:50
  And it's a very inefficient way. Whereas with speech, almost
• 34:54
  everything that can be created into her head can actually be
• 34:56
  uttered by her mouth. With visual information it you
• 35:01
  need to be an artist to recreate that in reality.
• 35:06
  Of that, so good. If you were entering college as
• 35:09
  a freshman right now, with everything you know about
• 35:11
  business in the future, what would you study?
• 35:16
  You know, I have no idea for actually getting the right job
• 35:22
  because so many of them are in flux.
• 35:24
  I think computer science is still worth exploring.
• 35:29
  I think that fundamental STEM degrees like physics are also
• 35:35
  interesting or or mathematics because they train your brain
• 35:39
  into thinking a specific way. And ideally it should be
• 35:44
  something that goes to 1st principles because I think first
• 35:47
  principle thinking is still is still king in in creating
• 35:51
  something new today. And if you really understand
• 35:53
  something at first principles you can create a great version
• 35:55
  of it. If you don't, you can only
• 35:57
  create your surface level perception of what good looks
• 36:01
  like. So liberal arts first principle
• 36:05
  thinking like learning how to think first might be.
• 36:09
  I wouldn't. Yeah, absolutely.
• 36:10
  I wouldn't dig liberal arts. I, I, I think we, we don't have
• 36:14
  the, that in, in Europe, but, but I think that people should
• 36:19
  go, especially nowadays with where we're, we're eliminating
• 36:23
  the need for Nbas, middle management just kind of going
• 36:27
  away and, and AI is taking over a lot of some material science
• 36:32
  work, accounting work. University should be a place
• 36:36
  where you go for the joy of learning and to maximize your
• 36:39
  learning about topics and, and better ways of thinking more so
• 36:45
  than necessarily learning how to do a job.
• 36:48
  I would. I would rather do an
• 36:49
  apprenticeship for that instead or or start a startup and and
• 36:52
  learn by failing. Cool.
• 36:55
  We're just about a time. I just have a couple more
• 36:57
  questions. What are your favorite AI tools?
• 37:01
  V7 of course. V7 is a big fan of cursor.
• 37:06
  It's increased our productivity quite a bit, but it's probably
• 37:08
  one that gets gets spoken about a lot.
• 37:11
  We have a lot of fans of granola for recording meetings.
• 37:15
  It's it's very comfortable to use.
• 37:17
  It's one of those great single use tools like a potato peeler
• 37:21
  that you just wouldn't use anything else for.
• 37:24
  And the 7th is the third one. Any any kind of documents that I
• 37:29
  ever received doesn't have to go through my eyeballs and mostly
• 37:32
  goes through an agent. Love it.
• 37:36
  If you could live in any historical era except for now or
• 37:39
  the future, which would it be? I'm from Rome, so the, you know,
• 37:45
  seeing seeing the empire at its peak would be pretty
• 37:48
  interesting. It probably wouldn't be
• 37:50
  lucrative or lead to any particularly interesting
• 37:53
  discovery like going at the start of the universe and seeing
• 37:56
  The Big Bang, if that's even possible physics wise.
• 38:00
  But you know, it would be a nice it.
• 38:04
  It also depends on whether I'm stuck there or I'm able to go
• 38:07
  back to the future. So that that's, that's very
• 38:11
  dependent. And I assume that can only go in
• 38:13
  the past, but I, I wouldn't mind going to the Roman Empire and I
• 38:16
  could actually live the rest of my life there with what I know
• 38:19
  from the future and, and do pretty well for myself.
• 38:22
  So it could be fun, but. You know, there was a, a meme
• 38:27
  going around a couple years ago where women were finding out
• 38:30
  like how often men think about the Roman Empire on a daily
• 38:34
  basis, you know, and they're asking, I think they asked me.
• 38:37
  And I was like, I don't know, three or four times a day.
• 38:40
  How many times do you think about the Roman Empire in a day?
• 38:43
  I think it's hard not to if you grew up in Rome because so much
• 38:47
  of your indication is sent over on that.
• 38:48
  So if you see a piece of leafy bread, it's like, Oh yeah, this
• 38:52
  is what the Romans were, were, were creating or masonry or, or
• 38:59
  just a map. I, I don't know.
• 39:02
  I think it's a maybe, maybe the male brain has this slight,
• 39:06
  slight longing for, for empires and some degree of, of order.
• 39:10
  Sometimes that comes through her heads.
• 39:12
  I'm not exactly sure where it's coming from.
• 39:15
  It it also just looks cool. I think there's there's definite
• 39:17
  element to that. It's drip.
• 39:20
  I mean, you got all the red and like the the cool, like kind of
• 39:23
  skirts and leather and brass. Oh, it's cool.
• 39:27
  Do you have a favorite Roman? Augustus is kind of a classic
• 39:33
  response, but Marcus Aurelius, I think a lot of people would
• 39:39
  would have read Meditations it it's really nice.
• 39:42
  It's sort of like a personal diary and you don't often get to
• 39:47
  read the personal memoirs of a historical character without
• 39:52
  propaganda being basically flushed through it to improve
• 39:57
  it. And there's a lot of really
• 39:59
  deeply personal. Anecdotes there and, and sort of
• 40:03
  emotional writings that make you realize, hey, actually being
• 40:07
  king is kind of a slog sometimes and sometimes it's the same as
• 40:12
  being CEO. It's, it has its upsides, but
• 40:16
  really it's the, it's the path of someone that is usually
• 40:20
  continuously dissatisfied and trying to fix problems.
• 40:24
  And like Bill Gates once said, once you make your, your first
• 40:29
  million or 10 million, it's always the same cheeseburger.
• 40:32
  And, and there's something in, in humans that make us
• 40:35
  continuously adapt to the situation that we're in and that
• 40:39
  and then seek new problems to, to, to solve or feel new
• 40:44
  problems that we want to solve. It's a great book to that
• 40:47
  illustrates that. Absolute classic and it comes up
• 40:51
  a lot on this podcast. I talked to a lot of founders
• 40:53
  and stoicism in general seems to be like a great salve for our
• 40:58
  founders and CEO's largely. And maybe it's the intention is
• 41:02
  to take the, the peaks off the highs and take the bottoms out
• 41:05
  of the lows because there's, there's definitely that roller
• 41:09
  coaster effect. What what's helpful for you to
• 41:12
  kind of like keep emotions in check as things get really hard?
• 41:16
  It's like that old story attributed to King Solomon where
• 41:19
  he asked the ring maker to produce him a ring that will
• 41:22
  make him sad when he's happy and happy when he's sad.
• 41:25
  The ring maker goes all around the empire to go and and find
• 41:29
  one. And he does find 1, brings it
• 41:31
  back and there's a big party and King Solomon is very happy and
• 41:34
  the ring is given to him. And he looks at it and he frowns
• 41:37
  and the ring says this too shall pass.
• 41:40
  And it's it's the same case in a start.
• 41:43
  IK all the good moments they will pass and all the bad
• 41:46
  moments they will pass. And there will be a lot of good
• 41:48
  moments and a lot of really bad moments.
• 41:51
  And I think it's, it's just understanding that even even
• 41:55
  Steve Jobs got fired by Apple and he probably had a really
• 41:58
  tough time after that and that and then it passed and he
• 42:01
  bounced back and he created something amazing.
• 42:06
  It's for sharing that story. I'd never heard that before.
• 42:08
  It's absolutely incredible. Where can people find you
• 42:11
  online? Al?
• 42:12
  At Alberto Rizzoli on, on I guess LinkedIn, Instagram.
• 42:18
  I'm a pretty open book online. And if you're in front of a
• 42:22
  spreadsheet at any point later today, think about B7.
• 42:26
  It's B7, la-bs.com. Or just Google B7 and see if any
• 42:31
  agent can do some of that boring work for you.
• 42:35
  I'm doing it this afternoon. Al.
• 42:36
  This has been awesome. I will leave you to go get some
• 42:38
  ice cold pints. Amazing on the show.
• 42:42
  Thank you. Thanks for having me.
• 42:43
  That's this. Have a good one.
• 42:44
  Bet 2. This episode is brought to you
• 42:49
  by MO Burst, the AI first growth engine trusted by iconic
• 42:53
  startups like Calm, Robin Hood, Reddit, Uber and Google.
• 42:58
  Moburst is a digital, full service, mobile first marketing
• 43:01
  agency that helps companies to scale and become category
• 43:05
  leaders. Whether you're finding product
• 43:07
  market fit or ready to accelerate growth, Moburst
• 43:10
  delivers creative, data-driven solutions that help with scale
• 43:14
  and efficiency and impact. At Moburst, we're not just
• 43:18
  guessing what works. We're predicting it.
• 43:20
  As an AI first agency, Moburst uses cutting edge tools to
• 43:24
  analyze user behavior, adapt creative in real time, and of
• 43:28
  our hidden growth opportunities. The result?
• 43:31
  Marketing that's faster, smarter, and built to convert
• 43:34
  from early stage startups to global brands.
• 43:37
  We've helped hundreds of companies overcome complex
• 43:40
  challenges, unlock new audiences, and optimize every
• 43:43
  metric that matters. At Moeburst, we believe that
• 43:46
  creative thinking, advanced technology, and data are the
• 43:50
  core drivers of success at scale.
• 43:52
  If you're serious about scaling your business with precision and
• 43:55
  purpose, it's time to talk to Moeburst.
• 43:58
  Visit moeburst.com and let's build something category
• 44:01
  defining together.