Over recent weeks there has been a lot of talk about how the rapid evolution of AI could trigger a SaaS-pocalypse, decimating valuations of well known names and even threatening their very existence (somewhat ironically, much of this hype was probably generated by AI). My view is that reports of the death of SaaS have been greatly exaggerated. AI adoption is growing, although not at the rate at which the hype around it has exploded, and organisations are starting to realise its transformative potential but a SaaS-pocalypse seems more like a clickbait headline than a reality.
There is no doubt that tools such as Lovable, OpenClaw and Claude Code can be extremely powerful, revolutionising the speed at which software can be developed. In fact, I would argue that these types of applications represent the most mature, effective and widely adopted tranche of AI-software to date. However, suggesting that this is going to result in organisations rushing to replace all their SaaS applications with vibe coded custom developments seems fanciful at best. Just because you can do something, doesn’t mean you should.
What is vibe coding? Vibe coding refers to the rapid development of software using AI prompts rather than traditional manual coding.
Even with the ability to develop software more quickly, the fundamentals when considering custom development as opposed to configurable off the shelf software don’t change. There is a need to understand what needs to be developed, how it needs to work and the processes that it will be used for. Code still needs to go through test and validation protocols, AI generated code will still have bugs that need to be caught. If anything, testing needs to be even more rigorous if it’s AI generated to avoid inadvertently impacting other features/code that the AI wasn’t ‘aware’ of.
Governance, compliance, security, logging all need careful consideration and attention. Vibe coding a cool looking app in a few hours is one thing, building a sustainable solution to process huge amounts of content and data for mission critical applications at enterprise scale is another entirely.
There’s then the on-going overhead necessary to deploy an application at scale, manage the infrastructure to run it, ensure it’s available 24/7, keeping the stack current and secure, adding new features, fixing bugs, integrating with new systems, to name but a few. SaaS has been successful as it allows organisations to benefit from economies of scale in so many ways to make life simpler. Is it perfect? Of course not, but I don’t see organisations embarking on a mass development effort in the next few years just because AI streamlines software development.
This isn’t to say AI won’t have a big impact on the SaaS industry, it just won’t wipe it out overnight! There are also ways to mitigate some of the challenges that I referenced above. At Vertesia, we have worked with a number of customers to replace traditional SaaS applications with AI-powered alternatives. We’re combining the best of both worlds by helping customers to deliver custom projects but without the need for custom code. When applications are built on a platform, a lot of the heavy IT lifting is already taken care of so delivery is far faster than normal. Also, we use AI tools to speed up delivery further, for example using Claude Code to build application specific UIs incredibly quickly.
As multiple applications are developed on the same platform, this substantially reduces on-going operational costs and complexity, solving for many of the issues traditionally associated with bespoke applications. Not only are customers able to deploy more powerful applications but they’ve also benefited from cost savings from a combination of Vertesia’s modern architecture reducing unnecessary compute overhead and a usage as opposed to seat based licensing model.
I also expect to see more edge cases that can be addressed with AI technology. Use cases that were previously too company specific or niche for a SaaS solution but too complex for cost effective custom development. Again, we’ve seen several examples of this at Vertesia where the outcome can be delivered more efficiently with AI and the ability to rapidly deploy on a centrally managed platform leads to a viable ROI. There is an argument that these could be custom developed more cost effectively now but that can result in a plethora of applications needing on-going management, multiplying cost and complexity.
The need to adopt a different commercial approach and downward pressure on renewals are already starting to impact traditional SaaS providers. AI moves the value away from data storage, structure and users, which is the value bedrock that traditional SaaS is built on, to automation and insight. This presents a big risk to traditional SaaS providers, evidenced by their sudden rush to prove that they can ‘do’ AI. It will be fascinating to see whether these providers can pivot quickly enough to deliver the innovation customers want by bolting on limited AI capabilities to 5, 10 or even 20+ year old architectures or whether organisations will look for more modern, AI-native solutions. It then becomes a choice between augmenting or replacing existing solutions. There’s also a big question around skills, developing ERP/CRM/ECM solutions is very different from building AI-native applications from the ground up.
In summary, there is no doubt that traditional SaaS providers will be affected but I don’t expect a sudden extinction event even if it does make for a great headline! AI will continue to have a profound impact on the way that work is done and the tools that we use to do it. We all need to adapt and the success of many organisations, SaaS providers included, will be dependent on their ability to do so.
I want to end by saying that, for better or for worse, no AI was used to write this blog. Love it or loathe it, agree or disagree, it’s on me. AI can be incredibly powerful and it’s currently providing me with my livelihood but I still believe there’s a place for real (if limited in my case) intelligence as opposed to the artificial kind!