Artificial intelligence has evolved from simple automation into systems capable of understanding data, organizing work, interacting with software, and carrying out sequences of tasks with limited human intervention. This development is becoming increasingly relevant to business workflows that involve data processing, application interaction, and repetitive decision-making.
The U.S. Census Bureau reported that 18% of U.S. firms were using AI in at least one business function between November 2025 and January 2026. More than 20% of firms also expected to adopt AI within the following six months, indicating that adoption continues to grow, although not evenly across businesses. [1]
From Automation to Autonomous Workflows
Traditional automation typically relies on predefined rules and structured processes. AI agents, by contrast, can interpret instructions, retrieve information, process data, and respond to changing circumstances.
This makes them particularly useful for workflows that involve several interconnected steps rather than a single repetitive task.
Research from the U.S. Census Bureau shows that 57% of companies using AI applied it across three or fewer operational functions. Among AI-using firms, 52% used it in sales and marketing, 45% in strategy and business development, and 41% in information technology. [1]
This concentration of AI adoption within a limited number of business functions may suggest that many companies are taking a gradual approach to implementation rather than pursuing full-scale automation.
At the same time, the AI agent market is expanding rapidly. According to DataIntelo, the market was valued at $7.8 billion in 2025 and could reach $68.4 billion by 2034, representing a projected compound annual growth rate of 27.4% between 2026 and 2034. [4]
Customer Service Becomes More Efficient
Customer service is one area where AI agents can support tasks such as information retrieval, request classification, document preparation, and handling repetitive inquiries.
According to the U.S. Census Bureau, around 55% of U.S. workers reported using AI for work-related tasks, while 31% of workers who had used AI during the previous week said it saved them between one and two hours of work. [2]
AI agents can connect customer data, knowledge bases, ticketing systems, and internal procedures. An agent can classify a request, search for relevant information, prepare a response, update a support ticket, and escalate unusual cases when necessary.
Human employees can then focus on sensitive complaints, exceptions, approvals, and decisions that require judgment or accountability. This allows companies to automate routine activities while maintaining human oversight where it matters most.
Finance and Operations Gain Multi-Step Automation
Finance and operations often involve structured workflows divided into multiple stages. Invoice processing, expense classification, reconciliation, reporting, inventory monitoring, and exception handling may all require information to move between different systems.
The U.S. Census Bureau found that AI adoption was higher when measured by employment rather than by the number of firms, reflecting greater adoption among larger organizations. [1]
AI agents can review documents, identify inconsistencies, summarize financial information, and route exceptions for approval. In operational environments, they can monitor inventory data, summarize supply-chain events, organize service requests, and notify employees when predefined conditions require attention.
| Workflow area | Indicator | Potential role for AI agents |
|---|---|---|
| Overall AI adoption | 18% of firms | Business-function assistance |
| Sales and marketing | 52% | Research and workflow support |
| Strategy and development | 45% | Analysis and planning |
| IT | 41% | Technical assistance |
| AI use at work | 55% of workers | Task-level assistance |
Source: U.S. Census Bureau. [1][2]
AI Agents Are Changing Software Development
Software development is another area well suited to agent-based workflows because many activities can be evaluated through code execution, testing, logs, and other measurable outputs.
NIST’s 2026 AI Agent Standards Initiative states that AI agents can perform autonomous actions and identifies interoperability, security, agent identity, and authentication as important areas for development. NIST also notes that agents can interact with external systems and internal data, creating additional reliability and security considerations. [5]
AI agents can therefore support connected development activities such as code generation, error analysis, testing, documentation, and issue investigation.
Human review remains important before software changes are deployed, particularly when agents have access to production environments, internal systems, or sensitive data.
AI Agents and the Changing Workforce
The expansion of AI agents does not necessarily mean that entire business workflows will become fully autonomous. A more practical model divides responsibilities between AI systems and employees.
Federal Reserve research found that by the end of 2025, around 18% of U.S. firms had adopted AI within their organizations. A separate worker-focused survey found that work-related use of generative AI had reached approximately 41% in November 2025. These surveys measure different aspects of AI adoption, so their results should not be treated as directly comparable. [3]
The same analysis found that approximately 78% of employees worked for organizations that had adopted AI when adoption was calculated on an employment-weighted basis. This figure differs substantially from the 18% firm-level adoption rate because larger employers account for a greater share of total employment. [3]
This distinction illustrates why AI adoption can appear very different depending on whether it is measured across companies, employees, or individual workplace tasks.
A developing workflow model may therefore divide responsibilities as follows:
- AI agents: information retrieval, classification, summarization, and routine actions.
- Employees: judgment, exception management, approvals, relationship management, and higher-level decision-making.
- Governance systems: access control, monitoring, auditing, and escalation.
Security Becomes Essential
As AI agents gain access to applications, databases, APIs, and other business tools, security becomes a central requirement.
NIST’s work on AI agent security highlights new security challenges created by autonomous agents and indicates that existing cybersecurity practices may need to evolve to address these risks. NIST has also emphasized the importance of identity and authorization controls when agents can access organizational data, applications, and tools. [5][6]
Organizations deploying AI agents should therefore establish controls for identity management, authentication, authorization, monitoring, auditing, data protection, testing, and human approval for critical actions.
These safeguards become particularly important when agents are able to modify records, communicate externally, execute transactions, or interact directly with business systems.
The Future of Connected AI Agents
The next phase of enterprise AI may involve multiple specialized agents working together.
One agent could collect information, another could analyze documents, while another coordinates a technical or operational task.
At the same time, current adoption data shows that organizations are still deploying AI selectively. The concentration of adoption within a limited number of business functions suggests that enterprises are likely to expand their use of AI agents gradually as technical capabilities, security controls, and governance frameworks mature.
Conclusion
AI technologies are reshaping business workflows by bringing together data processing, decision-making, software interaction, and task execution within increasingly connected systems.
Government data indicates that AI adoption continues to grow. Around 18% of organizations had incorporated AI into at least one business function, while work-related use of generative AI among workers reached approximately 41% by November 2025. [1][3]
Customer support, finance, operations, IT, and software development are among the areas where AI agents may provide significant value.
However, successful adoption depends on more than automation alone. Identity management, authentication, security, testing, monitoring, governance, and appropriate human oversight will all play an important role in determining how effectively AI agents can be used in critical business environments.
References
- U.S. Census Bureau — The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks
U.S. Census Bureau Report - U.S. Census Bureau — AI Use at Work
U.S. Census Bureau — AI Use at Work - Federal Reserve — Monitoring AI Adoption in the U.S. Economy
Federal Reserve Report - DataIntelo — AI Agent Platform Market Research
DataIntelo — AI Agent Platform Market Research - NIST — AI Agent Standards Initiative
NIST — AI Agent Standards Initiative - NIST — Summary Analysis of Responses Regarding Security Considerations for AI Agents
NIST — AI Agent Security Analysis