AI agent value depends on business process outcomes

Business value from AI agents depends on process outcomes

Putting an AI agent into production shows that an organization can deploy it. Establishing business value requires another judgment: whether the process using that agent performs better. Enterprises continue to increase investment and integrate agents into workflows despite a lack of evidence of productivity gains or cost savings, often while the technical foundations for operating them remain inadequate.

As investment grows, most enterprises are moving from experimentation to execution, according to Malik. Survey findings suggest that enterprises recognize the opportunity but struggle to understand everything successful deployment requires. For technology leaders, execution therefore means coordinating, overseeing and assessing the work agents perform together.

This responsibility extends beyond the team deploying an agent because business processes often depend on decisions and systems elsewhere in the organization. An agent can work within a departmental workflow while the wider process remains difficult to manage. Readiness, agent collaboration and monitoring each reveal a different part of this operational challenge.

Readiness, coordination and monitoring expose different gaps

Organizational readiness is the first constraint: only 1 in 5 leaders said their organization was ready to redesign business processes for autonomous operation with AI agents, an August Deloitte study found. Deploying software within an existing process requires fewer changes. Preparing the process to operate autonomously also changes which decisions people must oversee and how work moves between systems.

Yet survey findings show that leaders expect extensive changes to those processes: nearly three-quarters of U.S. company leaders expect roughly half of their business processes to be redesigned or rebuilt around AI agents by 2030. This expectation concerns the future structure of business operations, while the Deloitte finding concerns reported preparedness for redesign. The measures describe different conditions, but both matter when leaders decide how to organize the transition.

Redesign also requires agents to work together, yet advanced agent orchestration, or coordination among multiple agents, remains uncommon. Survey findings put advanced orchestration at just 7% of enterprises, while more than half used basic or workflow-based handoffs. A workflow-based handoff passes work through a defined sequence; multiagent collaboration requires agents to coordinate their contributions. So many enterprises use limited coordination even as leaders anticipate processes built around agents.

Coordination also requires visibility into what those agents do. One in four AI agents operates without monitoring, according to New Relic data. New Relic sells monitoring software, so it benefits commercially from demand for greater visibility into agent activity. Its finding measures individual agents, whereas the orchestration finding measures enterprises; the two measures remain distinct.

Some companies are responding by adopting specialized tools to monitor complex technology environments. These tools help organizations observe agents during operational work, giving people information for oversight. But effective collaboration and improved process performance still need separate assessment, because observing an agent does not establish how well the wider process works.

Cross-functional processes require shared responsibility

The wider process becomes the management challenge when work crosses the boundaries of the team deploying an agent. Neil Ward-Dutton, research VP, agentic automation and AI technologies at IDC, said in a statement that complex business processes commonly span departments and systems. A department can improve its own task while the full process still depends on coordination outside its control.

These dependencies require systems to exchange information and work together, a capability known as interoperability. Without foundations that support work across functions, enterprises will continue to encounter isolation and interoperability problems, Malik said. Separate departmental projects leave the organization responsible for connecting their agents’ decisions and outputs across systems.

Making those connections requires coordinated, purpose-built systems in which agents work together. The design question expands from the task one employee or department can automate to the work the full process requires. Each agent needs a purpose that fits the group’s work and accounts for its relationships with other agents and systems.

Malik links this coordination requirement to governance, the rules and responsibilities for directing and overseeing agent activity. “The next phase of enterprise AI will depend on creating harmony among those agents while maintaining the governance and oversight necessary to build trust and deliver meaningful business value,” Malik said in the report. His recommendation makes oversight part of how agents work together.

The scale of that oversight explains why isolated projects create difficulties. “Businesses must learn to coordinate thousands of decisions, workflows and agents as a single, governed system,” Malik said. Managing activity at that scale requires responsibility to extend across the decisions and workflows agents collectively affect. Organizations relying on isolated productivity tools or departmental projects remain at a disadvantage when work crosses organizational boundaries.

Manage agent portfolios against process outcomes

Managing work across those boundaries requires decisions about agents across the organization. Ward-Dutton recommends a portfolio-level approach to agentic AI, AI that carries out work through agents, and to individual agents. Portfolio management brings agents into a shared investment and oversight discussion, allowing leaders to consider their contribution to processes that span departments.

Ward-Dutton connects that oversight to standards for building and monitoring agents, process priorities and measurement. “By standardizing how agents are built and monitored, prioritizing core processes that cross functions, and measuring results at the outcome level instead of counting deployments, these enterprises will see actual ROI, rather than just a collection of pilots,” Ward-Dutton said. He predicts that this approach will produce a return on investment, or ROI.

Common construction standards give the portfolio a shared basis for development decisions. Departments building agents independently can make choices that suit their immediate tasks while leaving others to resolve coordination problems. Organization-wide standards bring those dependencies into development decisions, helping teams account for agents built elsewhere that must participate in the same process.

Monitoring standards extend this shared responsibility to observing agent activity. One team’s monitoring choices affect the information available to everyone overseeing the wider process. A common approach makes visibility part of portfolio management and gives people responsible for connected work a consistent basis for oversight.

With construction and monitoring addressed together, cross-functional process priorities give investment a business focus. Leaders can consider the relationships between departments and systems alongside each agent’s task. These relationships matter because an agent’s value depends partly on the other work required to complete the process it serves.

Measuring the completed process then connects portfolio management to investment decisions. Deployment counts show delivery capacity, while outcome measurement assesses what changed in business performance. Otherwise, a growing portfolio can become the visible measure of success before its operational effect has been assessed, the risk behind Ward-Dutton’s reference to a collection of pilots.

For each deployment decision, leaders therefore need to define what evidence will demonstrate improvement in the affected process. This requirement belongs in the approval and governance of the agent portfolio, so the organization can assess the operational result and financial return. Better coordination supplies an operating foundation; measuring the changed process determines what the investment achieved.

Key takeaways for decision-makers

• Assess readiness for process redesign: Only one in five leaders reported readiness to redesign processes for autonomous AI agents. Technology leaders need to assess coordination and monitoring gaps before expanding autonomous operations.
• Assign responsibility across departments: Business processes span systems and functions, making agent coordination a shared responsibility. Process owners and technology leaders need common governance rules and clear accountability for agent decisions and handoffs.
• Manage agents against process outcomes: Portfolio oversight brings agent development, monitoring and investment priorities into a shared framework. Approval decisions need defined measures of process improvement and financial return so investment tracks business performance.