ANSWER

An agentic workflow is a business process where an AI system doesn’t just suggest work—it can plan, decide, and take actions to move a task toward completion.

In a traditional workflow, software follows explicit rules and humans perform decisions. In an agentic workflow, an AI “agent” can:

• interpret intent (goal),

• break work into steps,

• choose tools/actions,

• execute actions (sometimes across multiple systems),

• and escalate to humans when needed.

Why it matters to Business Analysts

Agentic workflows change what “requirements” must cover. It’s no longer enough to define screens and rules—you must define:

• Decision rights: what the agent is allowed to do without approval

• Boundaries: what it must never do (data, actions, users, systems)

• Escalation: when it must stop and hand off to a human

• Evidence: logs, explanations, and auditability (what happened and why)

Example

A “copilot” might draft a customer response for a human to send.
An agentic workflow might:

1. read the ticket,
2. pull order history,
3. decide if a refund is warranted,
4. issue the refund in the payment system,
5. notify the customer,
6. and log the full decision trail.

What a strong BA delivers

• A workflow map showing human + agent handoffs

• Risk tiers (low/medium/high) tied to required controls

• Acceptance criteria for audit logs, monitoring, and rollback (“kill switch”)