The deployment of software-as-a-service (SaaS) has defined the past decade, as organisations moved away from managing network infrastructure towards a service-led model. Using SaaS requires less configuration and installation, as well as less demand for specialist talent and skills. This means teams can focus their energy on building applications and features that drive revenue.
Now, as organisations deploy artificial intelligence (AI) capabilities across the enterprise, attention is turning towards what is already being used. Will the next generation of SaaS market leaders be those that own the customer journey from end to end, using AI to connect workflows rather than simply add another intelligent feature? Or will SaaS be defined less by software capabilities and more by how effectively platforms integrate with and orchestrate business processes?
If this is the disruption being caused by AI, then as capabilities become easier to replicate, competitive advantage is shifting away from feature lists and towards owning entire customer workflows. Essentially, is AI disrupting what makes SaaS valuable, or is it commoditising the features that once differentiated these platforms?
George Brocklehurst, a research vice-president in the emerging tech and trends research practice at Gartner, says that over the past 12-18 months, we’ve seen a progressive shift from a question-and-answer model towards AI-enabled workflows designed to achieve a business outcome, where the model becomes a component in a more composite system.
He explains that, in its current direction, AI will not simply update traditional SaaS, but is “truly rendering the dashboard UI [user interface] experiences obsolete”. Therefore, the next generation of SaaS is not going to be defined by software capabilities, but by orchestrating cross-domain workflows for end-to-end outcomes.
Is AI disrupting what made SaaS valuable by commoditising the features that once offered differentiation? Brocklehurst says AI is being used to connect workflows more effectively, rather than simply adding more features.
“This is how leading-edge agentic solutions are gaining scale in enterprises,” he says. “Lots of enterprises are dabbling in it, trialling it, failing with it, but where it’s gaining traction is where those solutions better interpret context and retain institutional knowledge. The key takeaway here is that the threat to SaaS incumbents isn’t this full sci-fi autonomy, it’s unifying capability.”
Brocklehurst says leading agentic services are progressively arbitraging this market (exploiting the difference between what tasks currently cost and what AI agents can deliver), while suppliers focusing their R&D on making individual users faster inside their own stacks are going to be fighting for commodity pennies.
“These are stateless tools that miss or forget that business context the minute the session ends,” he says.
“The user doesn’t want to log in to five different dashboards to pull a report: they want a single unified agentic layer that does the heavy lifting for them across those systems.”
Brocklehurst explains that a user can collect data points from different SaaS stacks and use AI to bring them together into deliverables. “The moment that an agent sits between the user and the software, the legacy application becomes headless,” he says.
Fatigue by training?
Brocklehurst says surveys conducted by Gartner show that enterprise buyers are fatigued by AI, and buyers don’t want to be sold another tool that their team needs to learn. They just need the tool to “deliver business value, automate the outcomes and make it work across the stacks”.
He says that once an agentic service can customise how it interacts with users based on personal, functional and organisational relevance, those interactions become more relevant.
“They reduce friction in working with the technology, and that earns you the right to operate across multiple domains,” says Brocklehurst. “It builds trust. In other surveys that we’ve done, trust drives the frequency of engagement, and so as you more frequently engage with the technology and get to better outcomes, that drives the return on investment.”
He says this is helping to scale enterprise adoption, as every time you introduce something that makes a workflow more fragmented, it adds more friction. What we see is that those providers and platforms that absorb that adaptation tax unlock fragmented value streams that stretch across the enterprise.
No value in SaaS?
The rise of AI is beginning to challenge the traditional SaaS model, with Forrester describing a potential “SaaS-pocalypse” for suppliers that cannot demonstrate clear value in an AI-first world.
It says that horizontal point-solution providers are particularly exposed, especially where products have low switching costs, limited integration into enterprise workflows and an unclear return on investment.
While some AI-native suppliers may also be acquired by larger technology companies as the market consolidates, larger enterprise SaaS providers are responding by embedding AI agents into their existing platforms and moving towards AI-first business models.
Their advantage lies in established enterprise relationships, industry expertise, data, consulting ecosystems, and the ability to combine technology with processes and governance. However, even these suppliers could face disintermediation as AI orchestration, control and semantic data layers increasingly sit above traditional application suites.
Merrick Lackner, CEO of Rently, says we are not on the way to a SaaS-pocalypse. Rather than SaaS being dead, he says, we will see a more in-depth network of partnerships emerge.
“We’ve seen Anthropic and Salesforce announce this partnership where they’re able to mutually benefit each other, and I think we’re going to continue to see that,” says Lackner.
“When it comes to just purely agentic services synthesising information, that does become increasingly commoditised,” he adds. “You can get that kind of service through OpenAI, through Gemini, through Anthropic, and of course, as a business, you could wrap that into your service as well with one of those providers or with your own foundation model.”
Fitting into AI
For organisations buying SaaS, this means reassessing how their existing applications fit into a wider AI strategy, rather than simply adding more point solutions. Companies will need to rationalise their SaaS estates, address technical debt, establish better data governance, and understand which capabilities genuinely differentiate the business and which could increasingly be replicated by AI.
Contracts may also need to change, moving away from traditional seat-based pricing towards consumption- or outcome-based models as AI agents become part of enterprise software.
How does this affect existing security suppliers and service providers? Lackner says that if you look at firms such as Zscaler and CrowdStrike, or companies such as Salesforce, they have tremendous amounts of data, and we know that even the frontier models have reached the point where they can ingest publicly available information.
“I think we’re going into a world of partnership, so I think they’re going to mutually enhance one another,” he says. “AI tools are great at synthesising information, but then there’s specialised information, too. So, when it comes to specialised information, the route to go is definitely with mutually beneficial arrangements.”
Lackner also says the addition of AI services will allow providers to use transformer technology to sift through and analyse data, and come up with solutions based on it.
“Prior to a partnership, a frontier or foundation model may not have access to all that data,” he says. “Then these partnerships get constructed where the data can be viewed in whatever silos or repositories, and then act on it.
“Being plugged into data sources and so forth is basically making certain repositories of information available that AI and GenAI [generative AI] would not necessarily have readily available.”
Autonomous modification
Forrester says that as legacy systems, commercial software and large open-source estates were not designed for autonomous modification, organisations must remain cautious about allowing AI to make large-scale changes without oversight.
Rather than simply replacing existing security testing tools, the shift could instead favour integrated software development platforms where security, validation and remediation are built into the development process. This puts greater emphasis on tools and controls that can provide context, constrain AI-driven changes and validate that remediation is safe.
Lackner says that as more companies build their own foundation models, while also partnering for sophisticated tasks that require greater depth and complexity with frontier models, learning and security will be at the forefront.
“My personal opinion is that’s going to be done through partnerships in the coming decade, for sure, but I’m still of the opinion that humans are driving a big portion of this,” he says.
“AI is not sentient and conscious, you know, in the same way that a human is, and even as a matter of first principle, whatever an AI does, it does need to get an instruction. Even in recursive development, there is an initial instruction set that comes from people.”
AI may not spell the end of SaaS, but it is likely to change what users expect from it, shifting value away from individual features, and towards the ability to connect data, applications and workflows. The SaaS platforms that thrive may therefore be those that become less visible to users, providing the underlying intelligence, data and infrastructure that AI agents need to deliver business outcomes.
