The Sales Agent

Relevant source files

The following files were used as context for generating this wiki page:

• agents/sales_agent.md
• agents/triage_agent.md
• app.py
• orchestrator.py
• memory_utils.py
• README.md

The Sales Agent, also referred to as the Sales Ops Agent, is a specialized AI component within the BrightPath AI Operations System. Its primary purpose is to qualify inbound leads, identify upsell opportunities, and manage the sales pipeline for BrightPath Training. It operates as the second stage in a three-agent pipeline, receiving processed data from the Triage Agent to provide high-context, professional sales responses and CRM updates.

Sources: README.md:5-11, agents/sales_agent.md:4-6

As a critical part of the multi-agent system, the Sales Agent interacts with file-based memory and the company CRM to ensure that every lead is handled according to existing contact history and corporate policies. This ensures that the business owner receives actionable summaries and drafted responses that are consistent with the company’s brand and pricing tiers.

Sources: README.md:14-17, agents/sales_agent.md:9-14

Architecture and Role

The Sales Agent is invoked by the orchestrator.py or app.py when the Triage Agent classifies an inbound message category as sales or complex. It functions as a specialist agent that consumes a comprehensive system prompt containing its role definition, current memory state, CRM data, and the specific triage notes generated in the previous step.

Sources: orchestrator.py:202-212, app.py:417-419, agents/triage_agent.md:14-19

Integration Flow

The following diagram illustrates how the Sales Agent fits into the broader message processing pipeline.

Core Responsibilities and Logic

The Sales Agent is governed by a specific set of rules and qualification tiers that dictate how it handles different types of leads.

Lead Qualification Tiers

The agent categorizes leads based on budget, decision-making authority, and timeline.

Tier	Description
Hot	Budget confirmed, decision maker identified, timeline within 6 weeks.
Warm	Interest confirmed, timeline unclear, requires nurturing.
Cold	Early stage, no budget confirmed, long timeline.

Sources: agents/sales_agent.md:21-25

Operational Rules

• Pre-response Check: Must check sales_notes.md for prior contact history before drafting any response.
• Pricing Restrictions: Never quote custom pricing; always offer a discovery call for large requests.
• Threshold Upselling: Suggest an "Annual Training Partnership" if the lead mentions more than 30 people.
• Competitor Handling: Acknowledge competitors briefly but emphasize BrightPath's unique value.
• Re-engagement: Treat any lead with no contact for 14+ days as a re-engagement case.

Sources: agents/sales_agent.md:27-33

Technical Implementation

Execution Interface

In the Python implementation, the Sales Agent is executed via the run_sales_agent function. This function constructs a system prompt by concatenating the agent's markdown definition, memory files, CRM data, and triage output.

def run_sales_agent(message, triage, memory):
    system = f"{load_agent('sales_agent')}\n\n--- MEMORY ---\n{memory}\n\n--- CRM ---\n{load_crm()}\n\n--- TRIAGE ---\n{json.dumps(triage, indent=2)}"
    return call_claude(system, f"Original message:\n\n{message}")

Sources: app.py:417-419, orchestrator.py:171-182

Data Structures (Output Format)

The Sales Agent returns a structured JSON object. This structure is used by the orchestrator to update the database, modify memory files, and inform the Manager Briefing Agent.

Field	Type	Description
lead_tier	String	hot, warm, or cold.
estimated_value	String	low, medium, high, or very_high.
draft_response	String	Full professional email response for the customer.
crm_update	Object	Contains status, notes, next_action, and deadline.

Sources: agents/sales_agent.md:37-53

Sequence of Operation

The following sequence diagram details the interaction between the Orchestrator, the Sales Agent, and the persistent storage layers during a sales request.

Memory and State Persistence

The Sales Agent is "stateful" across sessions because it reads from and writes to markdown files in the memory/ directory.

• sales_notes.md: Stores pipeline activity, competitor awareness, and specific activity logs (e.g., Sarah Miles board review dates).
• customer_history.md: Maintains per-customer notes, past workshop attendance, and sentiment history.
• open_tasks.md: Records outstanding actions identified by the agent, such as "Follow up if no reply in 5 days".

Sources: README.md:65-71, memory_utils.py:65-71, memory/sales_notes.md:1-12

Automated Updates

When a sales scenario completes, the system calls run_memory_update. For example, in Scenario 14 (Rachel Drummond re-engagement), the system automatically updates the status from "Cold" to "Warm" in sales_notes.md and adds a high-priority task to the database.

Sources: memory_utils.py:145-167, app.py:350-358

Conclusion

The Sales Agent is the primary driver for revenue-related automation in the BrightPath system. By combining rigid business rules (like pricing tiers and person-count thresholds) with LLM-driven empathy and historical context, it automates lead qualification while ensuring that the business manager remains informed of high-value opportunities through the subsequent briefing stage.

Sources: agents/sales_agent.md:1-15, README.md:1-5