What is agentic AI and why is Bain talking about it?

Agentic AI represents a shift from classic chatbots and generative models to autonomous systems that independently complete tasks across various applications. Instead of merely answering user queries, these agents interpret unstructured information, coordinate actions across multiple systems simultaneously, and make decisions within given boundaries.

Traditional automation based on rules or robotic process automation (RPA) has reached its limits in situations where workflows contain ambiguity and data is scattered across multiple platforms. Agentic AI can draw information from ERP, CRM, support systems, and emails, evaluate it, and take subsequent steps—such as approving an invoice, escalating a ticket, or updating a database record.

According to David Crawford, chair of global technology and telecommunications practice at Bain & Company, SaaS companies have spent two decades focusing on building so-called systems of record—central systems that record data. The new source of competitive advantage, however, lies in so-called cross-workflow decision context, i.e., the ability to interpret and act in processes that pass through multiple systems at once.

How much is on the table: Analysis of the $100 billion market

Bain estimates that providers currently capture only $4 to $6 billion of the US market. This means that over 90% of the potential remains untapped. If we include Canada, Europe, Australia, and New Zealand, the total addressable market reaches approximately $200 billion.

Interestingly, this market is not evenly distributed. The largest share goes to the sales area—estimated at $20 billion in the USA. The reason is not exceptionally high automatability, but simply the large number of workers in the field. Cost of goods sold and operations represent approximately $26 billion. Research and development, customer support, and finance each offer a market in the range of $6 to $12 billion.

Biggest opportunities by corporate functions

The degree of automation varies significantly depending on the business process. Bain identified key differences:

• Customer support and development: 40–60% of tasks can be automated. These areas have structured data, standardized procedures, and clear success metrics.
• Finance and human resources: 35–45% of tasks. Invoice processing and payroll are easier to automate than financial planning or resolving personnel relations.
• Sales and IT: 30–40%. Nuances in customer relationships and unpredictable security incidents limit the degree of trust in autonomous decision-making.
• Legal department: 20–30%. Contract review is repeatable, but the consequences of errors require stricter human oversight.

Who is already making money in the market?

The report cites specific examples of companies successfully commercializing agentic AI. The developer tool Cursor exceeded average monthly revenues of $16.7 million after just one quarter of growth. The conversational platform Sierra surpassed $150 million annually, the legal assistant Harvey has more than $190 million in annual revenue, and the enterprise knowledge search engine Glean reached $200 million per year.

Major players such as Salesforce, ServiceNow, and Workday are also investing in agentic capabilities. ServiceNow, for example, is strengthening its portfolio through the acquisition of Moveworks, while Salesforce is collaborating with Workday to connect sales and HR workflows. GitHub, in turn, shows how to expand from the main workflow—developer collaboration—into adjacent areas such as security automation and productivity.

What does this mean for Czech and European companies?

Bain explicitly mentions Europe as a region with similar potential to the USA. For Czech companies, this means two things: opportunity and challenge. The opportunity lies in the fact that the Czech market is part of this addressable space. The challenge, however, lies in readiness.

Agentic AI needs structured data, digitized processes, and connected systems. Many Czech medium-sized enterprises still work with isolated databases, Excel spreadsheets, and manual hand-offs between departments. Without a digital foundation, agents cannot function effectively. Investment in cloud infrastructure and API standardization should therefore be the first step.

At the same time, European companies must respect the EU AI Act, which places increased demands on transparency and human oversight of automated decision-making systems. Precisely in areas with high consequences of errors—such as finance, law, or personal data processing—regulation will be strictest. Companies that start with governance frameworks now will gain an advantage.

Six factors for automation success

Bain defines six criteria that determine whether a given process is suitable for agentic AI:

1. Verifiability of output: Tasks such as code compilation or resolving a support ticket have clear success metrics.
2. Consequences of failure: Tax returns or legal compliance require closer human supervision.
3. Availability of digitized knowledge: Agents need structured data and documented context, not just information in the heads of experienced employees.
4. Process variability: Repetitive workflows are easier to automate than unique, creative tasks.
5. Integration complexity: Workflows passing through multiple systems and APIs are technically more demanding, but that is where the highest value lies.
6. Political and security boundaries: Agents must operate within clear rules and permissions.

The greatest value is concentrated where no single system controls the entire outcome—that is, in inter-system coordination that currently consumes thousands of hours of human labor.

Recommendations for SaaS companies

Bain advises software companies to start by mapping specific customer workflows, not entire corporate functions. It is also important to assess data quality—whether it is comprehensive, tied to outcomes, and usable for automation.

Companies should close capability gaps through internal development, acquisitions, or partnerships. Examples include AppLovin with its own Axon platform, ServiceNow with Moveworks, or Salesforce with Workday. It is also key to adapt pricing models: from traditional per-user payment to payment for outcome or usage.

Source: AI News — Bain sees US$100 billion SaaS market in agentic AI automation