When ServiceNow dropped $2.85 billion to buy Moveworks in early 2025, it wasn’t just another tech acquisition. It was a clear sign that the software world is changing. Big companies are beginning to accept that the next generation of SaaS won’t revolve around apps we click through; it will revolve around AI agents that do the work for us. These agents are growing up fast. What started as chatbots that answered questions is now turning into digital workers that can understand a problem, decide what to do, and carry out tasks on their own. And businesses are paying attention. Deloitte expects that one in four companies using generative AI will experiment with AI agents this year, and that number could double within two years.
Investors and analysts see the same momentum. Analyst forecasts estimate the AI agent market at around $7-8 billion today, with some projections suggesting it could exceed $90 billion by 2032. Over the next few years, a growing share of enterprise SaaS platforms will include built-in agents as part of their core product roadmaps.
For entrepreneurs, this shift is bigger than a feature update. It changes the very idea of what software is supposed to do. Instead of tools that help people work faster, we’re moving toward tools that can actually do the work. And that change will reshape how software delivers value, and how customers pay for it.
From Co-Pilots to Autonomous Operators: Understanding the Shift
AI in software has grown in stages, much like the evolution of self-driving cars. For years, SaaS tools have relied on people clicking buttons, entering data, and moving through fixed steps. Even with the rise of AI co-pilots, humans have stayed in charge; they offer suggestions, but the user still does the work. Now, a new kind of AI is emerging. Autonomous agents don’t wait for instructions on every step. You can give them a goal like “cut our Q2 ad costs by 15%” or “fix all billing-related tier-one tickets,” and they figure out how to get it done. They break the job into smaller pieces, explore different options, make decisions, and then take action across different systems.
This shift requires a totally different kind of software design. Instead of building interfaces for humans to navigate, developers are creating systems powered by large language models that can use tools, remember context, plan ahead, and make decisions.
Industry surveys show how fast this shift is happening: many enterprises are already experimenting with AI agents, and adoption is accelerating each year. Most are still early in the autonomy journey, but a few have already reached more advanced levels in tightly focused areas.
New Business Models: From Seats to Outcomes
Autonomous software is quietly blowing up one of SaaS’s oldest assumptions: the idea that you charge per user. When the software does the work on its own, that model stops making sense. Do you charge for each agent? For each person it helps? For every task it finishes? No one really knows yet.
What’s becoming clear, though, is that companies are moving toward paying for results. Instead of buying access to software, customers are buying outcomes: tickets closed, leads qualified, documents processed, campaigns improved. It’s a shift that puts both sides on the same team, because everyone cares about what actually gets done.
This change ripples through the entire business model. It may cost more to win a customer upfront, since vendors must prove real results. But those customers tend to stay longer and spend more when pricing grows with their success. Churn becomes tied to performance: people leave only if the outcomes drop or someone else can deliver better ones. And as autonomous agents take on tougher tasks and deliver stronger results, the value and the revenue naturally grow.
Multi-Agent Collaboration: The Rise of Software Ecosystems
AI agents are already useful on their own, but things get far more interesting when they start working together. That’s where the real transformation begins. Picture a sales team run mostly by agents: one finds promising leads, another researches the company, another writes and sends outreach emails, another books meetings, and another updates the CRM. All of it happens in the background, with no one juggling tasks or handing work off.
The same idea works in customer support. One agent sorts tickets, another looks up the right answers, another drafts replies, another decides when a human should step in, and another watches for patterns in customer issues.
For these teams of agents to operate smoothly, they need systems that let them talk to each other, share what they know, divide tasks, handle disagreements, and be monitored like any other team. And this future isn’t far away. Gartner expects that by 2028, AI agents will make around 15% of routine workplace decisions, up from almost zero in 2024. In other words, software is starting to collaborate the way humans do.
Real-World Applications: From Hype to Production
For all the hype around autonomous agents, many companies are already putting them to work. At Genentech, an agent built on AWS now handles the repetitive research tasks that used to slow scientists down, giving them more time to focus on real discovery. The system breaks big research jobs into smaller steps and runs them automatically.
Rocket Mortgage is using Amazon Bedrock Agents to power a platform that sifts through enormous amounts of financial data. It delivers personalised mortgage advice in real time, an early sign of how agents could reshape customer experiences.
IBM is moving in the same direction, rolling out tools for managing groups of agents and offering ready-made ones for HR, procurement, and sales. Soon, its Watsonx Orchestrate system will integrate with Amazon Q Index, letting agents work across tools like Salesforce, Slack, and Zendesk.
Amazon Lex V2 has also been upgraded with agent-style features, including support for Bedrock Knowledge Base and Claude 3. And in July 2025, OpenAI introduced ChatGPT Agent, a system that blends Operator and Deep Research into one tool that can browse websites, build presentations, schedule meetings, fill forms, and run complex research workflows.
Challenges and the Path Forward
Autonomous software may feel like the next big breakthrough, but building it is far from easy. The tech itself is complex; agents need LLMs, orchestration systems, integrations with dozens of apps, and constant learning. And plugging them into existing company systems isn’t simple either, since every tool has its own API rules.
Security becomes a bigger issue, too. When agents can access sensitive data, companies need solid guardrails. But perhaps the biggest challenge is people. Workers often worry that autonomous tools will replace them, so teams need clear training that shows agents are here to help, not to take jobs.
Reliability is another sticking point. Enterprises can’t rely on software that gets things right “most of the time,” especially when an error from one agent can spread through a whole multi-agent system.
Still, 2025 is a rare moment for SaaS founders. The companies that will stand out aren’t the ones tacking on autonomy at the end; they’re the ones rethinking software around what agents can uniquely do. That starts with finding workflows where automation clearly beats manual work, rolling out autonomy gradually, and making sure customers can see and understand the agent’s decisions.
Autonomous software isn’t just an upgrade; it fundamentally changes how people and software work together. And the companies that recognise this shift now will define the next decade of SaaS.
Explore more insights on AI agents, SaaS disruption, emerging business models, and the strategies shaping tomorrow’s most successful entrepreneurs on Inspirepreneur Magazine.
