"Why should companies pay for SaaS (HR/CRM/ERP/etc.) when they could just vibe code them?" I get variations of this question or comment with some regularity (granted, it's sometimes just me talking to myself). Here are some biased (but hopefully, well-considered) thoughts: 1) I am a big proponent and user of vibe coding (what I call "agentic coding"). I do it every day, 7 days a week, including Sundays. It's amazing. 2) My company, HubSpot is a software company. We have hundreds of professional engineers -- just about all of them use AI for product development too. They are brilliant and know how to build production-grade products. 3) Even with this powerful army of talent, the number of internal, core SaaS applications that we have replaced with a vibe-coded variant is exactly ZERO. The number of applications we plan to replace is also exactly ZERO. 4) It's not the absence of talent that keeps us from rolling our own SaaS apps, it's the presence of focus. It would be silly to try and replace our HR, team collaboration, expense tracking and 100+ other SaaS apps we use when we can just buy them. Just doesn't make sense. 5) That's us -- as a software company at some scale. If you're a non-software company it makes even less sense for you. Doesn't matter how good the AI coding tools get. Let's say you *could* vibe code a replacement for that SaaS app you're using, who's going to maintain it? Who's going to keep up with industry trends? What are you going to do when the 20-something genius that vibe coded it over a weekend leaves the company? Who do you call when there's a major bug? 6) If you're a Fortune 500 company at some scale, perhaps you could pull this off for some discrete use cases and the tradeoffs are worth it. You have an IT/Engineering department that is larger than the population of some countries. You can take on the pain in return for the positives. For the millions of others, my advice is: Spend every calorie possible on creating value for your customers.
The last sentence says it all. I am continually baffled by the internal build strategy, especially where it is at best "build to parity" what can easily be acquired from any number of vendors. And that doesn't even scratch the surface of how you will then continue to test, patch, iterate, maintain the internally built product. The winners in enterprise will be the ones who are very opinionated about their architecture and data fabrics, and create frameworks where vendor solutions can be dropped into (or swapped out) rather than trying to build the application layer themselves.
It’s much more complicated than this. It’s very early in vibe coding, and most haven’t really done it - ai studio in Gemini and agent 3 in replit were both big shifts forward. Most companies are already on long term contracts and can’t make this switch this easily. We are still figuring out how to build solutions long term and the role of vibe coding in software development lifecycle. Does it replace the whole thing, or just certain parts. But the fundamental reason why people attracted to vibe coding over saas isn’t because of cost - it’s because saas never really fits perfectly for them, because it’s meant to fit for everybody. So you ‘accept’ a solution but always wish it looked a different way, and was built for the user, not for a scalable architecture. Companies can’t hire full in house teams to develop custom software - so we have saas. But vibe coding make this in theory seem possible if not right now, within 6 months. So much saas is a database, front end, and workflow engine with a permissions access framework embedded. This is so prolific in the training data that it’s hard to think of any reasonable argument that saas won’t get significantly disrupted. But - SaaS just needs to solve bigger problems.
I couldn't agree more! Plus, some of your SaaS tools -- like your HR platform -- hold highly sensitive and compliance related data, so you don't want to mess around with that. Rather choose a vendor who's core business it is (1) to ensure the data is safe & handled correctly, (2) all workflows & analytics work correctly and (3) they built an AI layer on top that maintains all relevant roles & permissions.
Even in a Fortune 500 company with unlimited budgets and a large engineering team, you still need a champion who is willing to commit years of their time to the product. I’ve seen one successful internal build of a SaaS at scale, but the champion behind it was a force of nature - driving the build, distribution, adoption, and ongoing maintenance. He essentially dedicated his career at the company to making it succeed. Code generation was never the bottleneck.
Completely agree, Dharmesh. Vibe coding and agentic tools are brilliant for quick internal workflows and experiments, but in an enterprise you cannot vibe code your way out of security reviews, compliance, data governance, audit trails, observability, and boring things like SLAs and support. Once a vibe coded workflow becomes business critical, you have created a real product, which means ownership, monitoring, incident response, and change management. Over time that usually increases headcount instead of reducing it, because you now need someone responsible for keeping up with changes (system/regulatory), platform updates, and evolving business logic, and the risk when that person leaves is very real. At scale you also get fragmentation, ten different versions of the truth, and a new kind of shadow IT that just happens to be built with AI. We build AI driven products every day and I still feel the same way. AI will absolutely reshape how we build and extend SaaS, but the systems that carry customer data and revenue still need strong product thinking, reliability engineering, and clear accountability, not just vibes.
Dharmesh Shah a good friend of mine has build a start-up. He is a talented developer and has vibe coded a CRM. It's cool. Now he is hiring some sales people. Two problems emerge. 1) He doesn't have time to continually update and develop the CRM as and anyway he doesn't know what best practice for CRM is. 2) It has a UI that only a developer would like - his sales people want a destination to go to everyday that looks nice and is consistent with the typical sales process, not just a chat interface. Now he is moving to HubSpot.....
Even with AI making development easier, the real cost isn't code - it's compliance updates, security patches, and explaining to the new hire why the system works this specific weird way. Focus is indeed the scarcest resource.
This is the part most people underestimate: building software is easy maintaining it, scaling it, securing it, and supporting it is not. AI can help you ‘vibe code’ features, but it can’t replace the ecosystem, reliability, and long-term ownership SaaS provides. Focus is still the ultimate competitive advantage. Dharmesh Shah
Totally with you on this. Building B2B software every day, I’ve had the exact same internal debate. Agentic coding is incredible. We use it constantly across our teams, and it unlocks crazy velocity for prototypes and internal tools. But when it comes to replacing full-blown SaaS… same conclusion as yours: still zero* replacements. (*to be honest, we did replaced a couple of them but nothing crazy 😏) It’s not a talent gap, it’s a focus problem. Real products need uptime, compliance, support, security reviews, roadmap continuity, deep expertise in the domain… all the invisible work that vibe coding doesn’t magically solve. From a sales perspective, this post is gold. The “we can just build it ourselves with AI” instinct is becoming a potential blocker. Having a clear narrative on why production-grade software is more than code will actually help reps handle those conversations with confidence. Love this framing. It’s simple, true, and exactly what teams need as AI pushes that DIY temptation.