Artificial intelligence is moving beyond helping people perform tasks. Agentic AI introduces systems that can make decisions, invoke other systems, access data and carry out work on behalf of an organisation.
That changes the service design challenge considerably.
AI is becoming an operational actor
Traditional technology has generally waited for a person to interact with it. An AI agent can become an active participant in a business service. It may have its own identity, permissions, tools and relationships with other systems or agents, while performing actions without somebody approving every step.
The question is therefore no longer simply whether the technology works.
Organisations need to understand what an agent is permitted to do, whose authority it is exercising, which systems and data it can access, how its activity can be traced and who remains accountable for the outcome.
This is an operating-model challenge
An organisation might successfully build an agent capable of approving refunds, raising purchase orders, resetting user accounts or modifying cloud infrastructure. Technically, the capability may work perfectly. Operationally, considerably more needs to be designed.
Someone must determine the limits of its authority. Human intervention points, ownership and escalation arrangements need to be clear. Monitoring and audit information must make abnormal behaviour visible and allow activity to be reconstructed. The organisation must also be able to stop the agent when necessary.
Traceability and accountability become service concerns
These are familiar Service Architecture and Service Design principles. What changes is the nature of the participants within the service.
A future service blueprint may no longer simply show customers, employees, applications, suppliers and processes. It may include human users, AI agents, agent identities, models, APIs, business applications, external providers and other autonomous agents, all contributing to one business outcome.
As those relationships become more complex, understanding ownership, dependencies, decision rights and failure modes becomes more important.
Agent sprawl could become the next service-account problem
Organisations may eventually have hundreds or thousands of autonomous identities. Without clear ownership and lifecycle governance, agents could become the next generation of unmanaged service accounts: operationally important, but poorly understood.
That creates questions about who owns each agent, what service it supports, which systems it can access, what authority it has, how changes are controlled and when permissions should be reviewed or removed.
AI makes documentation easier, but design judgement more valuable
Artificial intelligence will make documentation, process maps, support procedures and governance artefacts easier to produce. That may reduce the value of Service Design when it is viewed simply as document production.
But the value of good design has never really been the document. It lies in making the decisions behind it.
How should the service operate? Who owns the outcome? Where should authority sit? What level of autonomy is acceptable? What happens when something fails? How do suppliers, people, agents and technology work together? What evidence is required to demonstrate control?
As AI becomes an operational participant rather than simply another technology component, these questions become increasingly important.
The organisations that gain the greatest value from agentic AI are unlikely to be those that simply deploy the most agents. They will be the organisations that deliberately design the operating environment around them.