Bain & Company projects a substantial US$100 billion market opportunity in the United States for SaaS companies that deploy agentic AI to automate complex "coordination work" within enterprise systems. This significant estimate is detailed in the second report of Bain’s five-part series, which delves into the evolving software industry influenced by artificial intelligence. The report specifically investigates how agentic AI can foster new software markets and outlines strategies for SaaS providers to capitalize on these emerging opportunities.

The core of this burgeoning market lies in automating the manual, often fragmented, workflows that employees currently perform across various enterprise applications. These critical tasks frequently span multiple systems, including Enterprise Resource Planning (ERP), Customer Relationship Management (CRM), and support systems, as well as involving vendor management tools and email communication. Examples of such coordination work include extracting data from one system and cross-referencing it with another source, interpreting unstructured messages, and making informed decisions on whether to approve, respond, escalate, or defer action.

Traditional automation methods, such as rules-based automation and robotic process automation (RPA), often struggle with workflows characterized by ambiguity and information spread across disparate systems. Agentic AI, however, offers a solution by interpreting information from diverse sources, coordinating actions seamlessly across multiple systems, and operating effectively within established policy guardrails. Bain emphasizes that agentic AI is not primarily designed to replace existing SaaS platforms but rather to transform labor-intensive coordination tasks into efficient, software-driven processes. Current market capture by vendors is estimated at US$4 billion to US$6 billion in the US, indicating that over 90% of this market remains untapped. Furthermore, Bain projects that regions outside the US, including Canada, Europe, Australia, and New Zealand, could contribute a similar market size, pushing the global total in these regions to approximately US$200 billion.

The addressable market for agentic AI automation is not uniformly distributed across enterprise functions. Sales, for instance, represents the largest single share at an estimated US$20 billion, primarily driven by the sheer number of sales employees rather than an exceptionally high automation potential per task. Cost of goods sold and operations together account for about US$26 billion, where even modest automation rates can translate into a substantial addressable market due to the large operational workforce. Functions such as R&D and engineering, customer support, and finance each present an addressable market of US$6 billion to US$12 billion. These areas benefit from sizeable workforces and higher automation potential in specific workflows. Customer support and R&D/engineering exhibit the highest automation potential, with roughly 40% to 60% of workflow tasks being automatable, attributed to their structured data, standardized processes, and clear output signals. Finance and human resources fall within the 35% to 45% automation range, with higher potential in areas like accounts payable and payroll, while financial planning and employee relations require more human judgment. Sales and IT hover around 30% to 40% automation potential, limited by factors such as relationship nuance, deal-by-deal variations, and the unpredictable nature of security incidents. Legal functions show lower overall automation potential, ranging from 20% to 30%, as tasks like contract review and compliance, while repeatable, demand stringent oversight due to the high consequences of errors.

Bain’s report identifies six critical factors that determine the realistic extent of workflow automation by an AI agent. These include output verifiability, where workflows with clear verification signals (e.g., compiling code, reconciled invoices) are simpler to automate than those requiring subjective judgment. The consequence of failure is another crucial factor; workflows involving high regulatory or financial risk (e.g., tax filings, legal compliance, security incident response) necessitate closer human supervision, even when agents are technically capable. Digitized knowledge availability is also a constraint, as agents require access to structured data, documented context, and machine-readable inputs, including decision logic often informally held by experienced employees. Process variability and integration complexity further impact automation; workflows traversing multiple systems, APIs, authentication layers, and exception-handling processes are harder to automate end-to-end compared to those contained within a single platform. The highest-value areas for agentic AI are often concentrated where no single system of record fully controls the outcome, typically spanning ERP, CRM, and support systems. David Crawford, chairman of Bain’s global technology and telecommunications practice, highlights "cross-workflow decision context"—the ability to interpret and act within workflows that move through multiple systems—as the next key source of advantage for SaaS companies, who have historically focused on building systems of record.

The report cites several companies successfully adopting agentic AI, including Cursor, Sierra, Harvey, and Glean, all demonstrating significant revenue growth. For instance, Cursor surpassed US$16.7 million in average monthly revenue, while Sierra crossed US$150 million, Harvey passed US$190 million, and Glean reached US$200 million per annum. Beyond direct agentic AI solutions, the report also points to companies like GitHub as an example of leveraging data from an existing core workflow to expand into adjacent areas. GitHub, whose core business is developer collaboration and source control, used its repository and workflow data to support expansion into AI-assisted developer productivity and security automation. SaaS companies can expand through two primary types of workflow automation: automating core workflows where they possess domain knowledge, customer trust, and existing system integrations; or automating adjacent workflows that they do not currently serve directly, which requires meticulous mapping of customer workflows and the underlying data supporting decisions.

The advent of agentic AI also suggests a shift in pricing models. When agents deliver completed outcomes, such as resolving issues or processing invoices, outcome- and use-based pricing models become more relevant, moving away from traditional seat- and login-based models.

Bain & Company offers several recommendations for SaaS companies to navigate this landscape effectively. Companies should begin by identifying which specific customer workflows are now automatable with agentic AI, assessing automation potential at the subprocess level rather than treating entire functions uniformly. A critical step involves evaluating the quality of their data, ensuring it is comprehensive, tied to desired outcomes, and suitable for automation. Companies can address any capability gaps through internal development, strategic acquisitions (like ServiceNow acquiring Moveworks), or partnerships (such as Salesforce partnering with Workday), similar to AppLovin’s in-house development of its Axon platform. Essential prerequisites include securing AI engineering talent, developing cloud-native architectures capable of multi-agent orchestration, and allocating sufficient funding for model training and inference. Furthermore, Bain advises aligning pricing strategies and sales incentives with AI-driven outcomes rather than adhering to legacy seat-based models. Finally, SaaS companies will need to establish robust data and product foundations explicitly designed for agentic workflows, incorporating machine-readable hand-offs and systems that effectively capture decisions and outcomes from each workflow execution. David Crawford underscores the urgency, stating that the timeframe for SaaS companies to adapt and capture this market is "measured in quarters, not years," as AI-native companies rapidly accumulate deployment data with each automated customer workflow.