SaaS Apocalypse

The UI isn’t the product anymore. Agents pull value away from dashboards and seats toward context, workflow, and execution. What to build in-house, and what to buy.

Bart van der Meeren

February 26, 2026

9 min read

AI Strategy#AI#SaaS#Agents#Automation#Workflow#OpenSource

The UI isn’t the product anymore. Agents pull value away from dashboards and seats toward context, workflow, and execution. For builders, the rule is simple: buy systems of record, build systems of action.

Who this is for

By “builder” I mean anyone responsible for building business software and workflows:

SaaS founders and product teams shipping software
In-house builders (IT, data, RevOps, operations) building internal tools and automations
Agencies and freelancers building this for clients

This is not about developers in general. It’s about the people shaping how work gets done.

Why I’m calling it a SaaS Apocalypse (without the doom)

I don’t think SaaS is disappearing.

I do think a large part of the SaaS model as we know it is getting squeezed:

UI as the main differentiator
seat-based pricing as the default
tools that guide work instead of executing it

Agents change the entry point.

When the user experience shifts from “open app, click dashboard, export CSV” to “do this for me” and an agent executes, value moves away from the interface.

Interfaces will still exist, but more as:

oversight
exceptions
approvals
audit

Not daily click work.

System of Record vs System of Action

Here’s the question that makes a lot of decisions easier:

Are you building a system of record, or a system of action?

System of Record

Source of truth. Data integrity, audit trails, compliance, correctness.

Think:

accounting
bank connections
payroll
ERP

Mistakes are expensive. Maintenance is permanent. Integrations break. Regulation changes.

System of Action

The layer where work gets executed. The layer that decides:

what needs to happen
when
by whom
with which exceptions

Think:

lead routing and follow-up
revenue workflows
campaign logic
support triage
approvals and escalations

This is where agents actually move the needle.

The rule (explicit): build in-house vs buy

Buy systems of record. Build systems of action.

More explicitly:

Build in-house what’s unique to your company or niche: decision logic, context, exceptions, guardrails, orchestration.
Buy generic what’s commodity, or what is mostly risk and long-term maintenance: source-of-truth layers with compliance, audit trails, and fragile integrations.

Practically:

Start with a commodity base (open source if it fits)
Build your company-specific insights and actions on top

That’s how you iterate fast on the action layer without inheriting years of maintenance on a record layer where errors hurt.

What you usually shouldn’t build in-house

1) An accounting system with bank connections

No.

Not because it’s impossible, but because you’re signing up for compliance, security, broken integrations, exception handling, and liability.

Buy it. Integrate around it. Focus on the layer that executes actions.

2) Anything that’s mostly UI + standard CRUD

If your value is a UI on top of standard objects and flows, your moat is thin.

Not because agents are perfect, but because they’re often good enough to remove most of the screen work.

3) Generic horizontal tools without a wedge

Another project management tool. Another CRM for everyone. Another analytics dashboard.

You don’t win there with features. You win with distribution or extreme focus.

What you should build in-house

1) An action layer on top of a basic record layer

CRM is a simple example.

A CRM as a system of record can be basic:

accounts
contacts
deals
activities

Done.

What you usually actually want is:

insights that fit your business
actions that fit your process

Examples:

qualification logic: what counts as a good lead in your market?
next-best-action: what should happen now?
routing: who owns this, and when?
guardrails: when does a human need to review?
orchestration: executing actions across multiple tools

That’s a system of action.

2) Context you can explain

Agents only work well when context is explicit.

Context isn’t “more data.” Context is:

definitions
exceptions
permissions
tone of voice
criteria

If you build this well, it’s harder to replace than an interface.

3) Agent-first product thinking

Build as if your primary user is an agent.

That means:

solid APIs
clear schemas
idempotent actions
logging and audit trails
non-mysterious error handling

And yes, you’ll still need UI. But UI becomes the place for exceptions, approvals, and trust. Not for daily navigation.

A quick checklist for the next build vs buy debate

Is this a system of record with audit/compliance at the core? If yes, buy.
Are the main risks financial or legal? If yes, buy or narrow scope.
Is your differentiation mostly UI? Then you’re vulnerable.
Can an agent replace your UI via APIs without the user feeling pain? Then you’re vulnerable.
Is your advantage in context, workflow, and exceptions? Then you’re on the right track.
Can you measure outcomes and price on work done? Then you can price sustainably.
Can you run a real pilot in 2 weeks with real data and integrations? If not, scope is too big.
Are you willing to maintain integrations for years? If not, don’t build the wrong layer.
Can you start from a commodity/open source base and differentiate above it? If yes, do that.

Conclusion

The SaaS Apocalypse isn’t software disappearing.

It’s value moving:

away from dashboards and seats
toward context, workflow, execution, and outcomes

Accounting with bank connections? Don’t build it in-house.

CRM as a source of truth? Basic is fine. Buy generic, or start from open source.

But your action layer, your insights, and your execution? Build that in-house.

Want to sanity-check this?

Not sure what you should build in-house vs buy?

Call me. We’ll pick one workflow and make the build vs buy decision fast.