Agents will not kill SaaS: They will circumvent it

I don't agree with the current narrative of "SaaS is dead because AI makes it trivial to build software". No, it's not. It's hard to build great software.

I know it because I have built both bad software - sometimes -, in past companies, and great software now at MarsBased. We have been running for 12 years and have built over 150 different projects for companies of all sizes: Fortune 500, one-man projects, startups, scale-ups, non-profits, public administration and SMBs. It's hard. Freaking hard.

Cheap-ass software has always been easy to build. Now, thanks to companies like Lovable, Replit, Cursor and the like, it's accessible to everyone, but maybe because it should've always been like that. Much like Blogger and Wordpress helped to democratise blogging, back in the day, I think everyone deserves to be able to create software. It elevates our existence as species, one new skill at a time.

However, I won't give it to the fatalists and doomsters: enterprise SaaS is not dying because anyone can clone Salesforce over a weekend. You can't compare creating a new weather app for yourself to creating (or replicating) a behemoth of a platform like Salesforce. That is simply not realistic and such affirmations only seek to draw attention through aggressive clickbait and deceptive marketing tactics.

In my opinion, the SaaS market is being challenged for a different reason. Replication is not the real threat. Replication has always been there and it never killed any company in the first place when only the strong & prepared could do it, so how could John, from a small town in Fucksville, only working by himself, take down a giant in the industry?

He can't.

Hubspot, Salesforce, SAP... these are not weekend projects. They are decades of accumulated edge cases, integrations, compliance layers, redundancy, and institutional knowledge. The amount of knowledge thrown into building these products and platforms can't be condensed in a weekend. The cultural nuances, the historic and timely decisions dictated by market changes or the zeitgeist can't be replicated.

Agents do not magically erase that complexity. Agents don't need to replicate SaaS, they should focus on replicating the outcome.

If I want organised music, I don’t strictly need Spotify. I had a folder with thousands of songs, all correctly tagged using Musicbrainz, with their artwork and lyrics, before Spotify came along. I lost it after my hard drive malfunctioned during a backup so I decided never to have songs locally again. SaaS brought me convenience. It spared me from the hard work at the expense of ownership, which is what SaaS is good at.

We have delegated ownership to the SaaS platforms so we could focus on other equally important stuff: moats, marketing strategies, M&A, customer satisfaction, partnerships, multi-channel, client acquisition, creating new business lines, helping our employees be happier and work better, etc.

We have decoupled a lot of value from our company through SaaS platforms so we could collaborate better between companies, and that creates new markets, and new complexities too.

SaaS companies now face the API paradox: If they close their systems, they become irrelevant in an agent-driven ecosystem. If they open them via APIs, agents can extract value without the full SaaS experience.

SaaS companies must enable interoperability to survive, but interoperability weakens the moat. We're weakening ourselves further just so we can play along better with everyone else. For the greater good. But, at what cost?

We are about to see a sea of mediocre apps, like we saw the millions of flashlight apps when the app store was first released. Agents will fuel an explosion of "good enough" software. Remember the beer glass iphone app? Prepare to see that times one million.

Effectively speaking, not everyone will be able to create software, but enough of a large number of people will. That will add more apps to the marketplaces. Some will make the pie larger, some will corrupt the ecosystem. And that's the market that I think will be mostly affected by agentic AI and generative AI software platforms: consumer apps.

But let's go back to corporates and enterprise. Large organisations are not optimised for efficiency. In fact, they are inherently inefficient by design. Redundancies, slowness, politics and other inefficiencies are a feature, not a bug. Corporates optimise for for survivability, creating resilience at scale. One-person projects can't compete with companies with thousands of employees because no matter how many agents you deploy and run, you have multiple bottlenecks: your own time, your mental bandwidth, and the things you don't know: languages you can't read, local legislations you ignore, and more.

Agents are optimised for efficiency. Enterprises are optimised for resilience.

Also, a long tail of companies in B2B still do not even have proper digital infrastructure. Many do not have mobile apps. Some barely have functional websites. Others work with notebooks & pencils.

Agent adoption will concentrate in tech-savvy ecosystems, the rippling effects of which till take years to take over the entire market, if it ever happens.

I'll give you an example: I recently tried running OpenClaw in a secured environment. I have installed an instance of the miraculous lobster inside a Docker container on a Hetzner VPS. I made sure to make it super secure, so it couldn't have access to my files, apps or anything without my consent. I am running it with the least-permission principle, completely isolated, with controlled contexts and one finger always on the killswitch, in case something goes sideways.

The more secure I made it, the less useful it became. Loosen restrictions, and it becomes powerful but risky. For technical users, this is manageable but this ain't crossing the chasm anytime soon, as I see it.

So what does actually change? The interface layer.

The packaging of convenience becomes less defensible when agents can orchestrate outcomes across multiple systems with different interfaces. This is where the real riches are: make your agents skip what we don't need from SaaS and attack the outcomes. You don't need to replicate Hubspot: you have to replicate the outcome of using Hubspot, which is having an automated seamless customer relationship through marketing automation and sales insights. If you build something that does that, it won't look like Hubspot, because you can shave the "company debt" they have, but your version won't be of much use to other companies for exactly the same reason.

This is not the death of software. We will build more software to create more industries. We will create jobs that don't exist yet. We will find treasures in the sea of mediocrity because someone - or something - will figure out how to stand out. But we will figure it out.