For the first episode of the Axeome podcast, Michael and Eric talk about the advantages for stratups of using AI tools and what can be expected about their integrations in the market in the years to come.
You can also read our complementary blog article that dives deeper into available AI tools on the market and some limits entrepreneurs should keep in mind.
English subtitles available on the video.
Transcription
Michael Anthony-Clement
Hello everyone. Mike here with the first episode of Axeome from the Axeo incubator-accelerator. And I have with me Eric Chocron who is one of the members of our team, director of programs. And today, we're going to talk about AI tools that we can use as founders to accelerate the prototyping and validation of our startups. So with that, let’s get into it. Eric, do you have any initial comments on the use of the AI and how important it’s going to be?
Eric Chocron
Well to begin with, contextualizing what was happening in the past and what AI actually improves, I’d say would be a good thing to do. Contextually, when creating a prototype, an entrepreneur can themself look for a technical cofounder who can code and develop the prototype. But we can agree that it takes a lot of time to do that. In the past, it was like that. They can also look for a firm to work with them for a fee, or that could be partly subsidized for the startup to develop a prototype, if it’s sufficiently innovative or has the potential to drastically change a field or an industry. And they could also offshore it. In French...
Michael
In English.
Eric
Yeah. [Laughter] Basically employ people outside the country to help at a lower cost. But once again, all this IP that they’re developing, it would be outside Canada, which would come back to them partly through licencing. And also, delays will be quite lengthy. Which brings us to today, in 2025, talking about AI tools. It kind of gives super... It gives superpowers, in fact, to entrepreneurs who are in the prototyping stage, in the stage of wanting to set up an MVP, a first demo, a first way of doing things. And I think that you yourself, Michael, have had the chance to test out a tool.
We always talk about ChatGPT, in the past or in the present, which can help us code if we don't know code. There is also Microsoft's Copilot, which is also based on the same principles as OpenAI. In fact, it’s the same company behind it that manages it. So you can code Microsoft applications with Copilot. You can code general applications with ChatGPT.
But today, Mike, rather, you're talking to us about something else, and I want to hear from you, your vision, your perception of a new tool that you had the chance to use last week and the weeks before. So go ahead, go for it, and then I'm going to make comments on that.
Michael
Yes, perfect. And precisely, I wanted to... I saw Replit being mentioned. Replit is a tool that comes out of the United States. It was someone I follow on X, on Twitter, formerly Twitter, who said, "Wow, this is going to change the game for developing startups, it's going to speed up development, it's going to reduce costs." Then I thought, well, I want to fiddle with it, get my hands on it. I want to try it before I reach an opinion. Then what I set myself as a goal was... I subscribed to the Replit app, and it's really amazing.
It's like a normal mobile app. And it asks: "What do you want to build, Eric?" Then it's like, well, I want to build an application that can help me manage my incubation programs, and I think it should have this, this, this feature. I write it in normal text, no matter the language. Then it starts to build it. But it's not just a question of data, like the code that you then... Like in ChatGPT, it's going to give you a piece of code, and then it's going to say, "Insert this into your project." No, no, Replit will build the project from A to Z. The backend will open servers, it will open databases, it will start to do the frontend. Then it's going to say, Well, OK, version 1 is ready. Would you like to review it and tell me what you think?
Then inevitably, there are bugs, there are... You know, things that don't work. But what surprised me with Replit was that I’d say, "Well, no, it doesn't work, there's a bug, change it." Then it took maybe two, three, four times, but eventually the bugs were fixed. Then we moved on to another version. Now it's like, OK, we want to add another feature, we want to connect coaches with startups. Perfect. There, we started another cycle of creation and debugging. Then again, after three to four tries, we had a version that worked with a calendar. There, I could click the button, I could connect with a coach.
It was really... You know, we often say, when using certain technologies, that there's an AHA moment. I had an AHA when using Replit, because I thought to myself: as a founder of a non-technical startup, I am more on the business marketing side, and this gives me superpowers. Now, I don't necessarily have to depend on a technical cofounder or, as you said, offshore it to the Philippines or Vietnam.
It can allow me to build for... You know, we've already done projects together, but I think it's going to reduce the cost by a hundred... ten to a hundred times easily. Then the iteration time, probably ten times faster because you can do it on the fly. Now, I didn't test it for more than a few hours, so obviously I didn't validate it to the extent of building a project from A to Z. I may have done A, B, C. I have implanted [sic] some features that seem correct, but it’s showing me that it will be... if it's not possible right away, it will be possible very soon to make an application from A to Z without having technical skills.
Eric
If I can ask you a question, compared to the no-code tools that exist... Because that's an oversight on my part, when I was talking about the tools that are available to entrepreneurs. We've actually promoted a lot of no-code tools to our entrepreneurs, Mike, whether it's Notion, Airtable or Bubble to create their first MVP, or Figma to create a first iteration of a user story or journey through an application. If I compare this to no-code tools, is Replit, or apps like Replit, have better capabilities for entrepreneurs? Or is it the same thing? Does it save me time?
Michael
That's a good question. I think that in the end, you have to evaluate what are the needs you are trying to... what are the needs of the market. Because we agree that, if we go back to the beginning, when we launch, when we want to validate a startup, it starts with the market and the problem. So, the real question you should ask yourself as a founder is: what is the best tool I can use to solve the problem for the least cost, just to validate if I'm going in the right direction?
There may be answers that are: yes, Replit is more suitable because we need slightly more nuanced features that Replit could develop. It really takes some custom code. If it's something a little more generic, maybe no-code tools like Bubble, or you mentioned Notion, could do the job. So it comes back to the need, it comes back to all that. A tool, a technology, is just a tool. And we should be happy to use it, no matter the tool, when it’s going to solve the problem we're trying to solve for a market.
Eric
cofounder, technical or non-technical, Mike. Do you think it's a good idea for them to do everything from A to Z? You know, do the conceptualization, the vision, the perception, the added value of your project, including the... the polls, then to go behind the scenes to put it on some type of Replit? Or do you think that on your own, again, even if it gives you superpowers, it starts being a lot for you to manage?
Michael
Yes. That's a very good question. And this question will come up often. I think it was Sam Altman, the founder of OpenAi, who said, "We're soon going to see the first billion-dollar startups that are built by one person." And that's like, wow, that’s a lot of leverage that you can generate with existing tools.
I said that during a presentation yesterday to college students. I said, "Try to go as far as possible with the least amount of resources." That could also mean... It can also mean something like: do it alone for as long as you can. And the simple reason for that is, if you can, now, as a marketer, develop an aspect, like a technical skill with Replit or a tool like it, your feedback loop of learning is going to be so tight. That is, you're going to be able to talk to the market, take in the essential information, apply it right away to the product, do a few iterations with Replit, go back to the market, not two days later or two weeks later, but two hours later. "Hey, thank you again, thank you for your patience, here’s a new version." "Huh?"
And that, an entrepreneur who’s able to do that will have a huge advantage. As soon as you bring other people into your team, you complicate the matter. And that can be one way of looking at it. But you also increase the ability of the overall team to do more. And I know that you probably think in that way.
In other words, I see advantages and disadvantages on both sides. But generally speaking, with these new tools, I mean... My instinct makes me say: go as far as possible with these tools in your validation to keep costs low, but also to learn as quickly as possible.
Eric
I appreciate the comment, really. And yes, I'm going to be more on the team side, or at least find a cofounder to help you in your path, in your journey of discovery of building the MVP, the prototype, really. But I was going to say, actually, it’s pretty crazy. Last week, I saw from an influencer that you and I know on LinkedIn, Greg Isenberg, who actually reposted something on LinkedIn that talked about a job where they had for a role to fill for an AI agent. Well, for an AI. Then he said, ah... He said, "We were close to seeing this, but this is the first time I've actually seen this." He posted it saying, "Look, the position applies only to AI agents and serves only AI agents. And it says: if you're human, don't apply to this position. It's only for AI agents to fill." In fact, it was--
Michael
There’s prejudice there.
Eric
That's right. [Laughter] That's right. But was what I was going to say is that it goes along the same line as: can your cofounder be, in fact, the AI? Is that... You know?
Michael
Very much so. I think we're going exactly there, and it's not going to take as long as we think. I'm trying to stay informed--
Eric
But in Replit's logic, I find that... You showed me a few screenshots and how it worked. It looked a lot like a ChatGPT agent. But if your cofounder was an AI agent, it will ask you the right questions continuously instead of every time: what do you want to build today, or forgetting some stuff from the previous conversation.
Michael
No, that's right. Replit is like that. It's like, "Hey, I implanted [sic] the things you asked me to do. What do you want to build next?" And I even think I could ask for suggestions, and it’s going to give me good suggestions, because it’s going to understand the context of the application that I'm building, because I've given it that context. So yes, I think we're going toward a world where we're going to have to get used to collaborating with AI agents who are cofounders, colleagues...
Eric
Employees, colleagues.
Michael
Employees. And I don't think it's going to take that long, Eric. I try to stay on top of trends. I think we're at two to three years before it becomes quite common. We're going to hear a story, and then we're going to say, oh yes, that's another employee.
Eric
Question for you that you think will happen in the future to help entrepreneurs. Is it Replit and applications like Replit that are taking over? Or — and here, that’s my insight on what I see coming up really soon — is it the AI agents of each of these platforms who are employees of your company?
What I mean is: let’s say you already use the Notion platform for your backend, why doesn't Notion's AI agent take all the bugs, take all of that, and then do it for you? It becomes an active agent. You pay it monthly. Instead of paying it like an employee, you pay it, I don't know, a little less because it's a robot, but you're going to pay it. It’s your colleague.
You say, "Here's a bug that a customer reported." You send it to the Notion agent and the Notion agent fixes the bug. And in this same reality, even further: why isn't it a Slack agent, where you already have your bug channel, or a helpdesk agent, who becomes an active agent on your platforms, who will tell the Notion agent? So, the Slack agent talks to the Notion agent, the Slack agent talks to the helpdesk agent, and there's no more human doing anything, really. [Laughter]
Michael
It wouldn't surprise me, Eric. It's just obvious, you know. And the reason is very simple. An AI colleague will be more efficient and much less expensive, will make much fewer mistakes, will not need to sleep, not need to take vacations. And you know, it only costs electricity in the end, right?
Eric
Well, computing power. It's computing power. It costs the card, you know, the graphics card behind the...
Michael
Yes. But that's going to be amortized, it's going to be amortized over several years and several uses. But in the end, the main input will be electricity. In other words, I'm just trying to say that it's going to be so inexpensive that your colleague will be paid... You're going to be paid 100,000 a year, your AI colleague is going to be paid... $47 per year.
Eric
But, in fact, we're joking around, but... And we're not there yet, but ChatGPT has an automated agent component that costs $200 per company or per person who wants to subscribe. And I've seen, or actually listened to podcasts from The Verge and other electronic podcasts [sic] that say they've used it. It's very, very slow.
But just to give you an idea, when you use it... For example, they said: book a plane for me, a plane to go from Los Angeles to Florida, let's say, or to New York. So, it opens a window on your computer and there, you see it, in fact, clicking on things to do what you asked it to do. But we agree that, first, it's not... The speed wasn't very good, even today in 2025. Secondly, it costs $200 a month for one individual using this agent. Thirdly, it wasn’t getting the best deal, it wasn’t getting the best time. So, human intelligence was not yet adapted. But yes, two, three years or...
Michael
Easy.
Eric
Or four, five. Easily. I see it doing it, I see it going. And I see it easily replacing your cofounder, your employee. And the entrepreneurs will be even better priced [sic]. Now, will it replace the entrepreneur? Philosophical question. [Laughter]
Michael
OK, well, I think we have time for maybe one last... maybe one last remark on what you just said. And I’m telling you: it's not just that we're going to be replaced, but we're going to be amplified by AIs. That is to say, you, as a creative person with a vision of the future, it will be amplified, and it will increase the chances you have to achieve your goals. If I come back to the main topic of this video...
Eric
Tools.
Michael
The entrepreneurs who use these tools will be able to build startups faster, for less money. And let me remind you that most startups don't die because of competition. In fact, they're die because they can't find their product-market fit--
Eric
That's number one, by the way.
Michael
That's number one. Before either losing hope or just running out of money to finance it. And often, it's really just a matter of losing hope, because you can always try to keep going and pivot. As soon as you lose hope, it's over. So if we can reduce the costs of development, reduce the time it takes to validate in the market, it increases the chances of success for a startup that develops software.
Eric
Well said. We can leave it at that, actually. Thank you all for listening to us today. First podcast of the new Outaouais incubator, Axeo. Glad to be with you, Mike, as always.
Michael
Yes, sir. We're hoping to do that every week, at least. And we're going to talk about a lot of topics that will import [sic] entrepreneurs. New tools, new trends, how to better build a startup. That’s the focus of Axeome. Thank you all.
Eric
Goodbye.
Comments