_A note on the 80%: I don’t mean 80% of SaaS companies will shut down. I mean roughly 80% of the value SaaS captures today, the revenue, the margins, the market cap, will shift away from the UI and logic layers and toward data nodes and the agents on top of them. Some companies will survive by becoming nodes. Many will shrink. Some will disappear.
And to be clear, 80% isn’t a precise calculation. It’s a rough estimate based on what I see in my own work, in our customers’ teams at Aissist, and in where the market is already moving. I don’t have a crystal ball. But I’d rather make a clear bet and be roughly right than wait for certainty and be late._
A few weeks ago I did something I hadn’t done in years. I looked honestly at where my working hours go.
I’m the founder of Aissist, an agentic AI platform for customer service and sales. Like every founder, I’m a heavy software user. I pay for more than 20 SaaS tools: marketing automation, CRM, ads, email, social media, outreach, AWS, Azure, accounting, meeting booking, code repositories, dev tooling, and more.
The result surprised me. Over the past two months, I spent more than 30 times as much time in Agents (including the ones made by us) as in all of those tools combined. Some of them I haven’t logged into in weeks, even though they run every day.
I’m not using them less. I’m using them differently. And once I saw that clearly, I couldn’t unsee it.
I can’t stand the UI anymore
Here’s an uncomfortable confession: I’ve lost my patience for most SaaS interfaces.
Twelve clicks to build a report. Nested menus no one can remember. Dashboards that take eight seconds to load and still don’t answer my question. Every tool has its own idea of where “Settings” lives.
For a decade we accepted this. Learning the UI was the price of using the product. Companies even hired “admins” whose whole job was knowing where the buttons were.
That price now looks absurd. When I can describe what I want in a sentence and an AI agent does it across three systems in under a minute, a messy UI isn’t just annoying. It’s a tax.
I’m clearly not alone. Asana’s research finds workers switch between about 9 apps every day. Meanwhile, ChatGPT has reached 900 million weekly active users, more than double the 400 million reported in February 2025. People are voting with their time, and they’re choosing the conversation over the dashboard.
So why am I still paying for SaaS?
If I barely open these tools, why haven’t I canceled them?
To answer that, I broke SaaS down into what it actually is. Almost every SaaS product is three layers stacked together:
The UI. Screens, buttons, dashboards, workflows built for humans to click.
The logic. The rules: how leads get scored, when an invoice is flagged, which emails go out in what sequence.
The data and infrastructure. The records, the history, and the machinery that does real work in the real world.
Then I asked: which layer am I actually paying for?
The UI? No. I’ve stopped using it. My agents talk to these tools through APIs.
There’s one exception, and I think it’s telling: payments.
Anything that moves money, like our payment processor, banking, payroll, and paying invoices, I still handle myself, inside the vendor’s own interface. I’m happy to let AI draft, analyze, and reconcile. But I still hold back from letting it press “send” on money.
It’s not because AI can’t do it. It’s because a mistake there can’t be undone, and the trust layer around payments (authentication, fraud controls, approvals, audit trails) was built for humans, not agents. We think about this the same way at Aissist: an action that can’t be taken back is the highest-severity error class in our reliability standard, held to under 0.01%.
But notice what that exception says. The one category of SaaS I still log into is the one whose value is almost entirely trust and real-world rails. That’s exactly the kind of data node that survives.
The logic? Increasingly, no. I’d rather define my own logic in plain language and let an AI execute it than adapt my business to a vendor’s assumptions. A rigid “lead scoring” feature built for the average customer is worth less to me every month.
The data and infrastructure? Yes. This is the only part I’d fight to keep. And when you look closely, it’s more than just a database:
System of record. Years of customer history, transactions, and contracts.
Data that’s trusted, auditable, and hard to rebuild.
Real-world rails. An email platform’s real value isn’t the editor; it’s sender reputation, warmed IPs, and deliverability. An ads platform owns the auction. A payment provider owns the banking connections. A cloud provider owns the compute, storage, and uptime.Compliance and trust. SOC 2, ISO 27001, GDPR, financial regulations. Things a model can’t generate on the fly.
Network effects and proprietary data. Contact databases, marketplace liquidity, partner integrations built over years.
That’s the real product now. Everything above it, the UI and much of the logic, is becoming optional.
Satya Nadella said something similar back in December 2024. He described business apps as “essentially CRUD databases with a bunch of business logic”, and argued that it is the business logic layer that AI agents will take over. At the time, many people dismissed it. Not anymore.
I see it every day at Aissist
This isn’t just a theory for me. It’s what I watch happen inside our customers’ support teams.
Most of our customers run their support on a helpdesk: Zendesk, Intercom, Freshdesk, Salesforce, HubSpot, Gorgias, and others. When they deploy Aissist, something interesting happens to that helpdesk.
The AI agent resolves 83% of conversations on average, and up to 98% in mature deployments. It reads from and writes to the helpdesk, the CRM, the order system, and billing, and keeps them all updated, so no one has to touch them manually. The human team stops living inside the helpdesk UI and focuses on the conversations that genuinely need a person.
The helpdesk doesn’t disappear. It becomes exactly what I described above: a data node. The system of record. The ticket history. The audit trail. The UI and the routing logic matter far less, because the intelligence now lives in a layer above it.
The same goes for reporting. With Aissist Pulse, a support leader doesn’t build a dashboard to find out why customers are contacting them. They just ask, and get the answer in real time. No twelve clicks.
The market has already noticed
Wall Street has been pricing this in all year.
In the first week of February 2026, software stocks lost over $1 trillion in market capitalization on fears that agents would displace traditional SaaS workflows. The rolling correction traders now call the “SaaSpocalypse” has wiped roughly $2 trillion from software stocks since early 2026. Salesforce alone lost 40.9% of its market value in the first half of 2026.
The core fear is simple: AI agents threaten the per-seat licensing model that has underpinned SaaS growth for a decade. If fewer humans log in, fewer seats get sold.
And there was already plenty of fat to cut. The average company manages 305 SaaS applications, and 36% of licenses sit unused, costing about $19.8M a year in waste.
Am I an outlier?
Maybe today. I’ve been working with AI every day for over two years, so I’m probably early.
But I’ve seen this movie before. Early adopters of cloud, mobile, and the web looked like outliers too. Then the tools got easier and everyone followed. Every month, AI agents get better at operating software through APIs, and every month more people discover what I discovered: you don’t need to log in to get the work done.
Even conservative analysts are moving in my direction. Gartner predicts 35% of point-product SaaS tools will be replaced by AI agents or absorbed into larger agent ecosystems by 2030. I think that’s the starting point, not the ceiling.
Why 80%? Because most SaaS companies are mostly UI and logic, wrapped around data that’s either thin or easy to move. Those are the ones in trouble. The 20% that survive will own something hard: deep proprietary data, real-world infrastructure, regulatory trust, or network effects.
What replaces SaaS?
Not one giant AI app that does everything. Something more interesting.
I believe we’ll see a network of data nodes, each with a specialized service.
Each node does one thing extremely well: holds a trusted dataset, or runs a piece of real-world infrastructure. It exposes clean APIs. It’s secure, reliable, and compliant. It doesn’t need a beautiful UI, because humans aren’t its main users. Agents are.
On top sits the intelligence layer: AI agents that understand what you want, pull from the right nodes, apply your logic, and get the work done. The “application” is assembled on demand, for you, in the moment.
That’s the architecture we’re building toward at Aissist. Our multi-agent orchestration, AgentMesh, sits on top of the systems our customers already own, and our Connectors let it plug into almost any data node they rely on. We don’t want to be another dashboard you have to log into. We want to be the layer that makes the dashboards unnecessary.
That’s all we really need in this era.
This isn’t only about enterprise
The same shift is coming for consumers.
Think about how many apps sit on your phone just so you can book a flight, check a bank balance, order food, or compare prices. Each one has its own UI, its own login, its own maze of menus.
Now imagine telling your assistant: “Get me to New York next Thursday morning, aisle seat, under $400, and put it on my calendar.” The airline becomes a data and inventory node. Your bank becomes a payments node. Your calendar becomes a scheduling node. You never open any of the apps.
Consumer apps that are just a pretty wrapper around someone else’s data will face the same squeeze as enterprise SaaS.
What this means if you build software
If you’re building a SaaS company today, ask yourself one hard question:
If no human ever logged into my product again, would customers still pay me?
If the answer is yes, because you own critical data, real infrastructure, or real trust, you’re building a node. You’ll be fine. Probably better than fine.
If the answer is no, the clock is already ticking.
I’m betting Aissist on this shift. The future of software isn’t more screens. It’s fewer screens, better data, and agents that do the work.
What about you? How much time did you spend inside your SaaS tools this week, versus talking to an AI?
Top comments (0)