When AI Comes for the Most Defensible SaaS: What Founders Can Learn From the ERP Disruption

ERP was the category everyone agreed AI could not touch. Too entrenched. Too painful to replace. Hundreds of billions of dollars sitting behind the highest switching costs in software.

Then on September 1, coverage of an AI-native startup called Rillet made the rounds. The framing: AI vs legacy ERP, with an AI-native startup called Rillet going after Oracle, SAP and Workday.

If AI challengers are viable in ERP, no SaaS category gets to call itself safe. Including yours.

Here is what actually happened, and what it means for how you build and sell.

AI vs legacy ERP: what actually happened

Rillet’s CEO Nicolas Kopp went on CNBC alongside investor Roelof Botha, formerly of Sequoia. The pitch is simple. Legacy ERP runs on batch processing. Numbers get calculated monthly or quarterly. A CFO looking at revenue is looking at a period that already closed.

Rillet rebuilt the whole thing around AI agents that run continuously. Financials update in real time instead of every five to seven days. The full coverage is here.

Botha put the switching problem plainly: “changing your ERP or accounting system is like open-heart surgery.” That surgery is exactly why Oracle, SAP and Workday survived every previous technology wave. Nobody wanted the pain.

His estimate now: it would take rivals four to five years to catch up to a genuinely rebuilt architecture.

One startup does not kill three giants. That is not the point. The point is that the market now believes the most defensible category in SaaS is attackable. That belief changes buyer behavior everywhere.

The moat was pain, not love

Nobody stayed with legacy ERP because they loved it. They stayed because leaving hurt too much.

That is a real moat. It compounds for years. But it has a failure mode: it works only until someone builds a product where the gap in value is bigger than the pain of switching.

Read that again as a founder. If your retention is driven by lock-in, data gravity, or migration fear, you do not have loyal customers. You have trapped ones. Trapped customers churn all at once, the moment a credible exit appears.

The question to ask this week: if your product launched today, built from scratch with current tools, would it look like what you are selling? If the honest answer is no, someone is already building that version.

Batch vs real-time is the whole pitch

Notice what Rillet’s wedge is. Not “AI-powered.” Not “smarter ERP.” The wedge is one concrete outcome: your numbers are live instead of a week old.

That is a pitch a CFO understands in one sentence. No demo needed to grasp the value.

Every category has a batch process hiding in it. Reports that run weekly. Syncs that run nightly. Reviews that happen quarterly. Anywhere your buyer waits for information that could be continuous, there is a wedge.

Find the batch process in your category before someone else does. Then make it the entire pitch.

AI features will not save you

The second piece worth reading from the same day: a HousingWire column arguing that agentic AI is turning SaaS interfaces into invisible infrastructure. Users stop navigating screens. They state an outcome. The system executes.

The author, Blue Sage CTO Steve Octaviano, draws a line between AI embedded in the workflow and AI layered on top. The second kind he calls “AI theater.” Impressive in a demo. Disconnected from daily operations.

Most SaaS companies are currently shipping theater. A chatbot in the corner. A summarize button. An “AI” label on the pricing page. Buyers went through 2025. They can tell.

The uncomfortable test: if you removed your AI features tomorrow, would your product’s core value change? If not, the AI is decoration. Decoration does not defend a category.

The interface is losing value. The infrastructure is not

Here is the part most founders miss in the doom headlines. SaaS is not dying. The interface is.

When an AI layer sits between the user and your product, your beautiful dashboard matters less. What matters is what the agent executes against: structured data, clean workflows, reliable APIs, auditability. If your underlying system is fragmented and held together with manual workarounds, AI does not improve it. It exposes it.

That flips the roadmap conversation. The next moat is not a better screen. It is being the system of record that agents trust and build on. Boring, structural work. It compounds.

What to do about it this quarter

Not someday. This quarter.

Audit your moat honestly. Write down why customers stay. Separate “they love the outcome” from “leaving is painful.” The second list is your exposure.

Find your batch process. List every place your buyer waits for information in your category. Pick the one that hurts the most. That is either your next feature or your next competitor’s wedge.

Kill the theater. Cut any AI feature that does not change a core workflow. Ship one that does, even if small. One real agentic workflow beats five demo tricks.

Invest under the hood. Structured data, documented APIs, clean event logs. Unsexy. Also the price of admission for the agentic era.

Update your positioning. If an AI-native competitor entered your category tomorrow, what would your one-line defense be? If you do not have one, that is the strategy work to do now, not after they launch.

The ERP story is not about accounting software. It is a preview. The categories that felt safest are the ones where the gap between old architecture and new is widest. Check where your category sits on that list.

Common questions