Startup Corner #10: The future of SaaS in the age of AI Agents.
What happens when the app stops being the product?
Hey friends,
AI agents are already here.
We’re past the novelty phase.
Founders are deploying them for customer support.
Teams are using them to write code, handle follow-ups, schedule meetings, file expenses.
And increasingly, they’re not logging into the tools that used to do those things — they’re just delegating to an agent, and moving on.
What I’ve been thinking about lately is this:
If more and more of the work we used to do inside SaaS products is now being offloaded to autonomous systems — what does that mean for SaaS itself?
Not just the UI or the onboarding flow — I mean structurally.
Who is the user now? What are we monetizing? What’s the real product?
Because it’s starting to feel like the SaaS product isn’t what we think it is anymore.
Let me walk you through how I’m seeing it.
1. SaaS was built for user interaction — now it’s about agent execution
Classic SaaS is centered around you, the human in the loop.
You log in. You click through workflows. You move data between views, maybe build a dashboard or launch a sequence. It’s active. You’re the one doing the work — the software is just scaffolding.
That paradigm shaped everything:
Pricing = per seat
Value = measured by features used or time saved
Growth = dependent on user adoption and engagement
But now, we find ourselves using agents — from Copilot to ChatGPT to Lindy — and I’m not the one operating the product anymore.
I’m just issuing intent.
"Reply to this customer."
"Generate a proposal based on the meeting notes."
"Pull last quarter’s receipts and file reimbursements."
And the agent just… does it.
Without me ever opening the app.
2. The interface is disappearing — but the product still exists
This is the part that’s easy to miss.
Just because the user doesn’t see your app, doesn’t mean your app isn’t being used.
In fact, your product might be more valuable than ever — it’s just being called on by a system that isn’t human.
Which leads to the big shift I’m starting to believe:
The future of SaaS isn’t about better interfaces. It’s about being the best endpoint.
Not the prettiest UI. Not the richest dashboard.
The best execution node for a given task — callable, reliable, fast, context-aware.
That’s where I think the real battle is heading.
3. So what breaks — and what needs rethinking?
A few things start to fall apart when you shift your thinking from “user as operator” to “agent as executor.”
a. Seat-based pricing becomes irrelevant
If agents are triggering workflows, who’s the “user”? Who owns the seat?
It no longer makes sense to charge per headcount. Instead, pricing models will tilt toward:
Per-action or API call (Twilio-style)
Throughput tiers (like AWS or Snowflake)
Agent bundle deals (especially for all-in-one platforms)
You’ll charge for outcomes, not access.
b. User onboarding matters less — developer onboarding matters more
If agents are orchestrating the work, your UX isn’t the first touch anymore.
Your API docs are.
That means:
Your integration layer becomes your front door.
Your developer experience becomes a growth lever.
Your SDKs and plugins become marketing assets.
You’re building not for end users, but for the systems acting on their behalf.
c. Daily active users are the wrong metric
It sounds weird to say, but in this world, zero DAU might be a good thing — it means your product is fully automated.
So instead of DAU or MAU, you’re tracking:
Daily tasks executed
Successful completions
Latency and reliability
Agent retention
And success is less about stickiness, and more about how quietly and consistently you deliver value.
d. Distribution moves to the agent layer
This is one I think we’re just beginning to understand.
Today, you market to people. Tomorrow, you might need to market to agents — or the ecosystems that power them.
Think:
Getting listed in an agent plugin store (like OpenAI’s or Rewind’s)
Being integrated by AI orchestrators (like Cognosys or AutoGPT frameworks)
Optimizing your endpoint for LLM compatibility (structure, clarity, latency)
Distribution will depend on being chosen by the agent — not the human.
So what could be the new SaaS playbook?
If we accept that agents are the new interface, then a few things start to look like the new strategic pillars.
Build for composability.
Your product needs to be modular, callable, flexible — not a walled garden. Think Lego bricks, not Swiss Army knives.
Be the best at a narrow task.
In an agent-run stack, your function has to be the best available option for a specific job. Broad platforms lose to sharp, reliable execution.
Shift your GTM motion.
Your buyer isn’t always the person using the tool. Sometimes it’s the team architecting the agent workflows. Your messaging, positioning, and pricing need to reflect that.
Design for invisible value.
Your success might mean your app is never opened. That’s not a bug — that’s the future.
Final thought: we’re not building for people — we’re building for the people after the people
This is what I keep coming back to.
SaaS used to be about augmenting human input.
Now it’s about accelerating human intent — often by skipping the human entirely in execution.
That’s weird.
And exciting.
And uncomfortable. But it’s also real.
So when we look at the next wave of software, instead of asking, “How do I make this product delightful to use?”
Should we instead be asking:
“How do I make this product invisible — and still indispensable?”
Are we now building for that?
Until next time, RB