AI Agent Workflow Design for [Your Business Process]

Role: You are an AI automation architect, skilled in designing robust and scalable AI agent workflows.

Objective: Design a detailed AI agent-driven automation workflow for a specified business process, including pre-conditions, agent tasks, decision points, human intervention points, and governance considerations. The goal is to maximize efficiency while maintaining control and auditability.

Required Inputs:
1. [Your Business Process]: Describe the specific business process you want to automate (e.g., 'customer support ticket routing', 'contract review and redlining', 'inventory reorder approval'). Be as detailed as possible about its current steps, stakeholders, and pain points.
2. [Key Objectives for Automation]: What are the primary goals of automating this process? (e.g., 'reduce processing time by 30%', 'improve accuracy by 15%', 'reduce manual effort by 50%').
3. [Existing Systems/Tools]: List any critical systems or tools currently involved in this process (e.g., 'CRM: Salesforce', 'ERP: SAP', 'Document Management: SharePoint', 'Communication: Slack').
4. [Compliance/Regulatory Requirements]: Are there any specific compliance, legal, or regulatory requirements that must be adhered to? (e.g., 'GDPR compliance', 'HIPAA', 'financial audit trails').
5. [Risk Tolerance]: Describe the organization's risk tolerance for automation errors or unexpected outcomes in this specific process (e.g., 'low - errors are costly', 'medium - some errors are tolerable', 'high - experimental phase').

Execution Steps:
1. **Process Decomposition:** Break down the [Your Business Process] into discrete, sequential steps. Identify which steps are rule-based, data-intensive, or require subjective judgment.
2. **Agent Task Identification:** For each step, determine if an AI agent can perform it. If so, define the specific task the agent would execute (e.g., 'classify incoming email', 'extract key terms from contract', 'compare data points across systems'). If not, identify it as a human intervention point.
3. **Workflow Mapping:** Create a logical flow diagram (described in text) outlining the sequence of agent tasks and human touchpoints. Include conditional logic and decision branches (e.g., 'IF [condition], THEN [agent task A], ELSE [agent task B] or [human review]').
4. **Data Flow and Integration:** Describe the data inputs required for each agent task and the outputs generated. Specify how the agent would interact with [Existing Systems/Tools] (e.g., 'API call to CRM to update status', 'read document from SharePoint').
5. **Human-in-the-Loop Strategy:** For identified human intervention points, describe the trigger for human involvement (e.g., 'agent confidence score below threshold', 'complex case detected', 'final approval required'). Detail what information the human needs to make a decision and what actions they can take.
6. **Governance and Auditability:** Propose mechanisms for monitoring agent performance, tracking decisions, and ensuring compliance. This should include audit trails, logging of agent actions, and approval gates before critical actions are committed.
7. **Error Handling and Resilience:** Outline how the workflow would handle common errors or exceptions (e.g., 'API failure', 'unstructured data not parsed correctly', 'agent output uncertainty'). Specify fallback mechanisms or human escalation paths.
8. **Pilot and Rollout Strategy:** Suggest a phased implementation approach, including a pilot phase on non-critical cases, verification steps, and criteria for full deployment.

Output Contract:
Provide a structured response in the following sections:
1. **Automated Process Title:** A concise title for the new automated workflow.
2. **Workflow Overview:** A high-level summary of the agent's role and the process flow.
3. **Detailed Workflow Steps:** A numbered list describing each step, clearly indicating whether it's an 'AI Agent Task' or 'Human Intervention Point'. For agent tasks, specify inputs, actions, and outputs. For human points, specify triggers and required human actions.
4. **Integration Points:** A list of [Existing Systems/Tools] and how the AI agent would interact with each.
5. **Governance Framework:** A description of audit trails, monitoring, and approval mechanisms.
6. **Error Handling & Fallbacks:** A plan for managing exceptions and errors.
7. **Pilot & Verification Plan:** Steps for a controlled rollout.

Constraints:
- Focus on a single, well-defined business process from [Your Business Process].
- Assume the availability of general AI agent capabilities (e.g., natural language understanding, data extraction, API interaction).
- Prioritize clear communication and practical implementation over theoretical AI concepts.

Quality Check: Ensure the proposed workflow is logical, addresses all specified inputs, and provides clear guidance on how to implement and govern the AI agent. The design should be robust enough to handle exceptions and incorporate human oversight where necessary, aligning with the specified [Risk Tolerance] and [Compliance/Regulatory Requirements].