If you use Asana, Salesforce, or most other business SaaS products, you’ve probably had this thought before:

“Ugh … it’s so annoying to have to do it this way.”

Even when they’re really useful, SaaS products are a pain. When we install Monday or Zoom or Slack, we usually have to adjust our workflows to accommodate the way the software works – not the other way around.

Until now, we haven’t had much choice. SaaS products are entrenched in our workflows (and the only option on the market), and we’re so used to the status quo that we can’t imagine a different way.

But AI is changing that quickly – in fact, it’s reaching the point where it can outdo SaaS products at the same workflows.

I think in a few years, AI will be able to outdo SaaS at most workflows … and that’s bad news for every business that relies on expensive annual subscriptions for their business model.

Why SaaS rules right now

The main benefit of SaaS products is visibility. Software imposes consistency on process, so executives can evaluate what’s going on and track key metrics easily.

And most SaaS incumbents have cornered their markets, building reliance on the tech over time. QuickBooks owns the ledger, so they have a data moat no one can attack. Same for Salesforce – it’s too difficult to build a competitor, so everyone uses it even though it requires expensive consultants to manage.

But these tools come with significant rigidity. In order to make money in SaaS, solutions have to be generalized to a wide set of people. They get most users 80% of the way there out-of-the-box, and to get to the other 20%, you need upsells and pricey modifications.

So while these companies own the market share today, and enterprise leaders are clinging to their dashboards, eventually AI-native competitors who have agility to offer will edge them out.

How AI will disrupt SaaS

Even today, free or low-cost AI solutions are starting to pose threats to the justification for SaaS spend. And looking to the future, AI will change our reliance on these tools. Here’s how AI is disrupting Saas:

1. You can build the functionality yourself

A few weeks ago, I worked on a post about microapps, completely custom tools that are coded by AI from a simple text prompt. This includes apps like Bolt and Replit, which are essentially vibe-coding tools that let non-coders create fully functional, deployed apps.

The huge benefit of these tools is that they allow you to create solutions for specific, unique workflows for almost no money. Once these apps become mainstream, companies will be a lot less likely to subscribe to a SaaS tool that can only deliver a one-size-fits-all solution.

2. Node builders offer easier customization

Traditional SaaS tools historically don't play nicely together, meaning your workflows, data, and metrics are all siloed in systems that can’t talk to each other. AI-powered node builders like n8n, Pipedream, Relay, and Make are changing that.

They let you connect different tools and automate tasks across them, but with a key upgrade: AI adds intelligence to those workflows. So instead of just moving data, you can analyze, summarize, or act on it mid-flow.

Think of node builders as universal adapters – they unlock the data trapped inside SaaS tools and let you build custom, flexible systems that evolve with your needs. This essentially makes rigid SaaS interfaces outdated.

3. The semantic layer will change how we leverage software

There’s an emerging semantic layer that goes a level beyond nodes in terms of communication between legacy systems. It will allow you to describe how your business works to AI with words, and it will connect those descriptions to the data and tools underneath.

Think of it as a layer of business logic (context about customers, products, revenue, etc.) that sits above your tools and databases to define how everything relates to each other.

So if I worked in logistics, I could tell the AI, “We have 120 trucks that carry perishable goods along these routes.” The AI would use that understanding, plus access to my systems, to answer complex requests like “predict the percentage of our fleet that is likely to break down this year” without needing extra setup.

This type of AI knows how my business works, has access to all my tools, and can work across systems. But it isn’t available en masse yet – companies like Palantir are just beginning to roll this feature out to customers.

The phases of AI disruption

I’m under no delusion that the behemoths of SaaS are getting dethroned in the near future. There are quite a few steps between now and oblivion, but each of these stages will gradually erode the stronghold of legacy SaaS tools.

Stage 1: Layering of AI solutions

AI will be added on top of existing systems – not replacing them, but extending and enhancing what they already do. This step will actually make existing SaaS tools more useful and customizable without doing away with them completely.

This looks like:

• Using workflow automation tools to connect legacy systems
• Using new AI features in the SaaS tools to extend existing functionality
• Using no-code-built apps to fill holes in existing SaaS workflows

Stage 2: Replacement of legacy infrastructure

New AI-native businesses will begin rebuilding foundational systems from first principles, with AI at the core. These tools are not trying to plug into the old world; they’re reimagining how these processes will work entirely.

This looks like:

• A shift from AI extensions for existing systems to new AI-native competitors
• Tools that leverage business semantics over generic process UIs
• Less reliance on a singular ecosystem / proprietary tool in teams

Digits is a great example of an emerging AI alternative to Quickbooks. Digits doesn’t plug into the existing ledger (which has given Quickbooks an impenetrable data moat for years) because that ledger is not AI-compatible. Instead, they decided to redo the ledger completely. Now using Digits, I can actually chat with my books instead of my bookkeeper.

Stage 3: Agentic automation

Rather than clicking through interfaces or configuring settings, we’ll run processes via a series of autonomous AI agents that will handle tasks on our behalf – often proactively and invisibly.

This looks like:

• Chatting with an AI tool that “understands” your business logic
• Instructing the AI to execute on a process that it can do independently in the background
• AIs whose purpose is to achieve goals, not mimic human workflows

Stage 4: A full paradigm shift away from interfaces

Eventually, traditional software interfaces (like dashboards, dropdowns, and GUIs) will become obsolete – and that’s what we can’t wrap our heads around yet. This is a complete rethinking of how we work with AI, not just how we integrate with it.

This looks like:

• Expressing intent rather than executing a step-by-step workflow
• Language-driven processes vs. task-driven processes
• Apps that are not a destination you go to, but a gateway for your agent

I can imagine a future where you just sit on your couch and speak to your AI. No screen, no mouse – you won’t need them.

Obviously this is a long way out. We don’t even have real autonomous agents yet, just workflows using scattered bits of AI cognition. But it underpins the final boss of SaaS: a declining reliance on rigid processes and their associated interfaces.

What this means for SaaS leaders

I don’t think SaaS leaders need to worry about doomsday yet, but they do need to start thinking about how they meet this AI disruption. Because the reality is: Legacy SaaS tools will be replaced or go away.

If you’re a leader in SaaS, this is my advice to you:

1. Don’t ignore the little guys. New AI competitors seem harmless because they’re fundamentally different – but that’s exactly what makes them a threat. The longer you ignore these emerging companies, the more time they have to sneak up on you.

2. Don’t get complacent. SaaS companies love to drop “AI sprinkles” over their products to signal relevance. But the crux of your survival is in taking this seriously and finding ways to get AI into your offering, not just on top.

3. Don’t stay rigid. Obsolescence is knocking at your door, so you need to get flexible about the future of your product. Don’t dig your heels in on the way it’s always been done – it’s time to start imagining what AI transformation could look like.