There has been plenty of debate about whether generative AI is the death knell for enterprise SaaS, but most people are framing it at the wrong level.
One side argues that AI will make SaaS obsolete because software will become cheaper to build. The other? They’re utterly convinced that SaaS will survive because enterprises still need reliable systems of record. They both have a point, but the reality is far too complicated to wrap-up in a clickbait headline.
Enterprise software is not a single uniform category. It contains different kinds of logic, with varying ownership responsibilities, failure modes, and consequences when things go wrong.
The more precise question isn’t if AI will replace SaaS. It’s which parts of enterprise software become cheaper, easier to adapt and reconfigure on demand and more replaceable, and which others remain intrinsically complex, high-risk and institutionally important.
The vital distinction is between shallow logic and deep logic.
Shallow logic is the surface and coordination layer of software: user interfaces, forms, dashboards, task routing, workflow orchestration, notifications, approvals, configurable rules, and common cross-system coordination. Don’t let the name fool you, this layer still requires craft and judgement. It simply usually sits closer to how work is presented, sequenced, and adapted, rather than where the highest consequence obligations are owned.
Deep logic on the other hand is the high consequence logic embedded in the operational reality of the enterprise. Your regulated calculations, entitlement models, financial controls, data integrity constraints, exception handling, audit-sensitive decisions, security boundaries, privacy obligations, reconciliation logic, not to mention years of institutional edge cases. It is where software becomes part of your organisation’s control environment.
The future of enterprise SaaS hinges on this contrast.
AI is likely to continue compressing shallow logic significantly. It may also assist with deep logic, but it does not make deep logic easy to own, safe to change, or simple to govern.
The dividing line is not always obvious. It’s not about how complicated something looks; a screen can appear complex while the underlying logic is relatively shallow. Whereas a rule change can look inconsequential while carrying material legal, financial, safety, or regulatory consequences.
The question is: What happens if this is wrong, and who must prove that it was right? That is why regulated and mission-critical domains matter so much.
The "so what" for enterprise SaaS is that defensible value is pushed further into the stack. AI may compress shallow logic, but deep logic must still be correct, evidenced, and changed safely.
AI does not have to replace SaaS to weaken it
We don’t subscribe to the common “SaaS apocalypse” theory, that claims every company on earth will use AI to vibe code its own finance system, HR platform, CRM, or compliance system from scratch. That argument misses the point by confusing the cost of generating software with the burden of operating it responsibly.
There is a however a much more credible version of the “SaaS apocalypse”. One in which AI continues to erode much of the value currently captured by traditional SaaS interfaces and shallow workflow layers.
For years, many enterprise users have been forced through fixed screens, rigid workflows, duplicated data entry, awkward reports, and fragmented cross-system processes. In other words, the user has adapted to the system rather than the system adapting to the work.
AI challenges that arrangement. It may weaken the coupling between the system of record and the system of interaction. If a user can describe an outcome, have relevant context assembled, receive a recommendation, trigger the right workflow, update multiple systems, and importantly preserve the evidence trail, then the traditional application interface becomes less central. The system underneath may still matter, but the user may no longer experience it as the primary place where work happens.
That is the real threat to SaaS vendors whose value sits mainly in shallow logic: screens, forms, workflow convenience, reporting surfaces, and administrative coordination. But that doesn’t make all enterprise SaaS obsolete. The systems that calculate, reconcile, authorise, evidence, secure, and preserve critical enterprise data still carry a burden that cannot be dismissed as “just software.”
Shallow logic is undeniably exposed to AI, but the likely consequence isn’t that SaaS ceases to exist. It is that customers become less tolerant of paying premium prices for surface complexity. Poor UX, slow implementation, expensive configuration, static dashboards, and rigid workflows will be harder to defend.
If a product’s value is mostly shallow logic wrapped around a database, AI will challenge that position.
Deep logic carries institutional responsibility
While from a distance deep logic can look like a basic set of business rules, that is far too simplistic. Deep logic does include rules and calculations, but also the data model, control environment, permissions, audit evidence, integration contracts, operational constraints, regulatory interpretation, and the vital accumulated knowledge of what happens when reality refuses to conform to the happy path.
Deep logic is where software becomes institutional machinery. You can’t assess replacement feasibility based on whether AI can generate something that appears to perform the same visible function. That’s paddling at the shallow end when the real risk is deep.
AI does reduce the cost of creating software. Smaller teams can prototype faster. Domain experts can express requirements more directly. It can generate code, tests, documentation, data mappings, integrations, and user interfaces. It may also compress implementation work that previously required larger teams.
None of that proves that the long-term ownership cost of deep logic reduces to the same degree. The hard part of deep logic isn’t writing it down. It is knowing whether it’s correct, proving that it meets compliance obligations, changing it safely, operating it under pressure, and defending it when challenged.
Once an enterprise relies on deep logic, it owns the consequences. AI can assist with tasks, but it could never remove the need for institutional accountability.
SaaS is strongest where deep logic is shared
If a process is mission-critical and highly specific, individual enterprises may be motivated to invest in owning more of it. If the reverse is true and it’s low-risk and shallow, AI-enabled tools may make customisation or replacement attractive.
The case for SaaS remains strong when a process is mission-critical, high consequence, regulated, data-rich, and broadly similar across many organisations. Here, the value is not simply in the vendor hosting the software, it is in the vendor carrying a specialised burden across many customers.
Customers avoid owning a difficult non-differentiating burden, and the vendor earns the right to charge. AI does not invalidate that bargain, but it will challenge vendors to deliver it efficiently.
Data alone is a weaker moat
A common defence of incumbent SaaS is that it holds valuable data. There’s merit to that argument, especially if it has been accumulated over many years and is difficult to migrate. But possession of data is not the same as durable strategic control.
Data is powerful when it is connected to deep logic: within a system that understands what the data means, which rules govern it, which permissions apply, which decisions depend on it, which exceptions matter, and what evidence must be retained. In that setting, the vendor isn’t merely storing records. They're helping to preserve the operational meaning, integrity, and accountability of those records.
As AI enters the enterprise, it needs more than data access. It needs trusted context, business semantics, authority boundaries, and evidence trails. A system that simply stores data will be bypassed by a better interaction or orchestration layer. A system that governs high consequence data in context is harder to displace, because the value sits not just in the records, but in the disciplined interpretation and use of those records.
Data is not an automatic moat. Vendors still need to prove that their system leads to better decisions, safer actions, lower burden, or faster outcomes for customers.
The conclusion is not that SaaS dies
AI will not eliminate enterprise SaaS. But it’s likely to remove a large amount of weak SaaS value.
It will challenge the assumptions of vendors who have not evolved with the technology. Those that assume that users will continue to tolerate bad interfaces because the system is embedded, that data custody alone is a moat and that shallow workflow software somehow deserves premium pricing simply because it is inconvenient to replace.
AI compresses shallow logic, increases the importance of well-governed deep logic, and shifts value toward systems that combine trusted data, accountable decisions, and safe action. Within shallow logic, speed and adoption are sure to dominate. Customers will expect rapid adaptation and low friction. For deep logic, trust remains non-negotiable. Speed matters, but not at the expense of correctness, accountability, or auditability.
The future is not SaaS versus AI. The future is shallow logic becoming fluid, deep logic becoming more valuable, and every vendor being forced to deal with this boundary.
Disclaimer: Our industry moves quickly and what is true today may not be next week, let alone next year. We expect our leaders to think boldly and take clear positions, but we also want them to adapt to new information, shift their thinking and not be afraid to change their minds as technology evolves.