Article: The death of Enterprise SaaS has been greatly exaggerated
Authored by Adam Seeber, Domain Lead – Enterprise AI & ServiceNow at xAmplify
There’s a neat narrative circulating that AI agents, vibe-coded applications and intent-driven automation will make enterprise SaaS irrelevant. The logic is simple: if an agent can interpret intent, call APIs and stitch together outcomes across systems, then the UI becomes less central and the value shifts toward orchestration and outcomes.
There’s truth in that. The interaction model is changing, and for a growing share of work, the front door won’t be a navigation tree. It will be a chat interface, an embedded agent in Teams or automation living inside the workflow.
But jumping from “the UI is changing” to “enterprise SaaS is dead” ignores what enterprise platforms actually are.
They’re not just screens. They’re governance, auditability, identity models, integration contracts, compliance artefacts, operating models and risk allocation frameworks. They’re systems designed to withstand scrutiny at scale.
I say this as someone who is broadly enthusiastic about AI-native development. I spend time experimenting with agentic tools and have built small applications that solve problems I would otherwise have paid for. The productivity uplift is real, but even with that tailwind, the structural reasons enterprise SaaS exists don’t evaporate overnight.
Market sentiment vs enterprise reality
It’s true that confidence around some large SaaS vendors has shifted. Financial markets are dramatically repricing risk in a world where AI changes how software is built and consumed.
But share prices are a proxy for confidence, not capability. The current volatility reflects uncertainty about where margins settle and how defensible certain workflows remain, not proof that enterprise platforms are obsolete.
In Australia, particularly in government and regulated sectors, replacement decisions are rarely driven by demo quality alone. They are first and foremost risk conversations.
Procurement functions optimise initially for accountability and continuity, then for cost. Buyers gravitate towards suppliers with stable operating models, long-term viability and the ability to stand behind remediation if something fails.
Where the “SaaS is dead” crowd is right
AI will disrupt parts of the SaaS market, especially where products are little more than UX over a narrow workflow.
If your defensibility is “we built the nicest screen for this niche task”, you’re exposed. AI-assisted development collapses the time-to-market advantage of those products. Customers now expect embedded intelligence and competitors can ship “good enough” alternatives faster.
Thin layers that exist mainly because an underlying system is hard to use are also vulnerable. If an agent can reliably mediate complexity via APIs, the UI advantage weakens. That disruption cuts across both small point solutions and enterprise vendors expanding into adjacent specialist domains without deep domain capability.
The value of many tools will shift from interface design to orchestrability, meaning how well they can be safely integrated into agent-driven workflows.
Enterprise platforms won’t vanish – they matter too much.
Replacing an enterprise system is rarely a feature comparison. It is fundamentally a governance exercise.
Enterprise platforms are deeply embedded across finance processes, HR rules, identity systems, approvals, reporting obligations and regulatory requirements. Over time, organisations build their operating models into these systems, which makes them far more than just software.
As a result, switching is not just a technical migration. It implies strong operational considerations. In the Australian public sector, that risk calculus is even more pronounced. Data residency, audit evidence, delegated authority models, separation of duties and incident response frameworks are baseline expectations – and they need to be demonstrable, not implied..
Large organisations carry obligations around identity, auditability, encryption, logging, monitoring and defensible evidence — no matter how the solution is delivered.. They need suppliers with insurance, contractual accountability and supportability at scale.
These constraints reflect risk posture and AI does not reduce these obligations. If anything, it increases the consequences when they are not properly managed.
AI Acceleration & the Governance equation
As agents begin acting across systems, the security model changes: delegated actions become part of normal operations and need to be designed and governed accordingly.
That creates a clear set of requirements: permissions must be inherited and constrained appropriately, actions must be logged and attributable, prompt-driven misuse must be mitigated, and control evidence must be easy to produce for auditors
Once an agent is allowed to take action, trust becomes central. Government and enterprise buyers understand this. They’re pursuing AI capability, but adoption decisions increasingly hinge on whether governance and control can be evidenced.
This is why mature enterprise vendors are increasingly pairing their AI messaging with control frameworks: augmentation plus guardrails.
Where the disruption will actually land
The disruption will not be evenly distributed. It is most immediate for narrow point solutions that have not embedded intelligence in a meaningful way.
Internal line-of-business tooling in low-regulation environments will move quickly. In organisations that reward speed and have build-versus-buy capability, AI-assisted development will expand internal build confidence.
In some enterprise accounts, the competitor won’t just be another SaaS vendor. It will be an internal team arguing that a defined capability can now be built in-house, faster and cheaper.
There’s also pressure on parts of the services ecosystem. Historically, some advisory value has been tied to navigating embedded complexity. AI will demystify certain platform behaviours and shift some value away from navigating complexity towards higher-order design and assurance.
In its place, services value will shift toward higher-order judgement, including operating model design, governance, architecture and outcome optimisation, rather than simply helping organisations work through the maze.
The Human Capability Factor
Giving someone a nail gun doesn’t make them a carpenter. The same applies to AI.
AI can dramatically accelerate execution, but it does not replace judgement. Whether building software, configuring platforms or redesigning processes, the real differentiator lies in how work is specified, validated and tested, and in how failure modes are anticipated and managed.
In enterprise environments, something that simply “works” is not enough. Systems need to scale, withstand audit, and remain stable and maintainable under pressure. That layer of judgement, knowing what to build, how to build it, and how it holds up over time, remains both critical and scarce.
What’ll change
For Australian enterprise and government buyers, a few practical shifts are likely to show up in procurement and delivery over the next couple of years.
- UIs will be de-emphasised as agents become interaction layers.
- Procurement processes will pressure-test incumbents more rigorously.
- Buyers will compare vendor options against internal build more frequently.
- AI capability will be assessed alongside governance maturity.
- Total lifecycle risk will carry more weight.
In regulated environments, the “enterprise tax” of controls and evidence remains non-negotiable. That tax is often the price of being trusted with critical processes.
That is why enterprise SaaS continues to play a central role. It provides stable data models, proven operating patterns, supportability and clear accountability when things go wrong.
Where it’ll land
Enterprise SaaS isn’t dying. It’s being forced to evolve.
Agentic coding and AI-native build approaches are accelerating that shift — something I’m strongly supportive of and actively experiment with. The disruption is real, but it lands unevenly depending on risk posture, system coupling and operating constraints.
The UI will be de-emphasised as agents become the interaction layer. Narrow point solutions without embedded intelligence will be pressured. Some incumbents will be unbundled, others will re-bundle, and margins will tighten.
Procurement will shift with it. Buyers will compare incumbents not only against competitors, but against internal build. In some enterprises, AI will strengthen the case to build defined capabilities in-house.
Enterprise SaaS exists for excellent, proven, structural reasons. In regulated and large-scale environments, it remains the most accountable way to run critical processes: governed data models, auditability, operational resilience and clear contractual liability. Switching costs, integration dependencies and the “enterprise tax” of controls and evidence still matter — and in many sectors, they are non-negotiable.
The outcome is not SaaS extinction, but evolution – a reshaping of where value sits. Those in the ecosystem who treat AI as a new interaction model and adapt their products and operating approaches accordingly will find opportunity. Those who assume yesterday’s defensibility carries forward unchanged will struggle.