Most founders still think AI is a feature. Roy Pereira thinks that’s a catastrophic misunderstanding.
The 5x founder and CEO of Unified believes we are watching the collapse of the entire SaaS operating model — not because software is disappearing, but because humans are no longer going to be the primary users of software.
That changes everything.
In this conversation on the Ignite Podcast, Roy laid out one of the clearest frameworks we’ve heard yet for where AI infrastructure, APIs, and enterprise software are heading over the next decade.
And unlike many AI commentators, Roy isn’t theorizing from the sidelines. He’s lived through multiple technology shifts already — from the dot-com crash to cloud computing to SaaS — and says this current transition is bigger than all of them.
“The cloud transformation was massive. This is 10x bigger.”
From Building Companies to Building Infrastructure
Roy’s entrepreneurial path started early.
He got his first computer in high school, became obsessed with hacking and programming, dropped out of computer science because he was bored, and started building companies instead.
Unified is now his fifth startup.
But the company didn’t start with a grand vision about AI infrastructure. It started with pain.
At his previous startup, Zoom.ai — an AI executive assistant platform built long before ChatGPT existed — Roy’s team built roughly 70 integrations internally.
That almost killed the company.
Not because building integrations was hard.
Because maintaining them was brutal.
Every API behaved slightly differently. Documentation rarely matched reality. Standards existed, but nobody followed them consistently. And once you support dozens of integrations, engineering teams slowly stop building product and start fixing edge cases.
Roy realized something important:
Every B2B software company was repeatedly solving the same undifferentiated infrastructure problem.
So when ChatGPT 3.5 launched, Roy and his team saw a much bigger opportunity.
AI wasn’t just going to need data.
It was going to need all the data.
In real time.
Unified’s Thesis: AI Runs on Customer Data
Unified is essentially infrastructure for AI-native software companies.
Instead of every startup building and maintaining integrations into Salesforce, HubSpot, QuickBooks, applicant tracking systems, CRMs, ERPs, messaging platforms, calendars, and internal tools — Unified handles the data connectivity layer for them.
Roy compares it to Plaid, but for B2B software.
And timing mattered enormously.
The original SaaS era mostly used APIs for onboarding data — moving customer information from one system into another.
AI changes the equation.
Now software agents need constant, transactional access to live business data to make decisions, generate outputs, and eventually take actions autonomously.
That requires a fundamentally different infrastructure architecture.
Roy believes most existing integration platforms were built for the previous generation of software — not for AI agents operating continuously in real time.
That’s why Unified exists.
The Real AI Shift Isn’t Chatbots — It’s the Death of Interfaces
One of the strongest ideas from the conversation is Roy’s argument that we’re entering a world where software interfaces become optional.
For decades, enterprise software was built around human workflows:
Dashboards
Forms
Menus
Spreadsheets
CRM interfaces
Admin panels
But if AI agents become the primary operators of software, most of those interfaces lose relevance.
“Software for humans is basically going away.”
That sounds extreme until you think about how quickly behavior is already changing.
People increasingly ask AI for answers instead of visiting websites.
They ask assistants to summarize documents instead of reading them manually.
They ask agents to draft emails, update records, or perform tasks directly.
The interface is becoming conversational.
Eventually, the human may disappear from the workflow entirely.
Roy believes software companies will increasingly compete at the API layer rather than the UI layer.
And if that happens, pricing models, product design, and company structure all change.
The Seat-Based SaaS Model Is Breaking
One of Roy’s most important observations is that traditional SaaS pricing no longer makes sense in an AI-native world.
Seat-based pricing worked because humans were the operators.
But if software agents do the work, what exactly is a “seat”?
This is why companies like Salesforce are already experimenting aggressively with usage-based AI pricing and agentic workflows.
Roy thinks this shift is inevitable.
The future is software that performs work autonomously — not software humans log into all day.
That means:
APIs become products
Usage replaces seat licenses
AI agents replace interfaces
Data pipelines become strategic infrastructure
And it also means the structure of startups changes dramatically.
Why Tiny Teams Will Build Massive Companies
Roy believes one of the most underappreciated AI shifts is operational leverage.
Unified’s engineering team today is dramatically more productive than engineering teams from even a few years ago.
Not marginally.
Order-of-magnitude more productive.
The combination of AI coding agents, automated research, faster iteration cycles, and infrastructure tooling means startups no longer need massive teams to execute.
That has huge consequences.
Historically, venture-backed companies raised large amounts of capital primarily to hire people.
Roy thinks that era is ending.
“I don’t think we should go back to the way we used to build companies.”
Instead, he predicts small, highly technical teams operating alongside fleets of AI agents.
In some cases, one-person or near-one-person companies may generate enormous enterprise value.
We are already starting to see early versions of this emerge.
The Bigger Risk: Society Isn’t Ready
Roy is optimistic about technology.
But he’s deeply skeptical that society is prepared for the labor disruption AI could create.
And unlike many AI discussions, he doesn’t frame this as science fiction.
He frames it as near-term economic reality.
Roy openly questioned whether he would hire software developers in the future at all.
Not because coding disappears entirely.
But because the role itself fundamentally changes.
This creates a larger societal problem:
What happens when millions of knowledge workers lose economic relevance faster than they can retrain?
Roy worries the transition speed matters more than the technology itself.
The Industrial Revolution took generations to unfold.
AI may compress equivalent labor disruption into a decade.
That creates political, social, and economic instability risks most governments are not yet treating seriously.
Why This Conversation Matters
Most AI conversations today focus on tools.
Roy is talking about systems.
He’s describing a world where:
software becomes autonomous,
APIs become economic infrastructure,
human interfaces become secondary,
and startups become dramatically smaller and faster.
Whether his timeline is exactly right almost doesn’t matter.
The direction already feels obvious once you see it.
And perhaps the most ironic part:
Roy’s previous company nearly collapsed under the weight of integrations.
Now he’s building the infrastructure layer that may quietly power the next generation of AI-native software companies.
The founders who survive this transition likely won’t be the ones building prettier dashboards.
They’ll be the ones building the plumbing underneath the agent economy.
👂🎧 Watch, listen, and follow on your favorite platform: https://tr.ee/S2ayrbx_fL
🙏 Join the conversation on your favorite social network: https://linktr.ee/theignitepodcast
Chapters:
00:01 — Roy Pereira’s Origin Story & First Computer
03:12 — Why AI Is Bigger Than the Cloud Revolution
04:29 — What Unified Actually Does
08:11 — Why API Integrations Become a Nightmare
09:56 — Why AI Changed the Timing for Unified
12:48 — Founder Fear & Moving Faster
15:37 — The Biggest Threat to Startups
16:35 — Defining Product-Market Fit
18:50 — Scaling from 70 to 500 Integrations
20:15 — AI Coding Productivity Explosion
21:47 — Why the Integration Problem Is Still Unsolved
24:14 — Small Teams, AI Agents & The Future of Startups
25:31 — Why AI Infrastructure Is Inevitable
27:31 — Will Every SaaS Company Become an AI Company?
29:15 — Why APIs Still Matter in an AI World
33:02 — MCP, Abstraction Layers & AI Protocols
36:20 — Unified as “Plaid for B2B APIs”
38:39 — The Original Startup Idea Before Unified
40:49 — Inbound Growth & LLM SEO
43:25 — Why APIs Are Starting to Close
46:59 — Salesforce, AgentForce & Usage-Based Pricing
49:38 — What the World Looks Like If Unified Wins
52:25 — Do APIs Disappear?
53:48 — The End of Websites & Human Interfaces
56:55 — AI, Jobs & Universal Basic Income
59:16 — Why “Vibe Coding” Isn’t Enough
01:00:08 — What Roy Tells His Kids About The Future
Transcript
Brian Bell (00:01:15):
Hey everyone welcome back to the Ignite Podcast. Today we’re thrilled to have Roy Pereira on the mic he is a ex-founder with multiple exits and now the CEO of Unified.to a company building real-time data infrastructure for AI native applications. Thanks for coming on, Roy.
Roy Pereira (00:01:29):
Hey really happy to be here right.
Brian Bell (00:01:32):
Yeah, so I’d love to start with kind of the way I always lead these things off. What’s your background and origin story?
Roy Pereira (00:01:38):
I got my first computer when I was in high school. My mom spent a ton of money and she saw that I love computers. I was in the computer lab all the time. I wasn’t out playing football or anything I was in the computer lab and I was trying to hack it and I learned everything that I could and she thought it’d be good to buy me a computer so buy one, hack that and then just loved it I love how you can just build anything from scratch continued on university What kind of computer was it? you don’t want to know but went on to computer science in university and actually dropped out because I was so bored and I dropped out in my second year and I actually got a job in computer science and the rest is history as they say it’s
Brian Bell (00:02:19):
Amazing so four exits I mean what have you learned or not four exits but multiple exits but four four time founder is Unified the fifth company yeah Unified is the
Roy Pereira (00:02:28):
Fifth company that I founded personally lucky number five. Lucky number five. It’s interesting when I look-
Brian Bell (00:02:36):
It’s gonna be the biggest one too. It’s gonna be the, you know, $50 billion one. I’m pretty sure.
Roy Pereira (00:02:39):
Well it is already the biggest one in terms of revenue so that’s interesting we’ve just hit a nice milestone just recently and I do believe I agree with you and obviously I would but I do believe that we got the right timing in so much of life and tech entrepreneurship is about timing and we got the right timing on this one for sure we started when chat tpt 3.5 came out and we were like wow that’s gonna change everything in ways that I don’t think we even foresaw but
Brian Bell (00:03:07):
We haven’t even fully appreciated yet and it’s not even done changing things yet, really.
Roy Pereira (00:03:11):
No, there’s a tremendous amount of upheaval, carnage, transformation if we want to be positive about it. But you’re quite correct. This is definitely the start of something massive that I don’t think we’ve seen before. I mean, my first computer was a long time ago. And obviously I went through the dotcom era of 1999, 2000, 2001 crash. And, you know, the web and cloud was the transformational technology then. And it did kill a tremendous amount of incumbents. I don’t think that what we’re going through is even in the same category. It is worse or better, however you want to look at it. More transformational than we saw with the cloud coming in 2000.
Brian Bell (00:03:56):
Yeah, which was, I mean, dramatically transformational for startups, I think, because, you know, you used to have to spend millions of dollars just to get a basic, you know, website going. And now with cloud, you could, you know, just get on the free tier and just start hacking something together and really could argue like the angel investing community and really like SaaS 2.0 really took off with cloud right in the 2010s or late 2000s.
Roy Pereira (00:04:22):
Yeah. If you think about we’ve been living in a SaaS world for 25 years and it’s been the same playing field same rules as an investor so I’m an angel investor I know lots of VCs if you’re in that game you have this this Playbook this calculation and it’s the SAS calculation right and you know that it’s going to increase X percent every year a year ago 18 months ago it started to change and now it’s massively different and I think we’re also going through a transformation so a second level order of transformation in in the investment community that invests in tech because tech is going through this massive
Brian Bell (00:05:00):
transformation yeah well we can get into that I’d love to talk about the impacts of AI and what’s changing in my world and your world. But let’s talk about what you’re building at Unified First and why does it matter right now?
Roy Pereira (00:05:11):
Yeah, so I was a math geek and then I became a data geek, I guess, when I really got into computer science. And for the longest time, I’ve been into APIs because APIs to me is where the data is. It’s where I get access to the data. And I’ve always been building APIs in my companies. I’ve always built lots of different APIs. When my companies were acquired, I would go into the acquiring company and I would always look at how they were using APIs. I would always be disappointed. And so the last company was called Zoom.ai, which was a fantastic name at the time. The other Zoom talked to us a bit, but we were friends. Anyways, so we built a ton of integrations. We built like 70 different integrations in-house. And that was so painful. And that was a really good lesson that no one should ever go through that pain you know the joke is like I had a full set of hair before we started now I don’t because of the pain not quite true but and so
Brian Bell (00:06:08):
that’s how in the car Roy is clearly doesn’t have all his hair for anybody just not watching the podcaster no I’ve definitely given up on that
Roy Pereira (00:06:17):
And so when we were going through this acquisition in 21, all the strategics were interested in what we had built at Zoom, but they were really excited to ask us questions about the integrations that we built, because we built 70 in-house and we maintained them. And so they were always very curious and we knew that there was a lot of value there. There was a lot of things that we knew that we didn’t do right. But then when ChatGPT 3.5 came out and we were thinking about what to do next, we knew that AI was going to need data. We had built this amazing
Roy Pereira (00:07:17):
Yeah, Boomi or whatever. Zapier. How about that? Very popular. So when you need to gain access to your customer’s data, you need to authorize access. So they need to give you access to authorize their Salesforce account or QuickBooks. That becomes a bigger pain. And then when you have a new customer and they tell you, hey, I’m not on QuickBooks. I’m on Xero or Sage Accounting. And then you’re going to have to build an integration. And building Sure that could take it used to take months it’s obviously faster now but the maintenance of it is what actually kills companies and I know this firsthand our last company almost died with the 70 that’s what we didn’t tell the potential acquirers but yeah the weight of fixing bugs on these 70 integrations almost killed us because we couldn’t build any new functionality any features
Brian Bell (00:08:05):
So that was a really good lesson because when we started Unified Does the complexity of doing these APIs, I would imagine it decreases as you support more, but the support tickets increase, right? So what is it about having more integrations that makes it harder?
Roy Pereira (00:08:19):
There are standards out there for a lot of these APIs, both authorization like OAuth 2 or REST API or GraphQL, but everyone does it slightly different. And a lot of them don’t even adhere to the standards. They just are like standard-ish. And so there’s all these little differences. And then also there’s differences between the API documentation that they have and the actual API. That is almost like 75% of the APIs that we support. And we have like almost 500 today. So something that looks really easy from the outside is actually quite detailed because you have to get everything correct. It’s not like me and you conversing in English and maybe I’ll use a different adjective or my grammar will be different than yours and you’ll still understand. APIs don’t work that way and so that was the major issue that when we started unified we actually looked at that how do we maintain these so that we then don’t take the cost of maintaining which is the biggest issue and just transfer it over to our customers no let’s cut that out so that we offer you know 10% of the incumbents in terms of price and then also do a 10x in terms of speed
Brian Bell (00:09:34):
to grow and so forth.
Roy Pereira (00:09:35):
So that’s what we’ve done. That’s what we did at the beginning when we started the company.
Roy Pereira (00:10:05):
Everyone built APIs because they’re publicly available and you just give the API documentation to your partners and then they build stuff. which has been the case for 25 years the more verticals the more companies within a vertical obviously you have to do more work and that that was the normal for SAS and really SAS APIs are mostly around onboarding so you just get a new customer and they have data stuck somewhere else and so what are they going to do upload a CSV file sure or plug into an API. So that was the major use case. And I think when you look back at all these previous technologies that have tried to address that, that’s really their use case. Could be accounting, could be sales, whatever, but it’s the same use case. Two years ago, ChatGPT 3.5 comes out three years ago and it starts to change how data is being used because now data is not just for onboarding like upload your employee list to a payroll system data is being used transactionally for training and to get results so obviously first for insights but now more and more with agents as actual actions to change data. And so that needs very different architecture. And that’s something that we saw like right away three years ago. And we’re like, honestly, we had built a real time system in the past. So it was kind of easier for us to go in this direction instead of the older, more ETL. But that’s what We were skating towards, as they say, right? You skate towards where the puck is going, not where it is. And we saw where the puck was going. It was going to, AI was going to eat up all the data that it can. It ate up all the public data already, but the public data is all public. So there’s no differentiation. Like if you’re building a B2B application, who cares? All right. What you really want to do is feed it your customer’s data. So it really understands who your customer is and how you can help them.
Brian Bell (00:12:02):
So you’ve built companies in multiple spaces and you know what’s a belief that you had back then in the first four companies that you now think is completely wrong?
Roy Pereira (00:12:11):
So after my ad tech company was acquired in 2000 I don’t know 12 realized that what I thought about how to run a company was wrong so that was my biggest insight and I still remember it to this day and basically it turns into I or the founders are the biggest restriction speed bump speed bumps a better word speed bump to the growth or potential and what I mean by that is that We’re all human and we all take our beliefs, our neurosis, our fears, all of that in to the company, our trauma, right? And we’re always trying to pattern match. And so we’re always trying to stay alive in some ways because it is risky, right? You know, like, especially like you’re A 23-year-old kid who’s starting up a startup, you don’t know how you’re going to pay rent. So that fear is the speed bump, right? And I’m not saying, hey, just like drive the bus off the cliff because that obviously is not great either. Biggest insight that I had after my first acquisition was that I was the one that kept it slow. I could have gone so much faster when I look back because those fears, sure, they’re great, but they weren’t there, right? And so it was really just fear and it wasn’t really reality. And so what I’ve done after that was just go as fast as I can. Something breaks, sure, you fix it. And you do it fast, of course, but you try not to focus too much on what could break and kill you because what’s the worst that can happen yeah I like that that’s good advice and so like what’s the best that can happen if you go faster right and I think that’s the that’s the major thing that I’ve learned and I’ve tried to do at all of the companies that’s one thing now that everybody is always mentioned it’s like how the hell do you guys go so fast and I’m like because there’s a massive lion behind me and it’s going to eat us. And we have to go fast. Who’s the lion you worry about the most? There’s a lot of lions. Yeah. Technology moves so fast. Like, look at what’s happening right now. My God. Like, we are seeing a massive upheaval in the entire software industry. Like, I’ve never seen this. Like, we can get into that. But, you know, You know, maybe big guys coming in, competitors, but those I don’t really care. I never cared. If you listen to your customer, you will always outrun them. That’s not the issue. The issue, the big lion is time because tech changes and tech is changing faster. So it’s accelerating the change. And now I think we’ve seen a massive, massive acceleration. So it’s really about time. And can you stay ahead? Can you stay within that MVP window? Not MVP, sorry. Product market fit, PMF window, where you are are matching what the market wants because the market changes all the time. Your customers are churning, but they are talking about you, right? So you always get some churn depending on your vertical, but are your current customers talking about you? Can they live without you? I think the best PMF calculation was the superhuman guy. He wrote a paper and it was really good. And I thought that he really put it succinctly. It’s like, what happens if your customer no longer has the ability to use you? I think that is the measure of product market fit.
Brian Bell (00:15:34):
Yeah, it’s like really painful if your product goes away, basically. How many people are, I think it’s like Dan Ellis or one of these guys. How many people would be in significant pain if your product was taken away tomorrow.
Roy Pereira (00:15:46):
Yeah, exactly. And so, yeah, we’ve tried to create a product that is very sticky. It would be very painful if they lost access to it. But we’ve also tried to be good partners. And I think that’s super important. That’s if I go back to another lesson that I learned when I was at Cisco Systems in the com Silicon Valley, I was always from outside before I joined them. I was always like, why are they so popular? Their products are not super great. and I found out that yeah that’s kind of true but the reason why they were so popular was their team and the way that they treated their customers they treated them as partners and that’s something that’s also stayed with me and I think that’s also part of maybe that product market fit but partner market fit whatever go to
Brian Bell (00:16:33):
market market fit
Roy Pereira (00:16:35):
I think in a lot of ways products are never perfect, but if you’re treating your customers correctly and with respect and all of this and not just like a dollar sign, I think they’re more willing to forgive the little issues that pop up here and there.
Brian Bell (00:16:50):
So integrations are painful, like you said. You went from managing 70 to managing 500. How did you manage that order of magnitude shift in scale?
Roy Pereira (00:16:58):
Yeah, so, you know, we started with a unified API like literally 20 years ago at one of my companies early on in ad tech. And then when we built Zoom.ai, which was an AI chat agent, basically in 2014, we had a unified API. We kept the sort of same architecture. And, you know, lots of companies have done this, not lots, but several. Plaid was a very famous one in the fintech space. They basically created a unified API for all financial statements. So we built that again and we made it better. And we kept on doing that. But like I said, the maintenance was a big, big problem. We didn’t fix that at any of the companies until Unified. And so what we did here was we thought about how do we fix maintenance, but how do we build faster? I want to get to a thousand integrations this year, right? And how do we do that? And like, do I hire 30 people? No, I really don’t ever believe in that, but I certainly don’t believe in that in 2026.
Brian Bell (00:17:56):
and so we came up with and for anybody listening who’s living under a rock and doesn’t know what’s what’s happening and why it’s different now tell us why my
Roy Pereira (00:18:04):
engineering team is literally 10x more productive yeah and it changed since
Brian Bell (00:18:09):
do you think that or do you think that or do they think that or you guys both think that
Roy Pereira (00:18:12):
so I would say that not everyone is 10x um I would we can get into details like I think the the seniors are are x almost like it’s and Then the junior juniors are also like 10x. I think the ones in the middle are struggling because they’ve learned how to build efficient code and that is not necessarily that important anymore. Efficiency is great, but the detailed artistry that developers had is not that important anymore. It’s really about first principle thinking about the requirements and the curiosity and the trust of letting an AI coding agent actually Research and then build. Anyways, I think that is very changing right now. I think we’re going through wave, letting it do everything and then taking the repercussions of that or just letting it do research and then going back to coding and back and forth. So that is a very, very fluid thing that’s happening right now.
Brian Bell (00:19:10):
So the integration space already has some players, some pretty well funded as well. Why do you consider this problem unsolved?
Roy Pereira (00:19:17):
Yeah, you’re right. You can go back like 2020, 2005, Boomi comes out, right? Pass. Brand new industry, if you will, integration provider as a service, fancy term. And that came out because... companies were starting to move to the cloud you needed to move data back and forth and before that we had just plain ETL extract transform load of databases so that’s where the data was but it was all really internal and then you know a couple of other players and so forth and you’re right 2020 saw 2020 and 2021 saw an increase in integration solution providers they were all kind of based off of older technology but they all tried to add in some sort of innovation and they all got funded to your point if you remember 21 everyone’s getting funded there was lots of money going around I’ll tell you a good story though with this so we started in early 23 after 21 23 early 23 was when the interest rates started to go back up
Brian Bell (00:20:17):
It was a tough year to be an investor I mean great year to be an investor
Roy Pereira (00:20:22):
tough year to raise money yeah so we can’t raise any money that was the thing and we were like hey we have a couple of exits we’ve been around kind of thing and like nobody was paying attention and it wasn’t us Hopefully it wasn’t us. And so the long short story of that is we basically made some decisions there that have been instrumental in our success in terms of automating, in terms of relying on AI, even early AI that have helped us propel. because we couldn’t rely on getting funded and then hiring a massive group of engineers. So we didn’t. And we figured out a way how to be much more capital efficient. And to be honest, I actually don’t think I or anyone should be going back to the way that we used to be. I used to go and raise a ton of money and then hire a ton of people and spend it. As we grew, I think those days are kind of over. We’re just so much more productive.
Brian Bell (00:21:13):
Yeah, what does the world look like now then?
Roy Pereira (00:21:16):
The world looks like very small teams that are leveraging multiple agents, whether it’s a coding agent or marketing or whatever. But you can get so much done by understanding the problem of whatever it is, whether it’s coding or marketing, but you don’t need that many people to actually execute on it. So you could get it executed by an agent I think we’re still not quite We can have a conversation around that But I do believe that small teams can run as bigger teams. And we’ve always known this middle management and all this. We saw Facebook fire all their middle managers a couple of years ago, all of that. I think that’s all true. But you have these little pods, very small pods, much bigger than like an extra large pizza box that we used to think about in terms of engineering teams. These pods are smaller and they can just run with the additions of these agents that are doing research, maybe coding, testing, just all of that.
Brian Bell (00:22:14):
So you’ve positioned as AI data infrastructure. What actually changed with AI that makes your product inevitable now?
Roy Pereira (00:22:21):
Yeah, the way that I like to think about it, and this is what the thesis was when we started. Again, we saw ChatTPT 3.5 come out. We played with it. The previous versions weren’t great. We had built our own AI at the previous company. Like I said, Zoom.ai was a chat-based agent. It was a secretary, an AI executive admin. And we had built all of our natural language processing. We built a neural network to understand who you are and how you like to work and all this, like a good secretary. And so we understood the requirements for AI data, but we also understood that data changed. This type of data wasn’t like a static, like language to understand language. Language doesn’t really change much. Business data changes all the time. Like every second, there’s a new piece of data, new signal. And so that was one of the very first things that we We looked at how do we make this real, real time, not like fake real time, but then how do we get all the data, all the data from everything, not just like one thing. Those two components were pretty critical and they were based off of our understanding So we wanted to be and we still that this is still our core thesis. We just want to be the plumbing plumbing isn’t very sexy, but infrastructure plumbing because what is carried in our pipelines is not oil. It’s data. Data though, data is a new fuel for AI. Why is that important? Because AI actually, even though today AI is sort of a bolt on in a lot of ways of applications, AI really is becoming and really will become the core of AI software.
Brian Bell (00:24:01):
Well, it’s kind of like the history of electrification. When electricity came to the factories, they had to redo the factory floors in order to lay them out properly for the reality of electric motors. And I feel like what’s happening now, and I feel it in Ignite in my venture firm I see it in my startups in the portfolio and lots of other places we’re sort of having to retool and rekit companies and organizations around the reality of this like electricity of AI do you believe that every SaaS company becomes an AI company or is that flawed somehow?
Roy Pereira (00:24:36):
I think the definition of what is an AI company is almost irrelevant. It has .ai in the domain name, right? It’s like being in 2001 and getting a .web. a domain. Why, right? And I’m not like, we can say why now. In 2001, there were people doing that, right? Because they were like, ah, we’re a web app. So that was cool. That wasn’t cool for too, too long. I think we all realized that it was just a piece of technology and it was the way of the future in the user space. So I do think that AI, it’s a piece of the core of how we, the stack, if you will, of how we build applications. It does so much that we used to do hand coded, if you will. It adds so much more power.
Brian Bell (00:25:23):
Yeah, now you can just dump me the JSON and feed it to the AI. The AI figure out kind of, you know, what’s in the output and try to write code to kind of conform. That’s why you could do 500 of these instead of 60, whatever. And you struggled, right?
Roy Pereira (00:25:39):
No, that’s actually not what we do. And I don’t think that that necessarily will work out. And I don’t think it is working out today, regardless of whether the These models will get better, for sure. They’re getting better every month, every week. Sometimes they go backwards. But I don’t think that that’s the proper way of doing software. I do think that software still has some deterministic qualities of it. You can’t just dump freeform text, documents, whatever, into AI and tell it to do something. Because if you do do that, how is OpenAI... You said do-do.
Brian Bell (00:26:11):
I just gotta make it fun.
Roy Pereira (00:26:13):
No, no, I know. But if you do that, So I do a lot of mentorship here in Toronto in the tech industry. And I’m always saying like, if you do that, if you just like are a rapper to ChatGPT API, like what’s to stop OpenAPI from just killing you? And we’ve seen this, right? So we saw the first versions of agents.
Brian Bell (00:26:33):
But I have really good prompts. My prompts are so good. Yeah. But I think a lot of, I mean, maybe in the 90s and 90s, people are like, isn’t that just like an interface on top of a database?
Brian Bell (00:26:51):
Isn’t it kind of the same argument, kind of, in a way?
Roy Pereira (00:26:54):
There’s actually a lot of similarities, and we can talk about that. And I’ve been through those, and I’ve been thinking a lot about that. Am I just old and grumpy? Because I don’t like this prompt stuff.
Brian Bell (00:27:04):
Yeah, no, I can feel myself getting older and grumpy. I think I think we’re roughly I’m in my mid 40s. So like I’m getting older and grumpier now. Yeah, for starting to get into that phase of my life.
Roy Pereira (00:27:13):
But just back, I don’t think you can just like throw JSON in and just say, hey, analyze this and you can you can do this one time kind of thing. But I don’t think that’s the future of software in general. I don’t think that that’s I think that’s very sloppy and it’s not going to scare But I do believe that at the core, we will have these models, whether they’re LLMs, whether they’re neural networks or some other.
Brian Bell (00:27:36):
Some sort of cognition thing. I mean, you could almost make an argument that humans are next word predictors with emotion, right?
Roy Pereira (00:27:44):
We’re always looking for pattern matching, as I said earlier. And just back to your abstraction layer. So, yeah, back in the 80s, you know, we were programming, I don’t know what, 90s came in and we had Pascal, this language, and it was an interpretive language. We’re like, no, that’s horrible. I want to code in like the natural language of the computer. And then we had C and C++. C++ was like, oh my God, it’s like massive abstraction.
Brian Bell (00:28:11):
The object-oriented programming boom, right? Yeah.
Roy Pereira (00:28:14):
And then JavaScript came out and we’re like, oh my God, no. Can’t do that. Is it Java? No, JavaScript. We say Java down here. We say Java. No, no, Java. Oh, Java and Java? I don’t know.
Brian Bell (00:28:26):
We say Java anyway.
Roy Pereira (00:28:28):
So JavaScript never compiles down to like a computer language. So old timers like me are like, no, that’s horrible. That’s an abstraction layer, it’s gonna be slow, horrible. But then you, so you think about all these abstraction layers on top of what actually What the computer is doing in zeros and ones. There’s a tremendous amount of layers. And so when you think about what AI is doing now, it’s just another layer. And a good example of that is MCP.
Brian Bell (00:28:54):
MCP, which is- Which is, what does MCP stand for?
Roy Pereira (00:28:58):
It almost doesn’t even matter. It’s model control protocol. It’s a protocol that Anthropik came up with for its AI LLM to talk to data anywhere, right? So like in Salesforce, but having the AI I talk to it directly, and like you said, get JSON, because that actually is what happens, by the way. It takes JSON in a string format and then throws it into the prompt, and then the AI looks at it. Kind of inefficient, if you will. But, to my point, or to your point, it’s just an abstraction layer. And if we have enough power, then we can do it, right? Because in the 90s, we were like, we don’t have enough power to run JavaScript. It’s going to suck up all the CPU power. Things have changed. So maybe, maybe that’s where we’re getting. It’s a very inefficient way of doing it. I’m hoping that we’ll get better protocols that are more efficient.
Brian Bell (00:29:47):
How do you, I mean, that’s part of what you guys are building, right? Do you guys think of yourselves as a tool or infrastructure?
Roy Pereira (00:29:54):
Completely infrastructure. And in fact, we are so agnostic to the actual transport. Like I mentioned, MCP, we have like the largest MCP server on the planet, 22,000 tools. And if you read some of my articles, I am crapping all over MCP. It was like the worst protocol ever designed in the history of computers. I think I said that.
Brian Bell (00:30:15):
The only thing worse is OpenClaw, right? I’m joking.
Roy Pereira (00:30:18):
How many people install that? And then before running, it were like, hmm, should I do this?
Brian Bell (00:30:23):
I was watching a YouTube video on this at lunch as I was eating my noodles. And I had the same experience after wrenching in with it for like 20, 30 hours. I was like, this thing is not working for me. And I’m spending like $100 a day on tokens and it’s not doing anything.
Roy Pereira (00:30:39):
Yeah, I was more concerned about what it was stealing or what it was going to steal. I never actually ran it because my I did security for like 12 years and so I’m always thinking about security. Yeah, so many vectors. Right. I was like, hmm, it’s good. And then luckily I didn’t do it. Anyways, so MCP, just another abstraction layer on top of APIs, which is an abstraction layer. I’m just hoping that there’s more efficient mechanisms to move data around. For us, it doesn’t matter. We support REST. We supported GraphQL, which nobody cared about. And I have opinions on that, which a lot of people don’t want to hear. But we supported GRPC, which is the Google thing that nobody ever uses. If someone comes up with some other protocol to move data, we’ll support it. If a customer wants like, I don’t know, 8-track tape, whatever they used back in the 90s, sure, maybe, we’ll look into it.
Brian Bell (00:31:39):
For the semi-technical audiences, the best way, everybody’s pretty familiar, like using Plaid to connect their bank accounts to things, is are you guys basically the Plaid for all data across all infrastructure and all apps and all AI? Is that how to think about it?
Roy Pereira (00:31:52):
Yeah, that’s a really good way of thinking about it. We are the plaid for B2B applications, basically. If you’re building a B2B application, you will need your customer’s data that’s stored elsewhere, right? Because there’s so many B2B applications. So I’ll give you a really good example. You’re building the next generation AI sales agent. You need access to your customer CRM because that’s where the source of truth is. That’s the core database, right? They’re not going to upload that stuff to you. If you’re building the next generation AI recruiter, oh my God, so many companies are doing that. You need access to the applicant tracking system. That’s where the data is stored. That’s the core. And it’s the same thing. You’re building a FinTech, like so many AI CFO applications are being built today. Where’s the core data? QuickBooks, right? And so, and it’s not just QuickBooks too, because if you, sure, QuickBooks is the monster. It has a majority of market share, but think about Xero or Sage Accounting, or think about going into Europe. Europe has a ton of other accounting. So like as you’re growing, you’re realizing that one integration is not sufficient. It’s holding me back. So that’s where, that’s why we built Unified basically.
Brian Bell (00:33:00):
Yeah and that’s why we backed you for anybody listening if you’re wondering how I know Roy we backed their company and one of the it’s not our only thesis but one thing we do look at when we back companies is like can it help our other companies you know and is it widely applicable across you know enterprise SaaS B2B software business use cases right and this is definitely like one of those things where you’re building something that has you know real infrastructure that everybody will need, you know, and that creates a little bit of lock in, right? Like, you know, I guess that would be a concern if people are listening and you’re building a B2B product, AI product, whatever it is, like, will I actually be able to rip out Unified or do I need to like keep it forever?
Roy Pereira (00:33:41):
So I’ve been an operator, as you said, is my fifth company. And I have some opinions about being a nice company to work for. Because again, like you can have the best tech, but don’t be evil as Google says, right?
Brian Bell (00:33:53):
Yeah, I know. That’s sort of gone sideways.
Roy Pereira (00:33:55):
But so the pricing model is really just and it’s transparent. There’s all of this stuff that is part of who we are and how I want to be treated as a customer myself. And honestly, when we started this, we were thinking about us as customers because we were actually thinking about building a totally different company.
Brian Bell (00:34:14):
So this is great, and the founders love hearing this. What was the original idea? And you’re like, you know what? Actually, we’re going to pivot over here.
Roy Pereira (00:34:20):
It almost doesn’t matter, but it was an AI sales agent. It was going to do it all. There’s lots of those now. But this was like three years ago. And then I said to myself, man, I’m going to need so many integrations because the data is all over. And it’s not just CRMs. I’m going to need like enrichment platforms and messaging and blah, blah, blah, calendar, all of this stuff.
Brian Bell (00:34:41):
yeah it’s a lot of different integrations to get that to work yeah and I got a
Roy Pereira (00:34:44):
headache honestly and I’m like I can’t do this again I just built 70 at the last company I built like 20 in the last company and this was going to be my third sales
Brian Bell (00:34:53):
how can I give myself a migraine and do it for other people right
Roy Pereira (00:34:57):
but it literally was and I actually had a conversation with Salesforce at one point and I actually told them this I said I was thinking that I needed to build yet another Salesforce integration at this sales startup that I was going to build and it was going to be my third and I’m like I just can’t do it you remember like the
Brian Bell (00:35:13):
force platform and things like that
Roy Pereira (00:35:16):
I just couldn’t do it and I just shut down. I’m like, I’m not going and doing that because I’m going to have to build another Salesforce integration and I can’t do that anymore. So I took the summer off and I’m like, I can’t do that. I don’t have any other ideas. And then obviously we had this idea for a while. Ironically, I built a third Salesforce integration.
Brian Bell (00:35:36):
So far, I mean, you just crossed the revenue milestone. Congratulations. I mean, growth has been heavily inbound and founder led so far. how are you looking to kind of scale is it is it just build a better product and people will find you word of mouth or yeah you have to actually go out there and
Roy Pereira (00:35:51):
pound people over the head with the message I don’t think we’re pounding anyone with the message so we we decided to do inbound and lots of different channels around that and not just paid like keep you know Google Ads there’s quite a lot of options there we haven’t done outbound I think the entire industry does outbound dialing for dollars. So we did try and get our awareness up, try to get our word of mouth up. The product speaks for itself, the amount of integrations, how fast we go. All of that sort of differentiates us, not just on a like, you know, you guys are better than last year’s. competitor. You guys are a totally different animal. You’re next generation kind of thing. We’ve played around with that because we do really believe that. To be honest, the first year, we didn’t even understand how different we were until we actually started getting customers. They were like, this is completely different than what I’ve seen in the market. You guys are able to do so much more. That took us a while to figure out the right messaging around that, but it always has been inbound. The interesting thing in the last Four months is the influence of LLMs in terms of inbound strategy. Right. And that’s been a massive change. 40% of our inbound comes in from one of the LLMs.
Brian Bell (00:37:12):
Crazy.
Roy Pereira (00:37:13):
That is more than organic Google SEO.
Brian Bell (00:37:16):
That’s wild. Yeah. We actually have a couple portfolio companies that help other companies, other team connect companies and others figure out that. How do you show up in LLMs?
Roy Pereira (00:37:26):
Yeah, so we’re using a couple of tools and now our team sort of understands a bit of the secret sauce.
Brian Bell (00:37:34):
Which tools are you using that you find success with?
Roy Pereira (00:37:37):
There is a startup here in Toronto that we’re using called peerview.ai. We’ve had great success. I know that there’s others for sure.
Brian Bell (00:37:44):
Yeah, Zome is one of ours and then Sitefire is the other one. So we have two.
Roy Pereira (00:37:49):
Yeah, it’s very reminiscent of SEO optimizers back in the day before Google got wise and
Brian Bell (00:37:56):
changed the algorithm I’m sure LLMs will get wise eventually and they’ll try to
Roy Pereira (00:37:59):
lock that down And you know that OpenAI is basically the Google of 20 years ago, right? It kind of is, yeah They’re about to launch ads
Brian Bell (00:38:07):
What does that make Claude, I think? What’s the analogy? What is the Claude equivalent? They’re the Amazon? AWS of AI, maybe? More infrastructure-oriented, more enterprise-y?
Roy Pereira (00:38:18):
I don’t know Maybe, yeah But I guess my point is they’re about to monetize with ads. This was Google like more than 20 years ago. And then Google figured out that people were gaming it with these content farms and making a ton of money between ads and SEO search. And then they switched it overnight and they created this page rank algorithm. They changed all of this stuff and killed an entire industry.
Brian Bell (00:38:44):
I think that’s a lot of businesses, right? Same thing happened with Facebook and the social media feed. I remember I was working in ad tech. I came out of ad tech, so I know a lot about it. Yeah, they changed the whole ranking algorithm and then they closed the API actually, the API access. So you couldn’t even push ads externally into their feed. You have to use their tools and things like that.
Roy Pereira (00:39:06):
So that’s an interesting topic as well. So one of the trends that we do see happening is the closure of APIs across the board. We saw it before with more old school enterprise companies that had very restricted Restrictive API access and sometimes would charge like tens of thousands of dollars.
Brian Bell (00:39:25):
Some ingress egress kind of fees, right?
Roy Pereira (00:39:27):
Yeah, even to their customers and sometimes to third-party partners.
Brian Bell (00:39:31):
That makes sense. Like, yeah, it’s like if you expose all my data on HubSpot, let’s call it, or Salesforce, and I can just pull it all out structured, I can just go vibe code something over the weekend to replace, replace.
Roy Pereira (00:39:43):
Yeah, so what’s happening now is that we’re seeing not just the enterprise companies do that, but a lot of companies thinking about, why do I have such a public API? Companies are just building stuff, biocoding, whatever, but they’re not using my platform, they’re just sucking up API. So I think that’s an interesting problem going into the wrong direction. The solution that they’re coming up with, the restricted, is I think wrong-minded.
Brian Bell (00:40:08):
Companies are closing their APIs. Yeah.
Roy Pereira (00:40:10):
Yeah, and I think that that’s wrong, a solution for the problem today. I think the right solution is what Salesforce has just done. And to really understand what they’re doing, you have to go back a couple of years. I’ve been looking at this because my customers are coming to me, they’re software companies, sometimes startups, and they’re thinking new things. And so I’ve gotten this front row about how the software industry itself is changing. And I’ve written it in LinkedIn and so forth. But the other guy who I think has gotten it even faster is Mark Benitoff at Salesforce. He had this epiphany. He brought up AgentForce, really pushed it. You can debate whether or not it’s successful.
Brian Bell (00:40:53):
Maybe a little too early, but sure.
Roy Pereira (00:40:55):
Yeah, but I think he understands the ramifications of these changes. He understands pricing model. He was the first one to talk about pricing model cannot be per seat. This is a vestige of when we used to buy CD-ROMs or even disks and plug them in and install it on a server and then we brought it to the cloud.
Brian Bell (00:41:16):
Increasingly software is not a place where you go do work it does work for you and if it’s doing work for you then it doesn’t make sense to price it per seat Exactly
Roy Pereira (00:41:25):
and what he’s launched and what he’s I think his next version of what he started with Agent4 is basically this agentic Salesforce that does not have a web interface. It is a bunch of APIs and you get charged on usage. That I think is where we’re all headed because in the end, like you said, it’s not you You and me that are going to be like, how do I filter this pivot, filter this table? No, we are not ever going to need to do that. The user interface that we’ve been spending $100 a month for on Salesforce is irrelevant. And he’s actually picked it up. I think that’s really kudos because in the end, Salesforce is just a dumb database. If you think about it in a SaaS application, and it will be dead. Obviously, they’re going to fight to stay alive, but it’s relevant.
Brian Bell (00:42:15):
You can do it as some of the other companies are doing, just shut off your API access.
Roy Pereira (00:42:19):
or you can do that. But that is, I think, a short term solution. You’re basically cutting off the supply of your partnership. And I think that is super important. It always has been and will continue to.
Brian Bell (00:42:32):
So if Unified works, what does the world look like in the next five to 10 years?
Roy Pereira (00:42:36):
So much data. I hope and I do believe that even though we talk a lot about the SaaS apocalypse, it’s just really a transformation of our industry. There will be more software applications. We see this today.
Brian Bell (00:42:49):
probably like more like uh like the seat apocalypse you know you heard it here first on on the ignite podcast but because it’s not it’s not like software as a service is going away it’s just seat-based software is going away yeah I I agree
Roy Pereira (00:43:01):
actually I agree software for humans is basically going away right and our interface will be agents uh and swarms talking to apis talking to apis to do stuff And each API will have benefits. And what are those benefits? And if your API is the same as another API, why use yours instead of someone else’s? So I think we’re going down to that level, which is interesting because we’re going down an abstraction layer, right? We talked about abstraction layers.
Brian Bell (00:43:26):
Yeah, kind of like, or headless is another phrase that people throw around, right? Yeah.
Roy Pereira (00:43:31):
Yeah, completely. So I do believe that after the whole vibe coding thing, we’re in it right now. There’s millions of apps, you know, your grandmother’s building one and like trying to show it to you at dinner.
Brian Bell (00:43:42):
I built one the other weekend for portfolio construction, vibe coded it. It took me, you know, a few days, like probably 20 hours, but it’s pretty nice. I’m pretty proud of myself.
Roy Pereira (00:43:51):
Some of the vibe coding apps we’ve seen come in here are better than anything commercial that I’ve seen. So that is what’s going on. I think we’ll move away from that because I don’t think that model is scalable. Like you’re not going to like, you’re not going to launch and launch your product and do GTM and all of that stuff, but most people aren’t. But I think we will get to this more stable state where there’s actually more software, maybe it’s smaller, but there’s more software out there. The core though is still AI, which is the thesis that we have, and AI is just powered by data. So there’s gonna be a ton of data back and forth. And back to my point about the current protocols aren’t too efficient. MCP is not efficient. Everyone is looking at LLMs as AI. That is one piece of AI.
Roy Pereira (00:45:04):
I don’t think they do disappear, actually. I do think that there are more of them. So that’s my point. So the API is sort of like the raw state of data. You will need agents to interface with them. Maybe it’s REST, maybe it’s some MCP or a next version of that. So I am so bullish on MCP and I think it’ll just have an evolution step to make it more efficient. So I think that is going to be the core where companies are focused on, not the user experience because there is no user to your point so I do believe that the user is just an AI right yeah and when I started zoom.ai in 2014 I actually thought that was the time the right time I thought that technology that version of AI technology was good enough and we were wrong we were all wrong chat didn’t work it didn’t understand us it didn’t generate like it was it was like not even close to what we see today but I think that’s getting better and better and better today much much faster plus all the other AI tech that we’re not even including so I think the web is sort of done in a lot of ways as the interface because of us yeah if we’re not there if a tree falls in the forest and no one’s there doesn’t matter so like why build the forest
Brian Bell (00:46:17):
it reminds me of Kevin Kelly’s Kevin Kelly wrote a book what technology wants and he kind of goes through technology the history of technology for thousands hundreds of years and technology never really goes away it gets you know out of favor and so the web will still be around it just won’t be in favor and there’ll be other mediums of collaboration and and and getting work done you’ll still be able to go to a website yeah and sometimes you do sometimes you’re like I just want to go to the website and read but it’s it’s it’s getting rarer and rarer now anytime I want to find information I just ask the AI or and now it’s increasingly do things just add this to the calendar craft a response on this email so that is definitely what most of
Roy Pereira (00:46:57):
the companies are looking at today I can tell you that Both startups and very large companies that we’ve spoken with recently. It’s not about reading data. It’s really the hard part is actually writing. And I could tell you that when we build API integrations, the writing, the updating, that is harder than reading.
Brian Bell (00:47:20):
That’s why you got a bunch of humans dragging stuff into AI right now. Dragging and dropping.
Roy Pereira (00:47:24):
Yeah, I did a little research for a LinkedIn post that I had an idea for, and I basically found a screenshot of a user interface from 3,000 years ago. spreadsheet and it was like I forget what it was it was like the number of bushels of grain or something from Egypt and it looks like an Excel spreadsheet you know columns and rows and all this and I’m like we have not changed in 3,000 years right we’re still using the same user experience elements that someone 3,000 years ago understood and so when we’re not involved an agent is we don’t need those spreadsheets we don’t need that
Brian Bell (00:48:06):
It can also think more multi-dimensionally too, LLMs. You know, we kind of tend to think in two and three dimensions, you know, that’s a spreadsheet, it’s two dimensions and then a pivot table is like a third dimension. Yeah, the LLM can kind of approach the data from lots of different directions.
Roy Pereira (00:48:20):
Yeah, and that’s, I think, a lot of unknown unknowns, right? Right now, yellow lungs are placating us in some ways, right? They’re interfacing with us. What happens when they don’t need to and they have freedom to figure things out themselves and communicate themselves? They’re going to create new ways of thinking that we can’t. Your two-dimensional, three-dimensional analogy is quite correct.
Brian Bell (00:48:45):
Where do you fall in the AI’s taking all our jobs and we’ll have nothing to do argument?
Roy Pereira (00:48:50):
I think that the industrial revolution took 200 years and it caused a lot of upheaval in jobs. There was a lot of people not having jobs until they realized what to do.
Brian Bell (00:48:59):
I’m old enough to- We had factories in the cities and that, you know, we had terrible job conditions and
Roy Pereira (00:49:05):
Yeah. I remember traveling to Poland, Warsaw in 1993 and seeing the post-Soviet wall collapse and seeing what that caused the population. Lots of people understood what to do. They were like, capitalism, awesome, right? Lots of people, even more people did not. And they were lost. It was sad. And that happened kind of quickly. And I think that’s kind of what could happen here. So we don’t have 200 years. This is moving so fast. I’m not sure that I will ever hire a developer in 2027 and beyond. Right. And I think I’m not the only one saying that. That’s just one rule.
Brian Bell (00:49:48):
Yeah. As long as you have already a developer on staff who can collaborate with the AIs. You need like one, you need one technical person to collaborate with the AIs and that’s maybe it or maybe not.
Roy Pereira (00:49:58):
I’m not sure that the role of developer will be needed in the future. Wow and I am a hardcore coder even yeah all of my executive jobs I’ve always like coded to de-stress so like me saying that is pretty big I actually don’t think because that that’s the way we’re going and it’s not just coding there’s a ton of stuff ton of jobs roles tasks so we’ll still have jobs there’ll be less of them yeah So what’s everyone else who either can’t transition quickly enough, and there’s no way that most people are going to be able to, we’re just not that adaptable. So what’s going to happen? And so I think that if we don’t want a pitchfork moment in history, we’re going to have to bring in some sort of basic income, guaranteed income, something like that.
Brian Bell (00:50:48):
We’ll just bi-code our way to success. I mean, we’ll just ask the AI what to do. No big deal.
Roy Pereira (00:50:53):
You can vibe code all you want, but who’s going to be buying it? Because you’re not, humans aren’t going to be the users.
Brian Bell (00:50:58):
Right I think I think it’s a it’s a little bit of like we’re living through the singularity and it’s hard to kind of like picture what’s on the other side of the event horizon of AGI and and super intelligence for that matter right like what is what does it mean to like exist and like do stuff when the AI can spin up thousands if not millions of agents and do its things and then you have the embodiment of AI which is robotics coming right behind it and Tesla is now you know shutting down whole product lines to build like robots. Yeah, I think we’re in for some crazy exciting times. If you’re a technical optimist like me, it’s the most exciting time to live.
Roy Pereira (00:51:33):
I agree. I agree. But what would you tell your kid to go and do?
Brian Bell (00:51:36):
That is a huge challenge. I don’t know. You know what I keep telling them because I do have three kids? It’s like, you know, just act like you’re a billionaire already because we all will be like super rich in the future because of AI and robotics. And like, what do you like to do? You know, what do you like to do? What do you like to work on?
Roy Pereira (00:51:52):
What human thing do you like? Because I think we for the longest time have done work that isn’t really human. Like we shouldn’t be doing it as humans. We’re so much better than like punching in numbers into a spreadsheet.
Brian Bell (00:52:08):
Right. Or so much better than email, you know?
Roy Pereira (00:52:11):
Yeah, like if you think about the potential that we have, the evolution, and here we are, it’s like it reminds me of Lost. If you remember that TV show where like every three minutes they have to go and press a button on a terminal, right? It sort of feels like that. If you like stop for a second and look at the fluorescent lighting in your office, like why am I here instead of not outside doing something nice? Anyway, so I do think that we’re going to get there more than what we’ve done hopefully but like I said I do worry though that we could have a pitchfork moment there’s a lot of upheaval and it’s such a massive scale that I think will cause ruptures in society if not taken at heart by governments right but
Brian Bell (00:52:58):
then we’ll have super intelligent AI to help guide us yeah I’m not sure we’re gonna
Roy Pereira (00:53:02):
have like AGI this whole like you know super intelligent yeah like they’re already more intelligent than you and I in some dimensions for sure yeah For sure, and I think that’s just going to get better for sure but whether or not they’re they are more human more it’s hard to define
Brian Bell (00:53:18):
spider diagrams like you know AI and then it just sort of slowly goes out to all skill levels and some things like you said are it’s way more advanced like this like spider diagrams way out here and then other places it’s not quite reaching human levels yet but like eventually that spider diagram consumes all human capabilities
Roy Pereira (00:53:35):
so I tell my kid plumbing by the way become a plumber that’s the last where AI is going to be like a nurse you know yeah human to human so well that’s the other thing right I love hiring human support technical support people because they are talking to other humans and I think that is important and that I think is part of our secret sauce actually but what happens when my customers are not human they’re agents I don’t need a human that actually becomes a liability so I think we will start to get into that just like we go to an ATM and get cash you know banks haven’t gone away but the way that we operate with the bank has we don’t need a human to give us cash anymore I think that is the direction well the
Brian Bell (00:54:18):
average the average size bank used to have a thousand employees you know the average like 50 billion whatever it is 100 billion AUM bank like in the early 80s had I think something like 800 or a thousand employees don’t quote me on this and
Roy Pereira (00:54:31):
now it’s gone down to eight people yep and I think that’s a great analogy that’s exactly where we’re going to go
Brian Bell (00:54:37):
Yeah, even with ATMs, there’s more branches than ever, right? Because now people want to go into the branches. So there’s more retail bankers than ever. So there you go. Let’s do some wrap up questions. What’s a decision that you made at Unified that felt wrong at the time, but turned out to be right?
Roy Pereira (00:54:52):
Actually, I thought about this question, and it’s actually what I just said, not having AI support. And actually, I know someone here in Toronto very well, and he runs a very successful AI support company. But I really wanted to do this more old school because I do still believe in the interaction that happens between humans and the trust. And I think humans, we... We operate under this trust model. We make decisions irrationally based off of trust and likability and so forth. And so I did not want to cut that expense of human support. And at the time, it felt really wrong. It felt like I was out of date, to be quite honest. And yeah, you know, as a startup, you’re looking at your dollars, right? And like, where do I allocate these funds to? And I could allocate it to sales instead, let’s say, to grow. But I chose to really sort of push hard on human support and and honestly every single employee at the company does support not just our support team and it’s something that I really push hard because I want everyone to understand what the customer problem is but I also want the customer to know everyone’s name and their face and how they act and all this because in the end humans like interacting with humans it reminds me of Terminator right when they’re trying to figure out who’s human coming in and
Brian Bell (00:56:13):
what so you’ve been an investor what’s an advantage that you have that you think is invisible to most investors
Roy Pereira (00:56:17):
I think it’s always been easy to dupe some investors because they’re not technical enough they don’t know exactly where the market’s going and what the meta of the market is a big strategic coming in or people moving to some other way so I think, for instance, I don’t invest if I don’t fully understand the market. And I don’t care necessarily about the technical stuff, but I have to understand the market and where it’s going. And I think a lot of investors are looking at numbers, whether the revenue is going up or whatever. And I don’t think that is sufficient. And especially tech has always been fast moving. Like I said, again, it’s so fast right now. You look at, you know, 2023 vintage companies or 2024 vintage companies that were doing vector databases. You know, those are mostly dead. Right. And so it’s very, very fast, even if you’re investing in AI or like agents. Agents came in. They’re still here. But I would say that these are next gen. These are B2 or whatever. The V1 things are they’re all dead. Like all those investments, just write offs, in my opinion.
Brian Bell (00:57:22):
Yeah. So speaking of that, what’s a question you wish more investors asked you?
Roy Pereira (00:57:26):
I think some of them do ask the right question, which is what the question I would ask, which is why do your customers choose you? But I think really sort of getting underneath, not just like great product, whatever, right? Everyone can say that, but like really trying to figure out why we’re different. And I think everyone does say, oh, we’re very scalable. We’re very next gen, all of these marketing fluff terms that everyone uses. but like how do you figure out what really is true and I think customers understand because customers have the pain and they have the reason for looking for a solution for their pain and so they really understand the differences and so asking those right questions about what does the customer say why do they say that tell me something interesting that the customer said that they wouldn’t say about anyone else kind of thing and I think you’ll find those little diamonds that sort of signify that yes there’s something different here now that is tech product I think when you’re doing early
Brian Bell (00:58:22):
stage investing it’s really the people yeah what’s something about your product that customers value that you didn’t expect
Roy Pereira (00:58:29):
I’ll tell you why so the first problem with data integrations is getting authorization and it’s it’s not an easy problem to solve because it’s not you authorizing access to your Salesforce account that kind of is easy because you know yourself and you’re like sure it’s really getting your customer someone who doesn’t necessarily know you doesn’t maybe trust you or whatever Maybe he’s not technical, probably isn’t So that was actually a really hard problem And something that we solved like very, very early And then we sort of made it work like seamlessly And we never thought about it We just stopped talking about it We didn’t even have it on our website And we realized that that was like a massive differentiator For like the previous generation of these tools, solutions And our customers came in and they were asking us all sorts of questions We were like, yeah, no, you don’t have to You don’t have to do that Don’t worry about that And we just didn’t realize how much we’ve solved that one singular problem until literally two years after we started the company. And then we’re like, hey, maybe there’s a lot of value here. Maybe we should promote this because people are like... really it’s really hard to get this to work properly and we have it and nobody ever talks about it nobody ever says oh there’s a bug with it like it just works and it works across everything so it’s it’s something that we just sort of forgot about it and then we have to back in terms of promoting it and talking about it because for us it was like yeah whatever it’s done it works what’s a belief about building
Brian Bell (00:59:59):
startups that you’ve changed your mind on recently now that you’re on your your fifth startup building in the age of AI um
Roy Pereira (01:00:05):
I’ll tell you what hasn’t changed. People, people, people. Just like building a group is always important, regardless of if you’re a startup or not. Just like getting like-minded people, but not too like-minded. You know, people that are going the same path. I think it’s super important. But something that has changed, obviously the way that We’re so profitable, productive, efficient is something that has surprised me. I knew that, you know, all of my companies have been very efficient, capital efficient, but now with the new technology and the new processes that that enables, it is insane how efficient we’ve become. And like we talked about, you’re going to have like a team of 10 people at a company who built amazing products who can get like $100 million of revenue. And I think that is going to be a common, common thing.
Brian Bell (01:00:59):
Yeah. Well, you could argue OpenCloud was the first one person unicorn, right? Just Peter Steinberger, open source project, acquired for what, 1.2 billion OpenAI, something like that?
Roy Pereira (01:01:09):
Yeah, a bit of an outlier.
Brian Bell (01:01:11):
Bit of an outlier yeah but kind of kind of the first one person unicorn you know
Roy Pereira (01:01:15):
Yeah, I think I read an article about this other guy out of L.A. or something who basically did the same thing. One person.
Brian Bell (01:01:22):
Oh, yeah, I saw that one as well. I got it to like 400 million in revenue, something like that.
Roy Pereira (01:01:26):
He’s going to hit a billion this year.
Brian Bell (01:01:28):
So that’s wild. Last question, where can folks find you online and find out more about Unified?
Roy Pereira (01:01:32):
Yeah, so Unified.to. What’s the .to? What is that? There’s a good story right there. When I moved back to Canada from Silicon Valley, I moved to Toronto. Toronto was if you don’t know third largest city in North America if you don’t include Mexico Mexico City is bigger than right and also third largest tech hub in North America right so lots of talent here and funny story is there is a Pacific Island nation called Tonga that its TLD its internet domain was TO someone here in Toronto 2020 30 years ago got this bright idea about buying it and licensing it basically and paying a royalty to Tonga for it. And so now we have our own city domain and I’ve heard it’s one of the largest revenue sources for Tonga is the registration fees for .to. So why .to? Not because I live here, because io.com.ai, all of those things were taken. And I do not like long domains. I had zoom.ai before. I had shiny.com before. Anyways, all one word. I wanted unified. Love it.
Brian Bell (01:02:45):
yeah we have we have teamignite.ventures teamignite.vc and we have tiv.com I’m thinking about switching everything to tiv.com because it’s just so short and punchy typing out teamignite.ventures is so like it’s so many letters so many
Roy Pereira (01:02:58):
characters yeah yeah it just irks me yeah so I’m always like one word and so you have to figure out how to get that and yeah whatever doesn’t matter anymore
Brian Bell (01:03:07):
Honestly so nobody cares. Well, I really enjoyed the conversation. Roy, thanks for coming on.
Roy Pereira (01:03:12):
Thanks Brian.
