Guy Nirpaz: What Does AI Do to Business Models?

• Slides
• Transcript
• About the Speaker
• Join the BoS community

Having spent 15 years building his first company Totango through multiple rounds of venture and ultimately a successful exit to private equity, Guy has just started a new venture armed with a new set of experiences and a lot of thinking about the future.

He shares some of the biggest lessons in his journey at Totango and his post-exit realization that it had not been a healthy journey for him. 40 lbs lighter and with a commitment to himself that he wouldn’t fall into the same trap again, he shares why AI can enable a healthier approach to entrepreneurship. He’s started his new venture by asking himself what does AI mean for entrepreneurs, business models and company building?

You’ll learn:

• AI is disrupting traditional business models by enabling faster product development with smaller teams.
• The future of software is shifting from click-based interfaces to conversational AI assistants that understand and execute tasks.
• Successful AI products should focus on amplifying human capabilities rather than replacing workers entirely.
• Technical competitive advantages are evolving, with process and user experience becoming more critical than traditional development skills.
• AI is transforming business operations by potentially reducing costs in areas like customer support, R&D, and team scaling.

Slides

Find out more about BoS

Get details about our next conference, subscribe to our newsletter, and watch more of the great BoS Talks you hear so much about.

Transcript

Guy Nirpaz 

Hi, everyone. My name is Guy Nirpaz. The accent is from Israel, but for the last 12 years I’ve been living in Silicon Valley.

I want to share the way I think about AI as I’m building the new business coming out of previous business. But before I start, just to make sure that I know who’s the audience. So let’s do a quick show of hands survey on AI usage. I’m going to ask how many of you are using AI product daily one, AI product daily two, AI product dailies three or more. Okay, and then we’ll try to figure out which ones are the most popular. So who’s using more than three AI products everyday?

Not embedded AI, like one of the new AI stuff that came out. So, copilot, chatGPT, Claude, or any other product. So three or more? So we’ve got a few. Two? Most of the people. At least one? No, not at least one. One? Sorry. And none? No, not using at all? All right, very few.

So most people here have been on the bandwagon of AI for quite some time. I’m sure you have your own thoughts. I’d like to make this talk interactive, because I was telling Mark today that two stories. One is, I grow up I have more questions than answers, so I’m not guaranteeing any answers. I’ll just tell you how I think about things as an observer. And the second anecdote is me talking about business models.

So you remember Zack that was here first time? So when I was in my previous company, I was asked by a VC, what’s your business model? And they couldn’t understand what I’m answering. So I picked up the phone and called Zack, who was an early customer of my previous company, because he’s an accomplished dude – Zack, can you explain to me what the business model is, please? So for a person that 10 years ago, didn’t know what the business model is, it’s pretty high bar to try to talk about business model and AI, so bear with me.

I see myself as an active participant in the software market. So everything I’m telling you is you can see that I’m acting on, but again, I’m not a consultant, I’m not a scientist. I’m not looking at that from, just it really guides me as what is the next thing that I should be doing. I think that’s the flow of this presentation. I’m just going to share with you my thinking about where all this is going and what needs to be done. What am I doing? And would love to get your feedback. So I know there are mics that can go around during the talk, so if you have questions, we don’t need to wait till the end. And if it’s a good one, I’ll actually answer it if I have the answer. So I’ll do my best.

So a bit about my background. I started as an entrepreneur when it was for SaaS. So I built Totango, the customer success company. I’ve seen, I mentioned Zack, but other folks in the audience who have been through the life of Totango customers. So thank you for that.

Software is Eating the World

Back then it was Andreessen, Marc Andreessen, Software is eating the World. That was the world of SaaS. And I just read yesterday someone writing AI is eating software now. So we’re kind of switching from one world to the other.

Just a bit of a background, I started Totango because of SaaS. It was SaaS for SaaS, it’s a customer success technology. The idea was that in a world that the big technology transition of SaaS is basically doing two things for software.

One is putting them on the cloud, and the second one is introducing recurring revenue business model subscription mainly. And there’s going to be a problem to retain customers and grow the existing customer base because the entire business model shifts from perpetual one time payment to recurring payment. And that is what we were trying to help with Totango. Customer success technology, the mission.

The idea is customers are going to renew if the company is going to deliver value. So let’s build technology that enables companies to be customer centered, so they focus on their customers by helping them scale the founder mentality of early stage, where founder cares very much about every customer, to the point in which companies scale. And then you’ve got to have a system that makes sure that your company cares about your customers through the entire customer journey, from everything wholesale, from onboarding and driving, adoption, managing escalation, automatic renewals, driving upsell and so forth. That was the the idea behind Totango.

The technological idea was, if we this how we start, if you capture usage, you have information about your customer, whether they use their product or not, and how they used it, and that’s a good signal of value that they’re getting. You can build a health score regular green as a result of that. I don’t want to bore you with a lot of the details. One of the lessons that I learned that I’m doing in the new company, was that it’s very hard to introduce a market for six. It takes a very long time. You got to write a book, if you don’t want to keep pitching people the same pitch time and time again. So that’s the book that I wrote.

If you’re a VC company, you lose a lot of your equity because the business doesn’t grow that fast, and you need more money. And VC gives you money but take a lot of your company with the money that they give you.

So, long story short, targeting an evolving market has its advantages. You go to a lot of conferences, you speak to a lot of people. You are very innovative. Money comes much later. So take it with a grain of salt the way you want to build your own business.

But the story was good. We were early in customer success. Early customers were a lot of, you know. So the meta here was that all of our customers were SaaS companies. I got first hand view into the SaaS market and the business model. If you think about this, in a recurring revenue in the SaaS business model, you got three ways in which you can make money.

New logos – that’s the initial subscription. And then preventing them from leaving – that’s churn. And then the third one is the upsell, and we cover two out of three.

In mature companies, actually, most of the revenue comes from these metrics – preventing churn, maximizing lifetime value, net dollar retention, net revenue retention. Don’t want to bore you with all these details, but I was able to see a lot of companies going through this transformation to this new business model and executing super well. Just to say, one company that really transformed through that was Zoom. Zoom was an early customer of Totango. So they paid us initially $50,000 and during the COVID years, they paid multiple millions, just because our pricing was tied to usage, and was very good for them, but also very good for us. So we had a great, great run.

We learned a lot. We started with early SaaS companies, then scale up companies, really across the industry, like the 2010 to 2021. And we ended up selling the business to private equity in 2021. I stayed for another year to run the business as under the private equity which is a completely different experience than. So founders either didn’t take money or have majority in the business. It’s a very different experience. But anyway, experience is good for life.

Then I had to decide what I’m going to do. I think Jim talked about this the minute that you’re not in the chair anymore. So you rest a little bit, and you lose some weight because you’re not responsible for the quarterly results anymore.

By the way, who here works on a quarterly booking schedule that has to hit quarterly goals, anyone? So I can tell you that what I felt is I noticed that come, let’s say, beginning of December or beginning of March or beginning of June, something changes in your body. I don’t know this is, do you feel the same thing? I don’t know. For me, it was and when I stopped running the business, it was very weird, very different. I still have this stress level around the quarter results, but I had no one to talk to.

So anyway, I had to decide what I’m going to do next. And it was just around the time where chatGPT 3.5 came out. And I didn’t mention, but my background is in technology and product. That’s kind of how I started. This is why the first company I built with Totango. The idea was you don’t need a business model. You build a great product and then stuff will happen. I don’t know, I had to learn that. Yeah, you gotta build distribution and marketing and all this the other stuff, so people actually know about what you’re doing. But long story short, what am I gonna do next?

I’m not a VC, meaning I’m a builder. I’m not a VC, I’m not a consultant. I was very happy to consult to startups, but every time there was an issue, I would jump right in. Let’s go do it. So I was kind of basically telling them what to do, which is not what they’re. They didn’t ask me to tell them what to do. Just wants to advise. So I’m not a good consultant. I mean, I’m happy to help anyone here, but I’m just a builder. I like putting my hands on the keyboard and start doing stuff.

It was chatGPT days, so the world was changing completely. For me, it was like a Netscape moment. You look at that and you can speed of ideas that goes through your head about all the things that are going to be different with this new technology, which we’ll talk more in a bit. Let’s go build something.

I was just at the end of renovating my house, and I had this idea for a new company. So once you renovate the house, you replace a lot of your old stuff, and I saw two different experiences for products that come to my home. I got the nest and ring and Sonos, which is just a joy to install, and start using them. No, seriously, you do a QR code, you go through it, you get this ding. You feel good that you spend the money on these products. And you get other products, like your fridge and your stove and your couch, and just dump them and put all these papers next to them, and good luck, and let us know if you need anything.

AI over PDF

So my idea was, let’s build this. Let’s use AI, because the key idea that everyone talked about is AI over PDF. Do you remember that or not? There was like a can we do AI over PDF? That was a big idea. So let’s take all these PDF, let’s take this product at a QR code, and create this great experience for all these new products. All right, I’ll make it more interesting. That was the idea.

So we sell to manufacturers or product companies, they put QR codes. Other folks, the consumers, get this QR code instead of the paper. Get this mobile experience. All good.

But now I knew, because the second company that I’m building, I knew, Okay, I gotta, I can’t just start building this. I gotta get some market feedback. Is it useful? Does anyone even wanted to begin with? So I went to talk to the guy that sold me the appliances, and he said that it’s a great idea. And but when asked, but who’s deciding on payment here? And then my boss would he paid? No, he’s not gonna pay for that. And so that was one thing.

And then my son, he’s now 14. Back then, he was 13. By the way, we start Totango, he was zero. So I always called him Totango kid, because I could see the company very tied to his age. And given that he, it’s kids, it’s all Xbox, and all the screens, they’re native. I don’t know about your kids, but my son works out of like, three or four screens concurrently, right? So he kind of plays the Xbox, and he has YouTube going and something is, I don’t know how they do it, but they’re all going to be pilots, I guess. Just under, their brain works differently.

But long story short, he said that he think all the folks that worked in my home and installed this stuff said, yeah, it’s a great idea. And my son said, I don’t think people are going to use that. It’s an extra step, the scanning of the QR.

Long story short, I think it’s a good idea, but we got to validate that. So I’m asking chatGPT, okay, what do I need to do? The answer is, market research. Okay, can you explain a little bit what market research is? You find people, you ask them, you validate. Can you come up with set of questions? And it did. Okay, one is for the companies, one is for the consumers. Can you ask me those questions? I just want to see how it feels, so it asked me those questions. It feels actually pretty, pretty nice, like a conversation. And after I’m responding, I’m asking chatGPT, if you analyze my responses, what do you think would I be a good candidate for this product? Would I buy it? And did an okay job.

Then I realized, Okay, I’m going to build this product. The product that does this process that I just described, which is basically able to go and ask people questions and summarize the results and the reason. Who doesn’t want to be the Levi’s when they’re searching for gold, right? Everyone is building new set of products. That was my original thinking. AI forces everyone to start to rebuild, whether it’s incumbents in existing companies. I’m sure every one of you that works at existing companies the last two years, the main conversation is, what do we do with AI? How do we put AI into our product? Because you have to, because otherwise you can’t compete.

And new companies, of course, want to take advantage of this break in the shift in the market. To take advantage of the opportunity and introduce new set of companies that didn’t exist before. So this is how the new business that I’m working on right now.

I raised a little bit of money. I called it perspective, right? You need a name and money, and that’s all you need to do. And, no, I’m joking. Actually, first built most of the product with chatGPT, and then with Claude and copilot. It’s so much easier these days to to build and that’s kind of you see on the screen. This is what it does. Basically, you define the question you want to get answer to. It creates a link. The link is a link to a feedback conversation. This feedback, once people complete the feedback conversation, it analyzes, summarizes and give you the answer.

And it was this dream product that I always wanted when I was the CEO of Totango. Because when you know feedback is critical for business. Every quarter you do this QBR, you are trying to kind of figure out why we’re losing deals, and what can we do about this. So you do this win loss analysis. And I realized that what I wanted, always wanted, is this question of, why are we losing deals? And what can we do about this? And I just want the answer – you’re losing deals because you’re missing two key features, and your pricing is 20% above what it needs to be, versus all the process that is involved, either going to consulting businesses, paying them $200,000 they interview 40 people come back three months later and give you this long presentation.

Or if you do this with a survey, the anecdote was that my VP of sales would always give me this report, which is top three reasons why we win user experience, top three reasons why we lose user experience. What do I do about that? I would say, Neil, come on, what? Why don’t we? We don’t know what to do. So that, this is this vision for for this company, it kind of covers some of the ideas, but I do want to share with you what are the thoughts that goes through my mind when I’m building this, or deciding to build specifically this, and how to compete?

So the first thing is, let’s learn from history. So, Netscape moment. Some people say that AI, the transition to AI is bigger, is like the internet, and some people are saying it’s bigger than the internet, but it doesn’t matter. It’s big and it’s very important. And we got to do something about this.

So the first thing that was so if we agree that this is the internet type of moment, let’s look at internet winners. So who won when we switched to the internet? There were some early winners, but ultimately you had one company. Because everyone had to build a website, there was one company that made indexing website super efficient and made finding websites super, super easy, and that became a big business – Google.

And the other company basically built a store on the website, and they competed against the other stores that had to invest in brick and mortar and employees and a lot of kind of cost, and they were able to build this website that you can actually sell through or sell with, and became a huge business. Both of these companies these days are over a trillion dollar business.

So how do you build? I mean, I’m not that motivated to build another trillion dollar business. I’m actually more motivated to zero to one, but assuming that the game plan is to build a trillion dollar business, what can we learn from these lessons?

So if we think about the Google lesson, there’s going to be a new reality. A lot of new websites. People are going to have a hard time finding them. Let’s build a service that makes it very easy for people to find content in websites, and put ad revenue, ad business model on top of that. And this how we’re going to win. I don’t think it was a business model innovation, but it was super interesting product. It was a business model on top of that.

I think what we’ve seen with Amazon as an example, is that they build a specialized website that basically disrupted the non web company. So what’s the equivalent? What’s the equivalent in this world? You’ve probably seen a slide like that in the past. Because people talk about types of innovation, if you search it up, like types of innovation, there are so many types, like business model innovation, product innovation, distribution innovation, and so forth and so forth. I don’t want to get into that. But what I think is super clear is, AI is a disruptive innovation, right? Would you agree?

I mean, we just, we had, like, what, 50% using two AI products per day. I think a year ago, the answer would have been, you know, maybe, maybe 10% of the audience.

I’ve been running those surveys for a while because I’m trying to understand where this is going, who’s winning in AI? And it’s like every six months, the metrics are completely different. So I think everyone agrees that this is a new thing. And the reason why it’s a new thing, completely new thing, is because when software was eating the world, which was big, right? All this SaaS revolution with software was eating the world, it did super well. Digital execution, right? Sending emails, recording deals, you know, whatever we build in workflow, workflow products for SaaS, that was a pure execution. So there was a clear separation. The human would understand, think and create. And the product, the software would execute. Fulfill whatever needs to be fulfilled, the record book, the ticket, and so on and so on. But I think what we’ve seen for the first time with chatGPT, I think that was the aha moment, is that the interface of this new kind of software can understand humans in a native human conversation way. Can think and can create, right? A lot of people try to kind of create those new blog posts or SEO and all these kind of plays. But can it do it as good as a human? It depends which human, of course, right?

But it’s clear that this is the direction it’s going. It used to be six months ago, it was super slow. It was super hard to incorporate that into apps. Now all the models are super fast, right? The response time used to be like 40 seconds. Now it’s five to 10 seconds and so forth, and it’s going to be even faster.

So, long story short, in a world in which the machine or the software can do part of the things that only people could do before. What does that mean? There’s a lot of conversation around ethics and humanity in general, and AI, and I don’t want to talk about that. I want to talk about what does it mean for business, specifically for business of software?

Because that’s what I want to build software business that is going to be successful. So I want to understand, how does it all work together? I also think that there are two important metaphors that are worth discussing. So how many of you are driving self driving car, complete autonomous cars? Some are, I’m sure, Tesla, right? Yeah. I know, I’ve got those Tesla friends that are driving from Palo Alto to San Francisco and reading your newspaper while the car is driving.

Fine. I actually saw someone doing this on purpose, like a big paper, just to show that. And I had this story that was driving in San Francisco, and San Francisco has the Google the Waymo cars, and there was this car that was cutting me off, and I was kind of honking at it, and everyone on this sidewalk was, you know, mocking me saying, Well, why are you honking? It’s a self driving car. Again listen.

Mark Littlewood 

I bet there are more Teslas in California than there are actually physical newspapers.

Guy Nirpaz 

Yeah, you’re right. Okay, so really reading X, whatever. But the point I’m trying to make is, we had one person here talking about self driving car. And I think where we are with AI and humanity right now, it seems to me that with regards to adopting self driving cars, our tolerance level is super low, right?

Even if cars drive better than humans these days and they make less accidents, we don’t feel comfortable adopting self driving cars until they are flawless, right? So the order the self driving AI has a much higher bar than an assistant that helps you do something. And then the human is still accountable for the outcome. So that’s one observation, because I live in California that I had around what type of AI are we going to use. Now we’re going to switch gears a little bit into into P&L, right? Who doesn’t like to read the P&L.

But the question is, the key question is, okay, I want to start a company. It is a new startup. I already decided that it’s going to be to an existing market, because I’ve had my share of inventing markets. I want to go into an existing market where demand already exists, and what I need is to build something much significantly better than what’s currently available. And in a world where all the incumbents are also adding AI into their products, and they have advantage of distribution, and they have advantage of brand recognition, and they have advantage of customer relationship. How do you compete?

On the other hand, is the I know that not everyone here is a new company executive or founder. Some of you have existing companies. How do you compete against new players that are coming with AR? So I put here something that, as I mentioned, initially, I didn’t know how to read that, but then over time, I had to learn, and that’s a classic P&L for assessed business. So the numbers here are, like $100 million numbers. Let me just kind of review it.

The top section is revenue is incoming. This is how you make money, and it’s broken into subscription and professional services. Then you have cost of revenue, which is how much money you’re spending to service this revenue that’s been created, which is in a SaaS business. It’s usually hosting an infrastructure, customer support/customer success, depends on how you want to allocate that. Professional Services – If you have professional services revenue, and that’s the total cost of revenue. What you’re left with is called gross profit. So it’s your revenue minus the cost of revenue, and then you’ve got operating expenses which are broken commonly. This is a simplified version broken commonly around sales and marketing costs, research and development, R&D and G&A (general administration), that’s your total operating expenses. And if you deduct that you got your operating income, and if you deduct interest expenses for loans and taxes, this is your net income. That’s how a SaaS business roughly makes money.

So what does AI impact on, let’s start with the operation itself. What does AI impact, right? So, it was discussed, I think yesterday, about support. So we’ll talk more about revenue, assuming that we can create more, better products and reach additional markets and so forth. That’s going to impact revenue. But from an operations perspective, it’s clear that customer support, with 1000 new startups that are targeting kind of AI for customer support. I think Zach talked about this, right? So the new Zendesk building is, you know, call center technology for for that. So that’s one.

I don’t know if the good news is why the hyperscalers, Amazon, Google, cloud, Azure, are doing so well. Because you actually need more hosting and infrastructure if you are building your own AI, if you’re just consuming. But there’s more of those services. And if we look at the rest of the last year in Silicon Valley, 50% of startups were customer service optimization startups, and another 25% were sales optimization startups.

So what are we learning from that? Is that a lot of folks are betting on this idea that we can take cost out of the business by replacing people – tier one support, BDRs, SDRs. It starts with a lower quality or lower importance. I’m sorry that I’m saying it like that, but let’s say that the bet is that because support is organized around tiers, so it’s anyway an escalation model, and the tier three is where the engineer seat, and tier two is like a mediation. Tier one is less specialization and more bodies, if you will. And again, I really apologize about speaking like that, because my belief is that’s not necessarily the right approach for AI. But a lot are betting on taking cost out of the business by hyper automating things that only people could do, and now machines can do as well.

The second big transition that we’re seeing on this PNL, someone wants to guess, I didn’t say sales, but unless, let’s believe in that. But you know, we’re seeing this right? Who’s using copilot or? Developers. So the cost of cost of research and development, which is in this case it’s 20%. It usually goes up to 40% depending on the company, from the revenue percentage, is also up for disruption. It’s already have been up for disruption with global workforce.

People started moving, during COVID specifically, started moving developers from expensive centers like San Francisco into less expensive cost of living areas and then into distributed areas. In Totango, we had developers in Israel, which was where we started. And then we started hiring developers in Ukraine, just because there was a talent war, it was super hard to hire, so let’s hire. Let’s expand hiring. But now with copilot, and what’s the biggest one right now? It’s Cursor. Hey, exactly. I’m using that too. With Cursor, it seems that you might need less developers to build a business. So for me, as an example, as a new company, I was trying to figure out, how long can I stay on my own until I need to hire assistance. I could build most of the product on my own as those AI products becoming better and better. I mean, cursor recently kind of released like this new, optimized workflow. But there are others. The competition keeps going, but it makes it super easy to execute. And if you think about this, when you already have a developer that can get into a code base without training, and you task them with some sort of a feature or a change that you want them to make, and they can do it, test it, and you just need to review and release it. You’re basically becoming more of a product manager and less of a developer. I felt that I’m actually not becoming a developer again. I’m doing the stuff that I’ve been doing in Totango as the head of product, if you will. But without physical product team.

You write specifications, you review designs, you review code that has been created the new product, and then you decide if it’s ready to go or not. So that’s a big one that is happening right now. I think these are the first few.

Now, when we talk about I want to kind of switch gears into this one, so this is like a simplified version of the of the P&L. I just want to show you that I know how to read P&L, but this is much easier to read.

So in this case, when I want to cover kind of each topic here, and I specifically put this on this supply and demand equilibrium, because it’s important also, not only to think about what is it that we can do as a business. It also what does it mean from a from a market perspective, because that impacts a lot of things. So let’s start.

So we talked about product efficiency, right? We can build product much faster with much smaller teams. I think anyone that is not doing that is risking being a much heavier business than anyone in the competition. You got to do it. It’s here, it’s available. It has to be done. And the tools are becoming better and better.

But let’s talk about the product output, not the product operations. Let’s talk about what products should we build in a world of AI? Because it’s fine, we can. But if I’m a startup and I want to compete, if I’m going to build a new version of server monkey, why would someone compete with me? Kind of buy from me? Because there’s already server monkey. They’re going to add AI into their product. So why would they buy from Guy with his new perspective company that has no customers? Why would they? So, I think the way I was thinking about this is twofold, and I kind of missed that point on the previous slide.

One is, do we only compete with product companies, or are we competing also with services companies? We call this software as a service, but we really meant so software delivered on demand. We didn’t really meant service for service for best practices, for human capital, for BPO and so forth. We used service offerings back when. When I needed to do win loss analysis, or when the PE bought the Totango, they had this report from from Bain. And we had to pay for this report out of the proceeds of the deal. And the report, and we had to pay $2 million for a report. And I told the PE look, if we do, if we do report by the pages, I’ll do it. I’ll take $10,000 for a page. But the point being is that the consulting firms, various agencies and so forth. It’s a very frothy business model. And the way in which they justified the price – reputation plus head count. And and they were the only ones that could do that. And I’ll give you again this, this use case of win loss analysis, or market analysis, or anything like that, where you need to go interview people and summarize their findings. You would pay $2 million to $100,000 depends on size and type of company and so forth.

Another anecdote, I mentioned this in the round table yesterday, my almost second AI company, I told you about my first one, which was the Fridge app thing. The second one was, I was thinking with a friend that’s build something for lawyers, because there’s a lot of text there. And because I live in Silicon Valley, I could see that in any coffee shop, every table talks about legal AI. So I realized, okay, that’s going to be a very busy market. But if you are into legal AI, I’ve got three domains that I’m happy to sell you legal GPT, lawyer, AI, I’ve got them all, so talk to me later. I’ll give you the domains.

And now potentially AI could do that right? Meaning, AI can interview, AI can analyze. AI can schedule the meeting, AI can analyze all these results. So when we were thinking about this, how do we compete? The question is, do I look for software business that is going to add AI into their business and try to compete with them? Or do I go after services business that it’s already an existing market, but you compete by delivering using AI to take cost out of this operating model.

So product question – first, are we competing with services business, or are we competing with existing AI SaaS business? So we understand the services opportunity, the challenges distribution. On the south, I decided to go with the software story eventually. And the thinking was that in SaaS, we’re basically delivering, the SaaS is always as part of an existing workflow. If we think about, you know, Salesforce, Opportunity Management, where does it sit? I go talk to the customer. I understand if there’s a deal to be done. If there is one, I go to Salesforce and update the opportunity probability, so my manager could have that in their report, and so forth and so forth. The SaaS is always a workflow. So, I mean, it took me, like, four years to learn that SaaS is workflow into tanker, right? I thought SaaS is data, and then I learned that it’s a workflow. So, you know, just saved you four years of your life.

Workflow Revolution

If SaaS is the workflow, then the question is, what is the you know, how can you compete? And I think the opportunity is to basically, the potential new product is 10x improve the workflow. Either take steps, remove the workflow all together and so forth. And the metaphor that I had in mind when I was thinking about the lawyer, the legal stuff, I’ve got a lawyer who’s been my lawyer for that business. Lawyer for the last 20 years. He’s still in Israel. He’s a partner at a very large firm, and because of the time dear friends, basically, most of what I need him to do, or he needs me to do, is either he, if he needs me, it’s a DocuSign, if I need him, it’s a WhatsApp go figure something out, and he. The entire workflow is done by a simple chat message that’s called the assistant, the assistant model. So I think the key question is, all these SaaS workflows where I need to log in, look at the UI, click on something, put something in.

AI Assistant Model

Can that be reduced to a text message to someone who is going to take care of everything for you? And these are the two big metaphors in AI today. I’m sure you all know that, but just in case, we want to be on the same page.

One is the assistant model, the second one is the agent model. And the Holy Grail is the combination. I’m asking an assistant to do something, this assistant go goes and do that. Go find me a domain that’s relevant for my business. And if it’s under 1000 bucks, just buy it. Which is very different than, I mean 5% if you’re going to do this business, okay? The domain purchasing thing. No, I just came on the fly. So, you know, finders fee.

So, long story short, I think the opportunity is in rethinking workflows and removing significant component of the because machines now can speak and understand much better. Why would I need to go to a dashboard and click on something? Why do I need to go to a home page and click on something? Why can’t A, it would anticipate what I’m looking for, or B, if I just tell it, it will do what I needed to be. And then add all these enterprise controls and make sure that I’m allowed to do those things and and so forth.

And then it goes and asks people, and then brings the result, which is people don’t like the beta. I mean, I’ve done this during fundraising. What do investors think about this new idea? So at the end of this, would you like to see a demo? Great. I’m sending them this perspective, and I got the true feedback of what they really think about me. So that was good. And, you know, some of them thought I’d need to lose some weight. But I don’t know that’s I don’t think it’s appropriate for VC to comment on that.

So in my case, and the way I apply this in perspective, is the way you create this prospective interview is you basically ask a question. I’d like to know if this idea is good, or I’d like to get feedback from beta customers about the new release. And that’s how I want this to end.

The other part, so we kind of talked about product and then what do incumbents do? So I talked about what startups needs to do. They have to basically disrupt existing if I can disrupt the Salesforce business model UI, if you don’t need to see it, that gives me an opportunity to be closer to the user. So what would incumbents do? This is a very this is a very good question. V1 that everyone did is some sort of auto complete version of took all the input fields in the in the in the screen, and try to kind of create this, either generate something, or try to predict something, and so forth. I think it’s nice. It’s good. But, think about a new version for HubSpot, like a marketing automation tool. Which basically say all the new user that signed up today because they heard me in Business of Software, send them a welcome email plus a coupon for 250 Credits, or something like this.

I’ve built campaign automation in the past, so who worked with campaign automation? That’s like a 10x faster than any campaign automation system that I know of. I’m sure that there are others, but. So, my point is that’s one thing.

Now, what do you do with the fact that AI is unpredictable, or it’s not, it’s not a it’s not a system of it’s not rule based, right? And in a regular UI, you would click, you know what’s, the bug will reproduce itself all the time, again and again, because the rule would create this bug. That’s the joke, but the idea is, and then people talk about hallucination. How do you integrate this all together?

Here again, I think that the metaphor of assistant like a human employee is exactly the same. Because when I talk to my assistant, I mean, I don’t have assistant, but if I had assistant, I would tell them something, and they would play it back to me. I didn’t get this or so you want me to book you a ticket? No, no, no, no. I meant for you to do something a little bit different. That is very human. And if you think about this, open AI anthropic Google are spending billions of dollars, 10s of billions of dollars of educating us all that we can. But AI doesn’t always get it right. And you can re introduce, re ask them something. And when I mentioned this anecdote to my developer, which I have now, I thought I’m so genius, I got this insight. And he said, Yeah, but we’ve been doing this for forever with Google, right? We would search and go and search again. So the interaction model is changing, which makes it very difficult for incumbents, because if you don’t incorporate, if you apply rule based system into a non more statistical system versus a predictable system, you’re trying to do things that this system is not designed to do. And for that reason, it’s more difficult.

One we mentioned cursor, the development assistance. One of the things that they did super, super well, is they provided the AI workflow of developers, plus the chat, plus exposing which engine you’re actually using. So you know that Claude gets it wrong sometimes, and you can ask it to do something again. So I don’t have, like, 100% answer for the for the incumbents. But I do think it’s, it’s something worth thinking. We’re moving from a metaphor of click through GUI, to a metaphor of an assistant. Doesn’t mean that doesn’t have GUI, but it’s a little bit different.

Distribution Challenges

What does it mean for distribution? What anyone has an idea? AI for distribution. I’m really looking for for ideas. I think the downside in distribution is because it’s so easy to build new products. There are so many new products in the market, and it’s much harder to get above the noise. So that’s an advantage for incumbents, but it’s super difficult. I think it’s it becomes much more difficult to stand out.

And another aspect of distribution, Google is killing SEO, right? You’ve all seen that. When you search now in Google, it simply tells you the answer, which basically means that it’s not going to send you to people that wrote blog five ways in which you should do something. And that basically, as a distribution channel, becomes a non available distribution channel, which is very good for them. Because you’re going to put more money on ads, but it’s not very good for us. It takes a channel that was very, very meaningful, very important. So all these ideas, I’ll build AI that’s going to create a lot of content, so I’ll rank up in Google. It’s another problem, because if other people are doing this, it’s much harder to rank up. But still, it’s a challenge that is becoming difficult. So I think distribution is a challenge.

I think team, kind of we talked about this. So we talked about the cost that we can build much smaller teams.

Future of Business Teams

Have you heard about the single founder unicorn? Have you heard about this? It’s this conversation among the elite of Sam Altman about who’s going to be the first company that’s going to reach a unicorn status with a single person, which basically means that people believe that you will be able to build soon using AI agents and so forth, a business that is much smaller.

The other part that I want to talk about team, was discussed in the previous session, is how do we make AI human friendly, versus taking jobs away from the economy? I mean even just for the for the purpose of being able to sell better, the perspective.

Amplifying Humans, Not Replacing Them

I wanted to build AI that actually amplifies humans and not replaces humans. And the reason is, I still think that humanity cares more about humans than anything else. I think it’s the CTO of Microsoft that said that you could do Formula One, going back to self driving cars, you can do Formula One today with autonomous driving, but nobody is going to watch. We don’t care. We want to see the drama between Verstappen and the rest of the crew, and Daniel Ricardo almost crying because it was his last race, and so forth and so forth.

So the idea is, we care about humans. Let’s build AI that amplifies human. I thought it’s a good message. This is why perspective, this is what we do. Because I think it’s there’s going to be a lot of pushback within companies to adopt AI that is going to take my job. I think people are going to have, it’s going to be harder. So that’s on the team side.

And the last one, which has been discussed heavily is, how much money do you need to start a new company? Or how much money do you need to run a new company, right? So maybe this entire VC model is dead. You don’t need $100M, $200M, $300M to build a business and distribute. Maybe all you need is, I don’t know, $5 million to get to a billion dollar value. It’s a good question, and none of that matters as a rule or like, that’s how it’s going to be. I think if we look at that as a trend, then the question for you is, what does it mean for me? What does it mean for my business? What does it mean for my my my future?

Because I think the trend is there. Whether we get there in six months, we get there in six years, that’s the trend. This is all going there’s going to be way more companies with much smaller teams. Innovation pace is going to be very, very fast, and it’s going to be very hard to compete, because you can see that even the AI models. What’s the true difference between chatGPT and Claude and Gemini? I’m sure anyone has you guys feel true difference, I think most people are going to say the true difference is this cost 20 bucks a month, this costs 15, this costs 12. So I’m going to go with the 12. And that goes to the other side here, of the demand, which kind of puts a lot of pressure. If we are software executive, it puts a lot of pressure on us, because the supply becomes much more frothy. Which means the power goes back to the customers. And then the question is, how do you how do you acquire a customer? How do you retain a customer? What does it mean? Which puts a huge price pressure, which puts a huge service pressure, which puts huge power in the hands of customers. And these are all trends, and I’m happy for you to fight me on that, but I think this is what’s happening.

So when I look at what’s great, what’s going on super well right now is, it’s cheaper to run a business. You can build great support with AI. If you care about innovation, you can move super fast. You can use AI to get into new markets that you could get before, because you need to have services tied to them. I mean, McKinsey type of businesses, small companies, cannot afford that, now they can. Because they can use technology that does that, which is what SaaS did initially, right? I mean, the way Zendesk came about, it was a mid market play for or low end market play for customer service, much better experience and data driven decision making.

No Technical Moats

What’s not so great is that competition is harder, price pressure. I mentioned higher investment, initial investment cost depends on where no technical Moats. That’s another thing that everyone’s trying to configure. How do I compete? You’ve probably heard, but who would I have this idea, right? You build this AI that basically records everything that your sales team is doing on your on their UI, and with an agent, it builds a replica of Salesforce, but it’s just much better. It does what you want to do only, and you roll this out, and you and it’s, and you can do that right. And there’s a company in Europe that has been discussed that that’s what they did, venta. So that’s a hard you know, technical mode is becoming much more it’s harder to find where it actually lies. For sure, maybe it exists, but for sure, it doesn’t exist where it used to exist, where I’ve got the best developers who know how to write the best code. I’ve got the best designer.

I actually think that the potential mode is I’ve got the best process that enables me to understand customers better and I can move faster. In other words, is user experience the outcome the UI, or user experience is the process that produces the best UI? And, yeah, you can read the rest.

So, long story short, there’s a lot that is coming ahead.

If anything that I said provoked you to think about something and you think I’m right or mainly wrong, let me know, because I’m learning as well. I want to build a great business in this new time. And if I can help anyone, I would love to do that, but happy to talk and help out. And thanks for listening.

Guy Nirpaz

Founder & CEO, Perspective AI

Guy is a Silicon Valley-based Israeli entrepreneur who was founder & CEO of customer success platform Totango for 15 years before exiting to PE. He founded Perspective AI in 2024 to build a different company in a new, healthier way, enabled partly by the development of AI and AI tools. Guy loves people and technology and has dedicated his career to improving the way in which business is done through innovation. 

Guy moonlights as the lead guitarist in a rock band based out of his garage in Palo Alto and used to grow oranges.

More from Guy.