Is SaaS Dead?

Six months ago, I couldn't write a line of code. Today I have a fully functioning, multi-tenant project management platform with 27 tools, AI built into the architecture, and live users.

One person. No dev team. No venture capital. Just thirty years of knowing what construction actually needs, and AI as a co-pilot.

If that's possible — and it demonstrably is, because you're reading the blog of the platform I built — then the traditional SaaS model has a serious problem. Maybe not today. Maybe not for everyone. But the foundations are cracking, and most people haven't looked down yet.

The Enterprise Trap

Most enterprises now spend somewhere around 8-10% of gross revenue on IT. In sectors like financial services and tech, it's higher. The vast majority of that is operating expenditure — subscriptions, licences, services. Layer upon layer of them, all stitched together to run operations that were never designed to work this way.

The obvious question: why don't they just build their own?

Because right now, that would be a nightmare. Their people are trained on existing systems. Their data lives there. Their processes are built around it. Ripping all of that out to build something bespoke isn't just expensive — it's the kind of project nobody wants to sign off on. I've seen contractors cling to a terrible spreadsheet system for years because the pain of switching felt worse than the pain of staying. It's the same dynamic, just at enterprise scale.

So they stay locked in. Paying the subscription tax. Living with "good enough." Using 30% of the features they're paying for and working around the other 70%.

Why AI Implementation Is Harder Than They Think

Everyone's talking about AI transformation. Most enterprises aren't anywhere near ready for it.

Their data is scattered across dozens of platforms that were never designed to talk to each other. CRM in one place, project management in another, finance somewhere else, documents in three different systems, correspondence in email. Sound familiar? It's the construction industry's spreadsheet problem scaled up to the entire business.

The current answer is automated workflows with an agentic layer bolted on top. Zapier-style connections trying to bridge gaps between systems that don't share a common language. It works, sort of. Like a site that runs on verbal instructions and hope — it functions until it doesn't, and when it breaks, nobody can trace what went wrong.

You can't build a mansion on a patchwork of foundations. And you can't build genuinely intelligent systems on data that's siloed across twenty different subscriptions.

Coding Blocks and Modular Architecture

Here's where it gets interesting.

As AI coding capabilities accelerate — and they're accelerating fast — we're heading towards something I think of as coding blocks. Modular, purpose-built components that can be assembled to match exactly what an organisation needs. Not a one-size-fits-all subscription. Not a bloated platform where you're paying for features you'll never touch. Architecture designed around how you actually work.

Think about the functionality most businesses genuinely need day to day: file attachments, comments, approval workflows, activity logs, status tracking, notifications, version control, distribution lists. These aren't unique to any one industry. They're universal patterns. But right now, every SaaS platform rebuilds them from scratch, locks them behind their own walls, and charges you a monthly fee for the privilege.

What if those patterns were standardised, composable blocks? Each one with a consistent structure — a service layer, a data layer, a user interface — that could be assembled to create exactly the system you need.

I'll give you a concrete example. When I built Construction AI, we needed a document approval workflow. Rather than building separate approval systems for submittals, then another for RAMS, then another for correspondence, we built the pattern once — a composable block with status tracking, distribution, comments, and audit trail — and used it across every module that needed it. Same block, different contexts. One fix improves everything.

Need attachments and approvals on your purchase orders but not comments? Configure it that way. Need the full suite on your project correspondence? Drop them all in. That's how construction works — you design for the site, for the client, for the purpose. You don't buy someone else's prefab and hope it fits.

Specialist AI Agents

This is where it goes from interesting to genuinely transformative.

Once you have block-based architecture, each block can have its own dedicated AI agent. Think of it like a site team. You've got your specialist trades — the sparky who knows every circuit in the building, the plumber who can trace a leak by sound alone. They know their domain inside out because they work in it every day.

Software blocks can work the same way. A specialist agent that knows its block completely — every bug it's encountered, every edge case, every decision made and why. When something needs fixing, the agent works on its block, tests the changes against every system that uses it, and only pushes updates once everything passes. Fix once, fix everywhere. No more hunting the same bug across fifteen different modules because someone copied and pasted code three years ago.

Scale that up and you get a site management structure for software. Specialist agents on individual blocks. A foreman agent coordinating groups of related blocks. A project manager agent overseeing the whole system. Software that gets more reliable over time because the agents accumulate knowledge, rather than the current model where every new developer has to learn the codebase from scratch.

That's not science fiction. The architecture for it already exists. I'm looking at it every day.

The Migration Problem

If modular architecture is the destination, then migration is the journey. And it's going to be enormous.

Think about what it actually takes to move an enterprise off its existing tech stack. Data in dozens of systems. Relationships between records that span multiple platforms. Organisational context that only exists in people's heads. I've watched contractors try to migrate from one project management tool to another and lose three months of productivity. That's one system. Imagine migrating twenty.

But imagine agentic workflows purpose-built to extract data from legacy systems — not a clunky CSV export and hope for the best, but intelligent agents that understand context, relationships between data, and organisational structure. Agents that can map a source system, identify the data model, extract everything cleanly, and restructure it into a unified architecture.

The companies that crack migration workflows will unlock one of the biggest opportunities in enterprise tech. It's the equivalent of the contractor who figures out how to do a live refurbishment without decanting the building — technically difficult, operationally complex, but extraordinarily valuable when you pull it off.

What True Agency Looks Like

Once an organisation has its data unified within a modular architecture with specialist agents maintaining each component — that's when genuinely autonomous systems become possible. Not chatbots answering FAQs. Not AI bolted onto broken foundations. Systems that understand the business and take action.

Your document agent spots that a critical specification has been superseded and alerts everyone who referenced it. Your financial agent identifies cost overruns against forecast before your QS has finished their morning coffee. Your approval workflow agent notices a pattern in rejected purchase orders — always the same supplier, always the same issue — and flags it before the next one lands.

That's the end state everyone's chasing. The problem is, most organisations are trying to get there by adding AI to a patchwork of subscriptions that don't talk to each other. You need the architecture first. The intelligence comes after.

What We've Done With Construction AI

I'm not writing this from theory. The composable blocks, the unified data layer, the AI woven into the architecture — that's what we've been building for the past year.

We mapped out the universal patterns before writing any code. Attachments, distribution, comments, activity logs, approvals, revisions, status tracking — all built as composable blocks with consistent architecture. The result is that new functionality which would take a traditional SaaS company months can be composed from existing blocks in days. A bug fix in one block benefits every module that uses it. And the AI layer isn't bolted on as an afterthought — it's built into the data model from the ground up, so it actually has the context it needs to be useful.

We did this with a construction manager and an AI. No dev team. No six-figure seed round. No legacy constraints. That's either encouraging or terrifying, depending on which side of the SaaS subscription you're sitting on.

What This Means If You're Running a Small Firm

The grand vision is interesting, but what do you actually do with it today?

Pay attention to where your data lives. Every system you subscribe to is a silo. Every silo makes future AI implementation harder. When you're choosing tools, think about whether the platform owns your data or whether you can get it out cleanly. This matters more than features right now.

Don't over-invest in platforms that lock you in. Long-term contracts with enterprise SaaS vendors might feel safe, but the landscape is shifting fast. The more dependent you are on a single platform's way of doing things, the harder it is to move when something better comes along.

Start thinking about what you actually need versus what you're paying for. Most small contractors use a fraction of their software's capability. If you're paying for a platform with 200 features and using 15, you're subsidising complexity that doesn't serve you.

Look for platforms built with this future in mind. Unified data, modular architecture, AI that understands your domain — not AI bolted onto a ten-year-old codebase. The difference will become obvious over the next two years.

Where This Is Heading

The subscription model isn't going to disappear overnight. Enterprises are too locked in, the switching costs are too high, and the migration tooling isn't mature enough yet. But the direction is clear. Software is moving towards modular, composable architecture where organisations own their systems rather than renting someone else's vision of how their business should run.

The firms and platforms that understand this — that build for composability, that design around how people actually work, that treat AI as architecture rather than a feature — will have a significant advantage. The ones clinging to the old model of bloated platforms and per-seat licensing will find themselves increasingly difficult to justify.

I've spent thirty years watching construction resist change until the economics made it unavoidable. Software is approaching that same tipping point. The economics of building are changing faster than the economics of subscribing.

SaaS isn't dead yet. But it's on notice.

---

Steve McKenna is a Chartered Construction Manager and founder of Construction AI — a purpose-built, AI-native project management platform for the UK construction industry.