A seismic shift is taking place in the Software-as-a-Service (SaaS) space.

Tech leaders from Microsoft’s CEO Satya Nadella to Klarna’s Co-founder and CEO Sebastian Siemiatkowski have suggested a new breed of AI-powered business application could drive wholesale transformation of the economics of software companies.

In a recent interview, Nadella even gave the provocative assessment was that “business applications may collapse”. Historically, SaaS companies have successfully delivered value to their customers by applying specific business logic to large sets of data. However, in what Nadella calls “the agent era”, AI can do the same, which could mean that the value SaaS companies have been able to extract by integrating with customer databases may evaporate.

Let’s use a typical SaaS application designed for human resources (HR) teams as an example of this shift in the status quo.

A company’s workforce database is a live record of vital business information. When certain fields in these databases change, SaaS companies can derive significant value by developing technologies that automate processes based on those triggers. That might mean starting an onboarding process for a new employee when they’re added to the database, or initiating specific payroll actions to make sure people are paid correctly.

Providing a smooth and cost-effective way for companies to run these processes is the key to many SaaS companies’ success. But in Satya’s vision, information – and value – will no longer be locked up in companies’ databases. Instead, agnostic “multi-repo” AI agents will scan any relevant dataset or piece of documentation to extract insights and automate processes.

This effectively gives companies a seamless way to connect internal and external knowledge sources - creating a new “AI tier” for business logic that could destroy much of the value that SaaS companies have created over many years.

And that potential impact isn’t limited to process – it may impact pricing, too.

The “classic” SaaS pricing model, with a monthly per-user subscription cost, reflected an era where the marginal cost of providing software on an ongoing basis was near-zero.

However, the comparatively high compute cost associated with building AI products and services might encourage – or force – more SaaS companies into thinking about new pricing structures.

To Christian Owens, founder of payments and billing SaaS company Paddle, “AI could well catalyse a volume shift towards more usage-based pricing models”. However, any shift here might not be solely down to AI.

In this way, companies leveraging AI agents – software that executes end-to-end actions for a user, like booking a flight or changing the date of a meeting – might charge companies a monthly fee while triggering additional costs every time a user activates an “agentic” workflow. This may help companies cushion themselves from pricing risks while longer-term AI compute costs are still relatively uncertain.

So how can SaaS founders determine the right pricing strategy for their AI agents?

According to Robert Maguire, CEO at Altruistic, which builds AI technologies and consults with large organisations on AI implementation, there’s no need to reinvent the wheel.

“The products might be new, but it’s still sensible to use internal time and labour costs as your benchmark for pricing an AI solution,” says Robert. “If you know that it costs a bank £1,500 to underwrite an average loan, for instance, you can offer immediate value with an agent that reduces the time to reach an underwriting decision and charges, say, £750 per decision. That might be a good starting principle for a per-workflow pricing point, and the same model is transferrable across different industries.”

The success or failure of this kind of pricing strategy will likely depend on how effectively companies can articulate the value their AI products will bring to the customer.

With so much attention being paid to AI innovation, SaaS founders could be forgiven for thinking that if they’re not investing in building AI products and services, they run the risk of being left behind.

Certainly, a large proportion of companies are actively exploring AI integrations: a McKinsey study estimates that 92% of organisations intend to increase their investment in AI over the next three years. But the number of startups that are designed around AI from the ground up is still relatively small: Beauhurst data indicates that there are some 2,361 high-growth AI companies in the UK.

In this context, planning and executing the transition to an AI-focused business model – like executing any strategic pivot - is a serious undertaking, particularly with the technology developing at pace.

Part of the challenge for founders building with AI is determining whether they are ahead of the curve or not. “AI startups and scaleups are building in an information-deprived market,” says Robert. “There is a lot of buzz, but we don’t yet have really authoritative sources of market intelligence for most AI tools. In contrast, SaaS benchmarks are very well established and understood.” The rapid pace of development in AI means that “a lot of companies are prototyping products that might be nowhere near first to market.”

Christian thinks that while AI has real potential to transform software business models, it’s still early days for the space more generally: “Everyone in tech is getting very excited by AI, but if you asked the average high-street business, AI would be a long way down the list of their most important concerns.” And while AI’s reputation for occasional errors remains, business owners may be unwilling to outsource important commercial processes. “People who aren’t involved in the AI space are shocked when I tell them that most AI models are, say, 70% accurate,” says Robert. “That might be OK for restaurant recommendations, but not for your tax return.”

Another important consideration for SaaS founders is that when so much public domain information has been commoditised by foundational models, the proprietary data companies hold on their customers could become increasingly vital. “If you’ve been growing a SaaS business for a while, one big prize you already own is your store of information on your customer relationships and user behaviour,” says Christian. “AI startups in your sector don’t have this critical advantage, and if it was possible to access that data, they would bite your hand off to do so.”

Whatever your size and scale, any SaaS founder deciding to reinvent their company’s business model in the face of AI disruption confronts tough strategic choices, so it’s crucial to understand the true scale of the perceived threat.

Existing competitors and new AI-native challengers present different threats to SaaS founders.

Instagram co-founder and Anthropic Chief Product Officer Mike Krieger suggested in a recent 20VC interview that the risk of “alienating your existing customers” means that more mature companies opting to transform their offerings with AI could be at a disadvantage compared to upstart challengers who don’t have to worry about damaging long-standing client relationships.

With many of the most advanced AI products focusing on developers and software engineers, the nature of your sector and vertical matters, too. “If you’re a SaaS company focusing on workflow automation and you’re selling to tech-forward buyers, I’d potentially be worried if you’re not layering in AI,” says Christian.

It's worth bearing in mind that new problems and frictions may emerge as AI continues to develop. “Companies are already becoming increasingly fatigued with ‘SaaS bloat’,” says Christian, referring to the frustrations experienced by companies that have adopted large numbers of SaaS products which end up being underutilised. (Gartner estimated that unused IT features would create a $750 million ‘overspend’ in 2023.)

Satya Nadella’s vision of wholesale turmoil for the SaaS sector is likely to entice some founders and worry others.

Certainly, AI could lead to founders rethinking parts of their business models, such as incorporating AI compute costs into their pricing structures. But even in a time of rapid change in tech, the datasets and customer relationships that SaaS companies have built over many years may provide valuable moats against bottom-up AI disruption.

The risks and opportunities of an AI-centric era for software are both real. The decisions SaaS founders make on AI could shape the direction of their companies for years to come.