Replit customer at B2B SaaS Company on prototyping and customer discovery with third-party APIs

We spoke with a product leader at a B2B SaaS company who uses Replit primarily for prototyping and validating third-party API integrations during customer discovery.

The conversation explores how Replit enables their team to demonstrate integration possibilities to customers, while highlighting limitations around matching their existing product's design language and the challenges of transitioning from prototypes to production.

Key points via Sacra AI:

At a high level, how is Replit used at your company today? What are the top few use cases?

It's primarily used in a prototyping use case. The places that have been most useful is when we want to test out third party software or third party APIs and be able to use those in discovery conversations with customers, which would be something that's difficult to do without an actual real product. We're starting to experiment with using it in the go-to-market use case, where I think there's maybe more opportunity for the product.

Can you tell me a little more about that go-to-market use case? What kinds of prototypes are you building? And who on the team is using them in customer conversations?

I'm the products person, and using Replit in a product context is challenging because it doesn't necessarily reflect the user experience of our products. So it doesn't really matter from a 0 to 1 context, but when you're trying to iterate on our existing product, it's less useful. However, when we have conversations where our customers are attempting to envision how our product might integrate or be connected into their product, that is something they really struggle with.

Replit can help our go-to-market organization tell that story, help a customer envision how our product could exist within their product. That's interesting because it's not precisely accurate, but the alternative is we have to either get a designer to make slides and images, which is expensive and time consuming, or tell them a story with words, which many people struggle to understand. The go-to-market use case is more valuable in a lot of ways than the product use case right now, given the limitations of these platforms.

What do you think is the biggest misconception you hear about Replit in business environments?

I don't think I hear misconceptions about Replit specifically. There is a sense with AI coding in general that the designer's going to go away or the engineer's going to go away or the product person's going to go away. Obviously, there are limitations in all these platforms. The biggest gap for me as a product person is incremental iteration on our existing product, which is very data heavy, B2B, and not open to the public. Even the ones that focus on it, like Magic Patterns, aren't particularly good at replicating our existing design and allowing us to iterate in an incremental fashion.

Who on your team tends to build with Replit day to day? Is it more product teams, engineers, or nontechnical roles? And is it the same stakeholders who sponsor it financially?

I didn't totally get that question because you said, is it product people, designers, or nontechnical roles? I think those are all at least in our context relatively nontechnical roles. Engineers aren't using Replit. Product people are using Replit to validate concepts. And then if we want to do anything beyond that, engineering is involved.

Which tools did you consider alongside Replit?

I would just interject there—that's not what I said. Product is the primary user among the groups you mentioned. But the team and the org that's going to use Replit and similar tools the most would actually be our go-to-market org. Product people have always worked with engineers, so it's an easier, faster way to work with engineering. But the marginal impact with go-to-market teams, who haven't had engineering coverage for a long time, is much more significant.

Which other tools or platforms did you consider before deciding on Replit?

I think it would be going too far to say we've settled on Replit. We've tested out V0, Lovable, Quad Code (which is not exactly a competitor, but there are overlapping use cases), and Magic Patterns. The one that I probably use the most of those listed would be Magic Patterns, but I think there's a lot going for Replit. The one that I've been least impressed with is Lovable.

What has Replit done well for your needs? What would you say is its greatest product strength? And conversely, its weakest or most frustrating aspect for the types of work you're doing?

Replit is more useful when you want something that's more full-featured. If you want to build a real app simulator, Replit stands out in that space. When we wanted to do customer discovery with third party services, to validate that some financial services API works in the way we want—and it's hard to validate that sometimes with sandbox data so we want to get real customer data—we'll build tools to use in a customer discovery context. That's been excellent.

When I wanted to tweak my existing product in this way or that way, Replit struggles or has been less effective. But in general, Replit is really good at what it seems to be focusing on, which is more the 0 to 1 use case.

It was interesting—I was just at a product manager's event, and the CEO of a company was there telling me how he had vibe-coded his entire company in a 4-hour session on Replit. It's amazing that's even possible.

Given your use cases, how well does Replit support apps that go beyond prototypes into production environments?

I can't really answer that question. We haven't tried. My engineering team would kill me if I even seriously suggested that. There's a real suspicion about whether these kinds of isolated apps are even plausibly useful for production environments with real, high security production customer software and customer data.

So how do you typically handle that transition from prototype to production?

When we're prototyping things or doing discovery with Replit, it's usually not a production-ready feature. It's about how we can validate a concept, validate a customer problem. It's getting information that feeds into a design document or a product brief. Then engineering will take it from there on implementation. But zero percent of what Replit writes is making it into production software.

Can you share a couple of concrete examples of apps or agents your team has built with Replit as part of your validation and discovery process?

I can give an abstract example of the kind of app we might build. Let's imagine we wanted to build an app where we're going to use Whisper Flow for transcription. One of the challenges with data-centric projects is you need to actually try the product out in reality to see if it works in your use case.

We might build an app with Replit that has Whisper Flow or some other similar AI-driven product integrated into it and then go out and do user discovery with that app as a proof of concept. That would allow us to start understanding what works and what doesn't work in the real world with the results of that product. That information would feed back into whatever our production use case was.

In these scenarios, are the end users for these prototypes internal team members, or do you also involve external clients during validation?

With an AI-driven product like that, we don't necessarily know how to evaluate it before we've used it with a customer. The space is evolving so quickly—there's so much to learn, so much that you don't know.

In that scenario, we would be working with customers or prospective customers and product people to do discovery—guided interviews or guided user research. We're trying to use the interaction with the vibe-coded prototype to gain knowledge that would help us understand how we might build the product in actual production.

Have you encountered gaps where Replit couldn't meet your needs for certain prototypes or use cases?

The most obvious gap is I've not figured out how to get it to iterate within the context of our existing app. In my ideal world, the best user research would be if the customer feels like they're interacting with their own data in our product. I don't know how I would do that in Replit.

It's missing our design language, our colors. I don't know how to bring in our own data to demo it. That's not been a massive blocker, but it prevents us from using it in more of a design context. It certainly limits whether we'd be using it for something that's actually developing something closer to production usage. We still need designers—Replit certainly doesn't replace that.

Where has Replit been the most durable in your workflows?

I think Replit solves new problems that we haven't encountered before, making new things possible. It doesn't accelerate existing workflows in significant ways.

Are there use cases that tend to stay relevant over time as new needs arise?

All this stuff is so new, it's hard to say what's going to be durable. Where it has been most useful is the discovery use case I've described a couple times already, helping us validate particularly external APIs where the UX doesn't matter, but the API or partner integration really does.

It's really valuable when we're going to do an interaction with Persona or Plaid or Modern Treasury or Stripe or something where it's a third party service and we want to convey the breadth of that experience. Replit is really valuable because it's difficult to explain or expose those services verbally or with other AI-driven products.

When the elements you're trying to explore are more standardized and more third party and are well documented in public ways, Replit is strong. Where it's less good is when they're more unique to you. It's not in your development environment, so it can't provide a lot of value there.

Are there other areas or tools where Replit intersects with your workflows? For example, connecting with automation tools like Zapier or n8n?

We're pretty heavy users of n8n and less heavy users of Zapier. Totally unrelated, not connected to each other at all. I could imagine how Retool and n8n might work together, but it's not obvious to me how Replit and n8n would work together.

One pattern teams use is they build an interactive UI front end or simulator in something like Replit or Retool and then use n8n behind the scenes to orchestrate logic, like pulling data from APIs, transforming it, or triggering workflows across tools. Does that mapping make sense for your workflows?

That seems like an awful lot of work. It's hard for me to imagine doing that, frankly. If we were going to do that, I would figure out how to get Replit hooked up to our internal APIs and use internal APIs rather than pseudo APIs via n8n or Zapier.

I get that conceptually, but it feels like really creative problem-solving. I would be skeptical that any organization with actual engineers would think that's a good idea. That sounds crazy to me.

There's engineering skepticism on whether a vibe-coded app can meet our engineering standards for security. But using n8n for your APIs is different. We use n8n a lot and trust it. But to say that's how you're going to let a vibe-coded platform interact with your app feels like you're stitching hack on hack on hack. That's me saying that sounds crazy, not my engineers.

How do you currently think about the value you're getting from Replit? What tells you that it's been worth it for the kinds of work you're doing?

We're using their lowest-end plan—a monthly plan per user. We don't have a corporate plan with Replit, so it's cheap and affordable. I just expense it. People on my team who use it just expense it. None of us use it enough that it's a material cost. Our OpenAI API costs and Anthropic API costs are much more significant, and that's where we put our energy.

It's almost an obvious one. I think I'm paying $20 a month. That's irrelevant in the scheme of my budget or even my team's budget on a per-person basis. But once you've got people trying Replit and Lovable and V0 and Magic Patterns and whatever no-code builder, and you're not actually using them on a regular basis, the numbers become a little more offensive from an accounting perspective.

I mainly think about costs from the perspective of being a responsible person, not from the perspective of these being really significant to me as a business.

Anything we haven't covered that you'd want founders building in this space, or investors evaluating companies like Replit to understand?

As the corporate leader of my function, we are thinking at a corporate level and have those conversations. But if you've got a small number of people doing something at $20 a month, it doesn't add up to something crazy.

That said, every AI product is attempting to charge an additional $5 to $30 a month. The cumulative impact is significant. While Replit at $20 a month is fine if people are finding value in using it, the bigger question is when people are using Lovable, V0, Magic Patterns, plus Cordova, OpenAI, and 40 other services. Suddenly you're at $100-300 a month in per-person spend. That's when the numbers start getting really significant.

If I'm one of these AI products like Replit, I want to be the last company standing because there will be budget compression eventually. These budgets can't expand forever.

What would Replit need to improve to potentially gain greater foothold as you face these cost and usage considerations?

The discovery use case that I discussed with you is intense but a relatively low frequency scenario. If Replit was better at building internal UX and designs, that's one area of opportunity. The other is building internal tools, where they'd compete with Retool. The discovery use case is too narrow, and I would hope they expand beyond that.

Anything else you'd want founders building in this space, or investors evaluating companies like Replit to understand?

It's a really crowded space. It's been hard for me to differentiate what Replit is good at versus bad at without trying all these products. I'd be curious what their marketing goals are.

It's such a bloodbath in the space. How they actually differentiate themselves and how they plan to go about distinguishing themselves seems really hard.

Which education formats have worked best for your team in adopting Replit or similar tools?

Live training from internal champions has been the most successful because they understand our specific business context. That depends on having a small number of people who go out and experiment with the tool, and they come back and bring it into the product.

If I were Replit, I would be asking myself how I can nurture those champions and equip them to do a really good job internally with a land-and-expand kind of model.

Have you seen moments where teams or users were surprised by what Replit could do, whether positively or negatively?

Every time I've used it, when I start sharing it with people who haven't used the products before, they're surprised by it. I saw that Replit just announced a v3 of their product yesterday, which I haven't had a chance to use yet, so I'm really curious how far this goes. It definitely feels like the state of the art is advancing pretty rapidly.

The part that Replit excels at is the full stack hosting environment where you can build a playground and then actually have real customers going through it. That really does impress people because they don't expect there to be a database or other things standing behind this product.

If you had a magic wand and could improve one specific feature or capability in Replit to make it more valuable for your organization, what would it be?

Two things that might be most impactful:

1. A way to bring our own data into the product—either a BigQuery connection or ability to really easily do CSV uploads that we can swap in and out. My experience is that once you load data in, it's hard to change in Replit.

2. Figuring out how to really easily bring in components from our design system so that the apps feel native would make a big difference.

Do you think Replit could take a stronger position in helping users with incremental iteration?

I currently use other tools for that, but if Replit could do it well, that would be a great asset.

This transcript is for information purposes only and does not constitute advice of any type or trade recommendation and should not form the basis of any investment decision. Sacra accepts no liability for the transcript or for any errors, omissions or inaccuracies in respect of it. The views of the experts expressed in the transcript are those of the experts and they are not endorsed by, nor do they represent the opinion of Sacra. Sacra reserves all copyright, intellectual property rights in the transcript. Any modification, copying, displaying, distributing, transmitting, publishing, licensing, creating derivative works from, or selling any transcript is strictly prohibited.