What happens when some of enterprise software’s most productive users never log in?
AI agents do not need dashboards. They need context, permission and a clear path to act. They can classify requests, check entitlements, initiate transactions in other enterprise applications, handle exceptions and complete workflows across different systems. As AI agents take on more work, user seats remain useful for measuring human access, but enterprises also need to measure the work performed by agents and the value that work creates.
For much of the software-as-a-service era, the seat bundled access, usage and value into one convenient commercial unit. More users meant more licenses and enterprises saw adoption as proof they were getting value from the platform. Agentic AI separates these measures. Now, enterprises need to know who or what is allowed to act, what work was done, how much it cost and what business results can be confirmed. I call this the agentic value ledger.
When software begins to perform the work
The first wave of enterprise generative AI was largely assistive. Copilots helped employees create, retrieve and summarize information faster. The person, however, still moved the process from one stage to the next.
Agents change the model when they begin to complete bounded portions of work. Consider an equipment failure reported by a customer. An AI agent could open the case, identify the affected asset, validate service entitlement, check parts availability, propose an appointment and route a warranty exception for approval. The process may touch customer service, asset management, inventory, field operations, warranty management and finance, even though the customer made only one request.
The basic unit of value is no longer just one person using one app. Now, it is a series of actions by both people and AI agents working together across a value stream. This changes how value should be measured. Daily active users do not show whether equipment was repaired faster or whether the process cost less to execute. Enterprises need to track completion rates, cycle times, exceptions, rework, execution costs and customer or operational outcomes.
Our 2026 report, The Blueprint for AI Leadership, found that many organizations measure AI return on investment primarily through process speed and efficiency improvements (28%) or cost savings and productivity gains (23%). Only 18% reported AI significantly impacting revenue generation. This reinforces the need to connect AI measurement to broader business outcomes.
Why ServiceNow has a distinctive position
Major enterprise platforms are adding AI agents. Differentiation will depend on whether a platform can give agents the right context, authority to finish work and enough control for the results to be trusted.
ServiceNow is well positioned because its platform already orchestrates workflows across the enterprise. Requests, incidents, cases, approvals, service relationships, assets, knowledge and many workflows pass through the platform every day. Together, they provide much of the context required to determine what should happen next.
ServiceNow will operate alongside other systems of record across the enterprise. Customer information may sit in Salesforce. Employee records may remain in Workday. Financial and supply chain transactions may be executed in enterprise resource planning (ERP) platforms. Specialized applications will continue to hold industry-specific logic.
ServiceNow’s larger opportunity is to become the governed system of action across these environments. This means coordinating work started by both people and AI, while preserving the authority of the underlying systems of record. It also means using the right identities, permissions, policies and approval steps when an agent goes from making a suggestion to taking action.
This matters because enterprise value streams rarely stay inside one platform. A customer issue may begin in customer relationship management system, require operational context from ServiceNow, invoke a transaction in ERP and involve an employee approval before it can be resolved. Companies need a controlled way for work to move across their technology systems, supported by shared context and clear operating boundaries.
The commercial model will be hybrid
As AI agents take on more work, software pricing will not simply move from seats to pure outcome-based models.
Seats will still be useful for human access and keeping budgets predictable. Models based on usage or actions work better for changing levels of machine work. Outcome-based pricing can work when the result is tangible and the software platform’s impact can be measured with confidence. In most business processes, attributing the outcome is more complicated.
Customer retention may depend on product quality, pricing, service and account management. An avoided outage may result from monitoring, infrastructure, engineering decisions and automated remediation working together. When several systems and teams contribute to the result, no single platform can claim the full outcome.
The most likely approach is a mix of user access, platform tiers, transactions, AI usage and selected outcome-linked components. What matters is that enterprises can connect that consumption to operating evidence. For each agentic workflow, leaders should know the starting point, cost, exception rate, quality standards and business results. Without that connection, AI consumption becomes another technology expense rather than a measure of business performance.
The hard part is delegating authority
The central challenge is deciding how much authority an agent should have.
The same study found that only 17% of respondents had high confidence in agent-initiated actions. As AI moves from recommendation to execution, confidence will depend less on what a model can demonstrate and more on the controls governing what it is permitted to do.
Some tasks are low risk, happen often and can be easily undone, making them suitable for agents to handle on their own. Decisions involving payments, regulated data, employee matters, security or material customer commitments may still require explicit human approval.
The objective is the highest level of delegated execution that the enterprise can defend. This also requires organizations to address the quality of the underlying process. Agents do not remove process debt. They execute it faster.
If knowledge is poor, agents can give wrong answers more quickly. Too much variation leads to more exceptions. Weak data can cause decisions based on incomplete information. Fragile integrations might let an agent finish one step but leave the rest of the process broken. So, agentic AI needs simpler processes, trusted context, clear ownership and strong controls for identity, permissions, escalation and rollback.
Designing for accountable execution
The hardest work rarely happens on just one platform.
A single value stream may involve ServiceNow, Salesforce, Workday, ERP applications and industry-specific systems. Leaders must decide which system owns the record, which platform governs the workflow, which agent is authorized to act and where human accountability must remain. This is a challenge we see repeatedly across enterprise application environments. No single platform needs to control every layer. The priority is to design how the systems work together around a defined business outcome. That requires simplifying the process, clarifying platform boundaries, improving data and knowledge readiness, governing agent activity and measuring value across the full value stream.
Recent product developments in agent orchestration and governance suggest ServiceNow is positioning itself for this role, focusing on coordinating work, agents and decisions across systems. If that approach succeeds, it could strengthen ServiceNow's position as a governed execution layer for complex enterprise value streams.
The real test is whether agents can act across platforms with clear authority, reliable context and measurable results.
The economics of enterprise software will increasingly depend on whether organizations can connect agent activity to accountable, measurable value.
