There is a quiet disruption happening inside the software your organisation pays for every month. It is not dramatic — no servers going dark, no sudden migrations — but it is accelerating, and the organisations that recognise it early will have a significant advantage over those that do not. The rise of AI-native products like Cursor, Perplexity, and Cognition is doing more than introducing new competitors to familiar markets. It is exposing a structural vulnerability that most traditional SaaS vendors have not yet found a credible answer to.
The vulnerability is this: the entire design logic of conventional SaaS platforms is built around helping humans manage work. Dashboards, menus, form fields, drill-down reports — every element of the interface exists to present information to a person, so that person can decide what to do next. But when an AI agent can simply do the work — write the code, find the answer, complete the task — the interface that was designed to assist human judgement becomes, at best, an overhead, and at worst, an obstacle. The question for senior decision-makers is no longer just which tools their teams prefer. It is whether the platforms they are investing in are built for a world that is already changing around them.
The Interface Was Always a Compromise
It is worth being precise about what traditional SaaS interfaces actually are. They are workarounds. Because software cannot read minds, vendors build graphical interfaces — menus, wizards, dashboards — to expose their system's capabilities in a form that humans can navigate. The better the interface, the lower the friction. But friction is never zero. Every click, every manual data entry, every navigation step is a tax on the human doing the work. UX design has spent two decades trying to reduce that tax. AI-native products are proposing something more radical: eliminate it entirely.
Cursor, the AI-powered coding environment, does not simply improve the experience of writing code — it changes the fundamental activity. Developers describe and direct; the system executes. Perplexity does not give users a better search interface — it removes the need to evaluate and synthesise dozens of results by returning a direct, cited answer. Cognition's Devin does not help engineers navigate a project management tool — it takes on engineering tasks autonomously. These are not incremental improvements to existing workflows. They are substitutions. The human is no longer the primary executor; they are the director.
Where Traditional SaaS Platforms Are Most Exposed
Not every SaaS category faces the same level of disruption, and it is important to think clearly about where the real exposure lies. Platforms that derive their value primarily from the interface — from making complex data navigable, or from structuring human workflows into trackable steps — are most at risk. CRM systems that rely on sales teams manually logging activities, project management tools that depend on team members updating task statuses, business intelligence platforms that require analysts to build and interpret dashboards: all of these are built on the assumption that a human is doing the navigating. If an AI agent can log the CRM activity, update the project status, and surface the key insight without anyone touching the interface, then the platform's primary value proposition weakens considerably.
The more insidious risk is vendor lock-in working in reverse. Organisations have historically stayed on legacy platforms partly because switching costs are high — data is embedded, processes are built around the tool, staff are trained. But if AI agents increasingly interact with SaaS platforms via APIs rather than interfaces, the switching cost of the interface itself drops to near zero. What matters then is the quality of the underlying data model and the API surface. Vendors who have not invested seriously in their API layer — or who have deliberately kept it limited to protect their interface moat — may find that moat has become a liability rather than an asset.
The Vendors Who Are Responding — and How
The established SaaS players are not standing still, but their responses reveal a tension that is difficult to resolve from the inside. Salesforce has invested heavily in its Agentforce platform, attempting to layer AI agency on top of its existing architecture. Microsoft is threading Copilot throughout the Microsoft 365 suite. Atlassian, HubSpot, and ServiceNow have all announced AI features at pace. The challenge for all of them is the same: they are trying to retrofit agentic capability onto products whose core architecture was designed for human-driven interaction. That is a harder engineering problem than building for agents from the ground up, and it often produces results that feel bolted on rather than native.
The AI-native challengers do not carry that legacy weight. Cursor was not adapted from an existing IDE — it was conceived with AI as the primary actor. That architectural difference matters enormously at the level of performance, reliability, and the kinds of workflows the product can support. For UK organisations evaluating their software stack, the practical question is whether a vendor's AI roadmap represents genuine architectural investment or a rebranding of existing features with a language model on top. The distinction is usually visible if you look at the API documentation, the data model flexibility, and whether the AI features work across the whole product or only in isolated modules.
What This Means for Enterprise Software Procurement
The implications for how organisations buy and govern software are significant. Procurement processes that evaluate SaaS platforms primarily on interface quality, user experience scores, or feature checklists will increasingly miss what matters. The more relevant evaluation criteria are shifting: how well does the platform expose its capabilities to external AI agents? How mature and stable is its API? Can workflows be triggered and completed without human interaction at every step? Does the vendor have a credible, architecturally coherent story for agentic use — not just a Copilot button on a menu?
There is also a governance dimension that deserves attention. As AI agents begin to act within SaaS platforms on behalf of employees — updating records, triggering workflows, generating outputs — the question of auditability becomes critical. Which actions were taken by a human, and which by an agent? On whose authority? With what guardrails? UK organisations operating under GDPR, financial services regulation, or sector-specific compliance frameworks will need clear answers. Vendors who have thought seriously about human-in-the-loop controls, audit trails for agentic actions, and permissioning models designed for non-human actors will be meaningfully better positioned than those who have not.
The shift from human-operated to agent-operated software is not a distant scenario. It is happening now, in pockets, across organisations that are early in experimenting with AI tooling. The organisations that will be best positioned are not necessarily those who move fastest, but those who think most clearly about what they are actually buying when they renew a SaaS contract. Is this platform built to serve humans navigating a workflow, or is it built to be a reliable, programmable substrate that AI agents can act within? That distinction — largely invisible in a standard demo or a vendor's marketing collateral — will define the useful lifespan of the software your organisation depends on.
If you are a technical lead or a senior decision-maker currently reviewing your software portfolio or planning a procurement cycle, the most valuable thing you can do is ask harder questions of your vendors. Not about their AI features, but about their AI architecture. The platforms worth investing in are those where intelligent agents and human oversight can coexist by design — not as an afterthought. At iCentric, we work with organisations navigating exactly these decisions, helping them build and integrate software that is designed for the way work is actually evolving, rather than the way it was done five years ago.
Which specific SaaS categories are most likely to be disrupted first by AI agents?
Categories where the primary value is workflow navigation rather than data storage are most exposed — CRM activity logging, project management status tracking, and BI dashboard interpretation are early candidates. Platforms where AI agents can replicate the core human activity via API without touching the interface will feel the pressure soonest. Data-rich platforms with strong API layers are more resilient, as their underlying asset retains value even if the interface becomes less relevant.
How can we tell whether a vendor's AI features are genuinely architectural or just a marketing overlay?
Look beyond the product demos and examine the API documentation — if AI capabilities are only accessible through the GUI and not exposed programmatically, they are likely surface-level additions. Ask vendors whether their AI features operate across the full data model or only within specific modules. Also consider whether AI-triggered actions appear in audit logs and whether the permissioning model distinguishes between human and agent actors.
Does this mean we should avoid renewing long-term SaaS contracts right now?
Not necessarily, but contract length and exit terms deserve more scrutiny than they may have received previously. For platforms where AI disruption risk is high, shorter renewal cycles or break clauses give you more flexibility to respond as the market evolves. For platforms where the underlying data model and API are strong, longer commitments may still be justified — the interface risk is lower when the asset is the data itself.
What does 'AI-native' actually mean in practical terms for a software product?
An AI-native product is one designed from the ground up with AI as a primary actor in the workflow, rather than a product that has had AI features added retrospectively. In practice, this means the data model, API design, permissioning, and core interaction patterns were all conceived to support non-human agents completing tasks — not just humans being assisted by suggestions. The difference shows up in reliability, depth of capability, and how naturally agentic workflows compose within the system.
How should we approach GDPR compliance when AI agents are taking actions within our SaaS platforms?
Under GDPR, accountability for data processing actions remains with the data controller — your organisation — regardless of whether those actions were taken by a human or an AI agent. You will need audit trails that capture which actions were agent-initiated, under what authority, and with what data. Ensure your SaaS vendors support agent-specific audit logging and that your AI tooling includes clear records of task provenance. Your DPO should be involved in reviewing any agentic workflows that touch personal data.
Are there SaaS platforms that are genuinely well-positioned to survive and thrive in an agentic world?
Platforms that own a critical, high-quality dataset — financial records, customer history, product catalogues — are better positioned because their core value is not the interface but the data. Similarly, platforms with mature, stable, well-documented APIs and a genuine investment in agent-friendly features (webhooks, event-driven triggers, granular permissioning) are more resilient. Vendors who have released dedicated agent frameworks or SDKs, rather than just Copilot-style chat overlays, signal a more credible architectural commitment.
How should procurement teams change their evaluation process for new SaaS purchases given this trend?
Expand your evaluation criteria to include API maturity, agent compatibility, and auditability of non-human actions alongside traditional UX and feature assessments. Consider running a proof-of-concept that involves an AI agent interacting with the platform programmatically, not just a human using the interface. Ask vendors directly for their roadmap on agentic workflows and what architectural investment underpins it — not just which AI features are on the product roadmap.
Could AI agents actually increase our dependency on certain SaaS vendors rather than reduce it?
Yes — if your AI agents are trained or optimised to work with a specific vendor's API and data model, switching costs could increase rather than decrease over time. This is a genuine risk, particularly if vendors design proprietary agent frameworks that create new forms of lock-in. Organisations should favour vendors who support open standards and interoperable APIs, and should architect their AI workflows so that the agent logic is decoupled from any single platform where possible.
What governance structures should organisations put in place before deploying AI agents within SaaS platforms?
At a minimum, establish a clear policy defining which actions AI agents are permitted to take autonomously versus which require human approval before execution. Implement role-based access controls specifically for agent identities, separate from human user accounts. Create a review process for periodically auditing agent-initiated actions, and define escalation paths for errors or unexpected outputs. Larger organisations should consider a formal AI governance committee with representation from IT, legal, compliance, and the relevant business functions.
Is bespoke software development becoming more attractive relative to SaaS in this environment?
For some use cases, yes. Bespoke software can be designed from the outset with agent-friendly architecture, giving organisations direct control over the API surface, data model, and auditability mechanisms rather than being dependent on a vendor's roadmap. The trade-off is the upfront investment and ongoing maintenance responsibility. The calculation is shifting, however — particularly for workflows that are genuinely differentiating, where a purpose-built system integrated tightly with your AI tooling may deliver more durable value than a general-purpose SaaS platform trying to retrofit agentic capability.
Get in touch today
Book a call at a time to suit you, or fill out our enquiry form or get in touch using the contact details below