In the RPA era, we used SDD (Solution Design Document) to capture workflows. It was enough to describe step-by-step tasks.
But things have changed: today we don’t just build bots that follow steps. We design agents that can make decisions, understand context, and escalate to humans when necessary.
That’s why we now have ADD.
What’s inside an ADD?
Think of ADD as the “user manual” of an agent:
- Purpose & KPIs: Why the agent exists and how success is measured
- Process Metrics & Story: The current process baseline and the agent’s role
- Functional / Non-Functional Requirements: What the agent does, and which quality/security constraints it must follow
- Prompts & Contexts: Instructions, rules, and business dictionaries guiding the agent
- Guardrails & Escalations: Boundaries and when to involve a human
- Tools & Reporting: Tools the agent uses and the visibility it provides
- Scope & Change Management: What’s included, what’s excluded, and how changes will be handled
Why Does It Matter?
ADD ensures agents are reliable, auditable, and scalable because:
- Clarity & Control: Boundaries are clear from the start.
- Auditability: Versioning and change requests guarantee traceability.
- Risk Reduction: Guardrails and context definitions minimize wrong decisions.
- Measurable Value: KPIs and reports tie automation to business impact.
- Scalability: One well-defined agent can be adapted across multiple processes.
- Human-in-the-Loop: Critical decisions always involve human oversight.
In short: SDD was the foundation of RPA, and ADD is the compass of Agentic Automation.
For detailed breakdowns and examples, check out my full Medium article 
For Template:
Agent Definition Document.docx (121.8 KB)