The SaaS moat is moving below the interface.

The horizontal-versus-vertical SaaS debate is useful, but I think it misses the more important divide. The real question is whether a product is a trusted system of record or primarily a layer of workflow around someone else’s data.

AI will not affect every software company in the same way. It is likely to compress the parts that are easiest to describe, generate and change—screens, approvals, task routing and narrow internal workflows—before it replaces the systems an enterprise relies on as its source of truth.

A system of record is more than its code

A CRM, general ledger or human-capital system carries years of operating history. Its value includes permissions, compliance, integrations, data quality, edge cases and the organizational confidence that the record will still be correct tomorrow.

That makes replacing it a business-risk decision, not simply a build-versus-buy calculation. Generating a credible interface is becoming easier. Recreating the accumulated reliability behind it is not.

This is why I would not assume that horizontal SaaS is safe merely because it serves many industries. The stronger protection comes from being deeply embedded, authoritative and dependable.

Vertical software needs a deeper answer

Vertical SaaS is more exposed when its main value is packaging a repeatable workflow for one industry. If a customer can build and adapt that workflow with AI, a per-seat product made up mainly of screens and process steps will face pressure.

But vertical does not mean doomed. Domain-specific software can remain durable when it owns authoritative data, encodes difficult rules, connects to regulated or fragmented ecosystems, or takes responsibility for a measurable outcome.

The useful test is simple: if the interface disappeared, would the customer still depend on the product’s data, decisions, network or operational capability?

Design for agents, not only users

As AI agents become the primary interface across multiple applications, the SaaS product may be less visible while becoming more important underneath.

The product work shifts toward clean APIs, explicit permissions, audit trails, reliable retrieval and safe write-back. An agent should be able to use the important capabilities of the system without weakening the controls that made it trustworthy.

My view is that SaaS is not disappearing. Its visible surface is being challenged. Workflows will become easier to create and replace; trusted records and hard operational capabilities will remain difficult to earn.

The next software leaders may win by becoming less visible—and more indispensable.