SaaS meets AI agents: Transforming budgets, customer experience, and workforce dynamics
As AI agents pervade the SaaS market, how businesses experience and leverage software will likely change—shifting business models, capabilities, and expectations
As agentic AI capabilities mature and enterprise software-as-a-service (SaaS) vendors build out their platforms to create, integrate, and orchestrate AI agents, how organizations purchase and use software could shift dramatically. In 2026, SaaS applications will likely become more intelligent, personalized, adaptive, and autonomous, evolving towards a federation of real-time workflow services that can learn from their experiences. This evolution should disrupt traditional pricing models. Subscriptions and seat-based licensing could give way to hybrid approaches that blend usage- and outcome-based pricing. All these advancements will likely introduce new complexity in both software implementation and monetization—potentially redefining the entire SaaS business model.
What is an AI agent?
In artificial intelligence, an intelligent agent is an entity that perceives its environment, takes actions autonomously to achieve goals, and may improve its performance through machine learning or by acquiring knowledge.1
AI agents could drive a gradual transformation of SaaS markets starting in 2026
To put things in perspective, let’s take a step back and look at how overall AI adoption appears to be evolving in the market. Deloitte’s 2025 Tech Value survey found that 57% of respondents were putting between 21% and 50% of their annual digital transformation budgets into AI automation, and 20% of respondents were investing 50% or more (US$700 million on average for a company with $13 billion in revenue).2 Nearly three-quarters of surveyed leaders said their organizations funded AI and generative AI technology capabilities over the last 12 months (the No. 1 area) and 39% funded agentic AI.
Based on this, Deloitte predicts that up to half of organizations will put more than 50% of their digital transformation budgets toward AI automation in 2026, and agentic AI will see an even higher percentage of companies investing, perhaps reaching 75%. Although the Tech Value survey focused on US respondents only, we believe that companies around the world will follow a similar path, possibly delayed by a year or two. SaaS is often foundational to digital transformation efforts, and treating these broader spending shifts as a proxy, we expect commensurate increases in spending toward autonomous AI agents as part of SaaS in the next year.
Where could all this investment and technological advancement eventually lead? There are some optimistic visions of the future getting attention. Some have stated that parts of, or even entire, enterprise applications could eventually be replaced by agents.3 Deloitte predicts that this future may ultimately come to pass for some enterprise applications, but it won’t be in 2026. It will likely take at least five years or more to come to fruition, even with the rapid pace of technological development and investment around agentic AI. There are challenges to this vision, as traditional SaaS providers have large footprints across complex workflows that will likely be hard to supplant.4
In 2026, we will likely see a lot of experimentation, a general augmentation of capabilities, and a slow restructuring of the SaaS market, with AI-first companies competing. This moderate pace is likely because the “agentification” of SaaS is often not only about technological change, but business and operating model change as well—for both vendors and users.
With agentic AI, SaaS is about to get more complex
Many CIOs and CTOs continue to face pressure to reduce costs and streamline the number of vendors that they use.5 In an agentic AI era, the question often arises, when and how should organizations start to shift their investments toward solutions with AI agents in the hopes of greater efficiency?
There are a couple of different paths some of the largest SaaS providers are taking in their approach to providing these capabilities to their customers. Many are adding AI agents to existing products and producing brand-new AI agent–powered products (Salesforce Agentforce, SAP Joule Agents, ServiceNow Now Assist AI Agents, and Workday Illuminate Agents are recent examples).6 Many are also creating agent-building frameworks built on top of current services and introducing new data management and orchestration capabilities to help make the creation and management of AI agents easier (Google Cloud Agent Development Kit, Oracle AI Agent Studio, SAP Business AI, Workday Build and Adobe Experience Platform Agent Orchestrator are recent examples).7
In an agentic AI era, the question often arises, when and how should organizations start to shift their investments toward solutions with AI agents in the hopes of greater efficiency?
In addition, some new AI-native companies appear to be developing agentic solutions that could potentially disrupt these incumbents. In the short term, “easier” business processes like customer service are more likely to be disrupted, but disruption could spread to more complex markets like ERP (enterprise resource planning) and CRM (customer relationship management). Significant amounts of investment are powering many of these startups.8 Many of these emerging companies are likely to get acquired in the next few years as incumbents look to expand their portfolio of agents and seek differentiation. In fact, Gartner® says, “By 2030, 35% of point-product SaaS tools will be replaced by AI agents or absorbed within larger agent ecosystems of major SaaS providers.”9
Today, organizations have access to AI agents through their existing SaaS providers, which can make it easier to test and learn how to build agentic solutions through built-in functionality. While organizations may take this agentic-by-default approach initially, as they gain more experience, they will likely shift toward a more deliberate tack. Building around their data, Deloitte predicts they will pick capabilities from a broad and complex agentic ecosystem, develop their own agents, and weave everything into an integrated and autonomous multi-agent system.
Navigating the transition to agentic AI
To more successfully get to this future, several challenges should be addressed:
Pricing becomes more complex
An area that will likely significantly impact both SaaS users and vendors alike will be how using AI agents will be priced and paid for. When software was mostly on-prem, you typically had a perpetual software license and paid for upgrades and upkeep. The SaaS revolution, driven by the cloud, shifted things to subscriptions. Today, there are a couple of common pricing approaches for SaaS. Generally, organizations are charged based on the number of users or seats that they have. These seats could include a tiered pricing option, where different tiers provide different sets of capabilities based on the type of user. Such pricing can be relatively straightforward and predictable. Usage or consumption-based pricing appears to also be increasingly common, and less predictable. This model is often based on the number of API calls or tokens (units of text or data an AI model processes) used.
As AI agents enter more widespread use, these traditional pricing models won’t be adequate to reflect the true value exchange between provider and consumer.10 AI agents could conceivably give one user the power of many users and reduce the need for the number of seats needed in an organization, impacting the revenue of SaaS providers. Additionally, AI agents operate autonomously, and their actions aren’t necessarily predictable; they may take novel or inefficient paths while completing their tasks.
There will likely be a lot of effort needed to shift to these newer models, and we expect to see pricing variety and experimentation in 2026 and beyond. It could take years for standard practices to emerge, if they ever do. There are a couple of pricing models that are expected to gain in popularity: usage-based and outcome- or value-based. Gartner says that “by 2030, at least 40% of enterprise SaaS spend will shift toward usage-, agent-, or outcome-based pricing.”11
Usage-based pricing
In usage-based models, a customer could be charged every time an agent takes an action or completes a task. Pricing could also be based on computing time, API calls, the number of tokens used for generative tasks, or how long an agent is in action (or a combination of all of those). There could also just be a flat fee per time period for the use of a single agent, like a salary for a digital worker. In a recent survey of SaaS companies, Maxio found that 83% of AI-native SaaS companies currently offer usage-based pricing.12 Usage-based pricing is often attractive because it is quantifiable and therefore auditable.
Outcome- or value-based pricing
Pricing model changes will impact multiple functions within organizations and may transform how SaaS vendors operate.
Outcome- or value-based pricing is based on the real business results that SaaS applications with AI agents produce—something that can be much harder to measure. This could be as simple as the number of customer support tickets that get resolved or how many employees were eventually hired because of an HR agent, or it could be as complex as an increase in overall revenue AI agents contributed to. There’s likely still a long way to go before there’s widespread use of this model, though some are pursuing it.13 Agentic systems still need to prove that they can produce consistent and reliable value.
These pricing model changes will impact multiple functions within organizations and may transform how SaaS vendors operate. First, there should be agreements around basic definitions for things like “an agent,” “a task,” “a process,” “an interaction,” and “an outcome.” What “value” is and how it is attributed should be clearly defined, communicated, and agreed upon contractually. This will likely take significant effort and coordination from engineers, sales people, legal teams, and others. Proving that an AI agent created value or a business outcome could be challenging, especially if multi-agent systems composed of agents from different vendors are used. Revenue for vendors and costs for customers could become less predictable and highly variable. System instrumentation and metering may have to become more advanced and data observability, billing, and financial compliance may have to become more real time and autonomous.
The sales models for many vendors will likely need to change. Sellers will have to educate customers on these new models and convince them that AI agents will create value and the shift won’t cost them more than their subscription-based services. Sellers will also likely have to be measured and compensated differently and may have to drive deeper relationships with customers.
Customer experience and user interfaces could become greater differentiators
AI agents are, by nature, supposed to be autonomous, so why do they need to have a user interface? Like APIs, agents are “headless.” They don’t have a direct connection to a user interface. However, someplace for interaction and visibility is necessary. So, what will that look like? Will there be a single, primary AI agent interface or multiple ones? Will a SaaS provider or third party “control” a gateway to agents?
Over the next few years, Deloitte predicts that the user experience and interface for SaaS AI agents will become more:
- Personalized and proactive: The interface will adapt to the individual user providing needed tools and tasks based on specific responsibilities and prior actions. It will provide tailored insights and suggest specific actions for users to take.
- Conversational: Interaction will move from menus and clicks to natural language and voice commands. AI agents will translate natural language into a structured series of API calls, eliminating the need for pre-defined workflows. It will be less about telling software what to do and more about asking software to achieve a particular outcome.
- Diagnostic: Because of the autonomous nature of AI agents, if something wrong or unexpected happens, users will have to be able to reconstruct the agent’s decision-making process and understand why it happened. Transparency, explainability, reversibility, and auditability will be crucial for trust.
Another open question is where will the interaction layer be? Deloitte predicts a lot will be done in stand-alone SaaS apps. Many SaaS providers want to keep users in their application as much as they can to maintain worker efficiency and keep users using their products. They will increasingly provide access to not only their own suite of agents, but agents from other providers as well. Interaction could also take place through a separate management platform. These might be provided by a SaaS vendor or they could come from a third-party company (like current SaaS management platforms). These “control centers” could integrate agents’ activities from multiple vendors and internally developed agents—tracking usage, expenditures, access, performance, status, security, and compliance.14 There will also likely be agent marketplaces, where internal and external agents get published and businesses can discover and integrate new capabilities dynamically.15 This interaction, or attention, layer has the potential to provide significant value, and there is likely to be considerable competition around it.
The bottom line
In 2026, the usage of AI agents through SaaS applications is set to rapidly grow, with many major SaaS providers working to implement more robust agentic AI solutions with their customers. We expect increased investment in all AI-powered automation, extending into SaaS applications. Organizations will be seeking process efficiency, cost savings, greater flexibility and personalized capabilities for workers. There will be a lot of experimentation and pricing variety. Overall, Deloitte predicts there will be a gradual move toward a future powered by integrated, autonomous multi-agent systems.
Considerations for SaaS customers to help prepare:
- Invest in data management: For AI agents, access, integration, observability, and data governance can become even more important. Data doesn’t necessarily need to be centralized in a single repository, but it should be consistent and accessible across an organization.
- Embrace the growing complexity: There will be more models, more agents, more vendors, new ecosystems, and new data relationships. Organizations will have to get agents from different vendors to work together, potentially causing pricing and operational complexity.
- Expect multifaceted pricing models: Pricing models for AI agents could create ambiguity as hybrid models that include a mix of licenses and usage-, value-, or outcome-based pricing become standard. Bolster your real-time finance capabilities.
- Help workers become AI orchestrators: More time could be spent managing AI agents like co-workers—setting goals, supervising their work, and validating and correcting their actions. When rearchitecting workflows, clearly define what humans will do, what agents will do, and what they will do together. It’s a cultural shift, not just a software upgrade.
Considerations for SaaS vendors to help prepare:
- Prepare for greater competition: As it becomes easier and easier for anyone to write code through generative AI tools, the cost of producing code approaches zero. This is going to create greater competition with AI-native companies and even customers themselves—requiring greater product differentiation.
- Focus on interoperability: Agents will have to operate across multiple systems, coordinate tasks, and share data and goals—all while maintaining security and compliance. Organizations should prepare for a more open and interoperable environment, in which customers can easily switch providers if expectations aren’t met.
- Shift sales models: With varied pricing models, things like revenue forecasting become more challenging. How sales teams are measured and compensated will have to evolve. Help customers predict costs and provide hybrid pricing models that are simple and flexible. Use conversations around pricing models to expose unmet needs and deepen relationships.
BY
David Jarvis
Sayantani Mazumder
Girija Krishnamurthy
Gopal Srinivasan
China Widener
Gillian Crossan
The authors would like to thank Diana Kearns-Manolatos, Faruk Muratovic, Rohan Gupta, Khusro Khalid, Laura Shact, Girish Srinivasan, Pavan Srivastava, and Prakul Sharma for their contributions to this article.
Cover image by: Jaime Austin; Adobe Stock