Bain Plans $100 Billion for Agentic AI in SaaS

A Colossal Market Potential for Agentic AI in SaaS

Bain & Company, a renowned global consulting firm, recently published a study forecasting a potential market of $100 billion for SaaS companies that integrate agentic AI in the United States. This emerging market primarily focuses on automating coordination tasks within enterprise systems, an area where AI could revolutionize current practices.

This ambitious estimate comes from the second report in a five-part series dedicated to analyzing the software industry in the age of artificial intelligence. Bain explores how agentic AI could not only create new markets but also provide SaaS companies with unprecedented opportunities to optimize their operations.

Automating Coordination Tasks: A Central Challenge

According to Bain, the targeted market essentially concerns the manual work that employees perform between different enterprise applications. These workflows often include systems such as ERP (Enterprise Resource Planning), CRM (Customer Relationship Management), and various support tools. They may also integrate supplier management tools and email platforms.

Typical tasks in this context include extracting data from one system to verify it against another source, interpreting unstructured messages, and making decisions such as approval, response, escalation, or waiting. Bain emphasizes that traditional automation methods, based on rules or robotic processes, show their limits when faced with workflows that involve ambiguity and information scattered across multiple systems.

Agentic AI, on the other hand, is capable of interpreting information from various sources, coordinating actions between different systems, and operating within established policies. Bain clarifies that agentic AI does not aim to replace SaaS platforms but rather to convert intensive coordination work into software expenditures. Currently, providers are already capturing between $4 billion and $6 billion of this market in the United States, leaving over 90% of the potential untapped.

Market Distribution by Function

The market potential is not evenly distributed across different business functions. Bain estimates that the sales sector represents the largest share, with approximately $20 billion. This dominance is attributed to the high number of employees in this area, rather than an exceptionally high automation potential.

Cost of goods sold and operations total around $26 billion. The substantial size of the operational workforce means that even modest automation rates can translate into a significant addressable market. Other functions, such as research and development, engineering, customer support, and finance, each represent between $6 billion and $12 billion in addressable market size.

Customer support and the fields of R&D or engineering present the greatest automation potential, with about 40% to 60% of tasks being automatable. These sectors benefit from structured data, standardized processes, and clear output signals. Finance and human resources fall within a range of 35% to 45%, with higher automation potential for tasks like accounts payable and payroll, while financial planning and employee relations require more judgment.

Sales and IT have an automation potential between 30% and 40%. Bain notes that the complexity of relationships, variability from one deal to another, and the unpredictable nature of security incidents limit automation in these areas. The legal sector presents an overall lower automation potential, between 20% and 30%, due to the need for strict oversight to avoid costly errors.

Key Factors for Automation According to Bain

Bain identifies six key factors for assessing the portion of a workflow that can be automated by an AI agent. These factors include the verifiability of outcomes, the consequences of failure, the availability of digitized knowledge, and the variability of processes.

Workflows with clear verification signals are easier to automate than those requiring subjective judgment. Examples include code compilation, invoice reconciliation, and support ticket resolution. However, workflows involving regulatory or financial risks still require human oversight, even if AI agents are technically capable.

Bain also emphasizes that the availability of digitized knowledge is crucial. AI agents need access to structured data and documented context, as well as machine-readable inputs, often held informally by experienced employees.

Integration complexity is another important factor, especially when workflows span multiple systems and APIs. Authentication layers and exception management processes add further complexity, making these workflows more challenging to automate end-to-end. Higher-value areas focus where no single record system controls the entire outcome, often encompassing ERP, CRM, and support systems.

David Crawford, president of Bain's global technology and telecommunications practice, stated that SaaS companies have spent the last two decades building positions around record systems. The next source of advantage lies in "cross-workflow decision-making context," meaning the ability to interpret and act within workflows that traverse multiple systems.

Examples of Companies and Adjacent Workflows

Bain's report cites several companies such as Cursor, Sierra, Harvey, Glean, Salesforce, ServiceNow, and Workday to illustrate the adoption of agentic AI. Cursor, for example, has surpassed $16.7 million in average monthly revenue after doubling its revenue in a single quarter. Sierra has reached $150 million annually, Harvey $190 million, and Glean $200 million.

Bain also mentions GitHub as an example of a company using data from a central workflow to expand into adjacent activities. While its core business is developer collaboration and version control, GitHub has leveraged its repository and workflow data to support expansion into AI-assisted developer productivity and security automation.

According to Bain, SaaS companies can expand through two types of workflow automation. The first type involves automating core workflows where they already possess domain knowledge and customer trust. Existing system integrations can support this automation. The second type concerns automating adjacent workflows that the company does not currently serve directly. These areas can be more challenging to identify, as they require detailed mapping of customer workflows and the underlying data that supports decisions.

Pricing models may evolve as agents deliver complete outcomes. Bain notes that outcome-based and usage-based pricing could become more relevant when agents solve problems or process invoices, contrasting with traditional seat-based and connection-based pricing.

Bain's Recommendations for SaaS Companies

Bain recommends that SaaS companies start by identifying which customer workflows can now be automated through agentic AI. The firm suggests evaluating automation at the sub-process level, rather than considering entire functions as equally automatable.

The report also highlights the importance of assessing data quality. Bain indicates that companies must check whether their data is complete, linked to outcomes, and usable for automation.

To fill capability gaps, Bain proposes several approaches: internal development, acquisitions, or partnerships. The report cites examples such as the development of the Axon platform by AppLovin, the acquisition of Moveworks by ServiceNow, and Salesforce's partnership with Workday.

Finally, Bain stresses the need for AI engineering talent, a cloud-native architecture for multi-agent orchestration, and funding for model training and inference. The firm recommends aligning pricing and business incentives with AI-driven outcomes rather than legacy seat-based models.

Bain concludes that SaaS companies will also need to develop data and product foundations designed for agentic workflows, including machine-readable transfers and systems that capture decisions and outcomes from each workflow execution.

David Crawford emphasized that the timeline for SaaS companies is "measured in quarters, not years," as AI-native companies gather more deployment data with each customer workflow they automate.