Agent provocateur: could agentic AI mean the end of SaaS?

Agentic AI that can independently carry out complex tasks could cause the SaaS market to “collapse” according to Microsoft CEO Satya Nadella. Adam Hughes investigates…

First came GenAI, fundamentally changing how we interact with technology – and giving telcos a path to drastically cut their workforces. Now software itself could be on the chopping block, as traditional SaaS “will collapse in the agent era” according to Microsoft CEO, Satya Nadella, in a recent podcast:

“The business logic is all going to these agents, and these agents are going to be multi- repo CRUD right, so they're not going to discriminate between what the back end is… and once the AI tier becomes the place where all the logic is, then people will start replacing the back ends.”

Nadella believes that this shift will lead to the collapse of conventional business applications; in his words, “they are essentially CRUD [Create, Read, Update, Delete] databases with a bunch of business logic.”

Agentic AI is artificial intelligence systems that can make decisions and take actions autonomously, without the need for direct human intervention. In short, AI agents don’t just “think” – they also “do.”

AI agents can reduce response times and operational costs through 24/7 personalised customer support, troubleshooting technical issues and handling complex queries, all with unprecedented precision thanks to advanced natural language processing.

Is 2025 the year that autonomous AI agents are deployed to carry out everyday tasks with minimal human oversight? By analysing individual customer usage patterns, AI agents acting on behalf of CSPs can dynamically generate personalised experiences, tailored services and recommend optimal data plans.

If AI agents can anticipate and automate complex tasks, why would we need back-end SaaS systems at all? The agent becomes the interface, eliminating the need for a UI and, according to Nadella, making SaaS redundant, the value captured by the AI layer running on top.

We can already see this happening to some extent with Klarna, which terminated its partnerships with Salesforce and Workday in favour of its own in-house AI built on ChatGPT. But this seems more a case of using GenAI to build their own bespoke applications rather than agentic AI meaning these applications need not exist at all.

Meanwhile, SK Telecom is launching its AI agent in beta for North American users this March, with the intent to release it as a white-label service to other companies, including other telcos.

Are we entering an era where intelligent agents become the primary interface for business operations? Are Agents-as-a-Service the new SaaS?

To take Nadella’s assessment at face value as he simultaneously peddles Copilot is like uncritically taking an umbrella salesperson’s weather forecast without question.

Many Big Tech execs have been making bold statements over the last few years, and they don’t always translate into reality; certainly not for Microsoft which, for all of its investments in AI has not managed to keep ahead of its competitors with the “messy rollout” of its Copilot agent, foisted onto Office 365 users with no notice – and a hefty increase to their subscription cost.

While AI agents are autonomous in their decision-making processes and can take up a great deal of grunt work, the human element of customer services remains irreplaceable. Retrieving customer information from a database is one thing, but nurturing an ongoing relationship, understanding interpersonal dynamics and nuanced problem-solving cannot be easily replicated by algorithms.

AI agents still require goals defined by humans, and complex use cases will still require traditional enterprise applications – where the data and processes live – and instructions and monitoring from humans to ensure that they carry out their tasks correctly.

Nevertheless, SaaS companies insisting on business-as-usual risk falling into irrelevance. The future of SaaS will no doubt incorporate a more conversational user experience to dynamically sort unstructured data, producing actionable insights, and drive long-term customer value and security. 

The most effective technological solutions will emerge from a symbiotic relationship between AI capabilities and human expertise. Success will depend on balanced, strategic implementation that leverages technological potential while mitigating potential disruptions… and not just building a simple GPT wrapper around legacy software.

By all means use AI to make your reps smarter, but let humans engage with humans.