Company Context & Policies
Relevant source files
The following files were used as context for generating this wiki page:
The Company Context & Policies system within BrightPath Training serves as the foundational knowledge base and regulatory framework for the multi-agent AI pipeline. It provides the essential business logic, pricing structures, and operational guidelines that agents must consult to ensure responses are accurate, compliant, and aligned with the company’s brand voice.
This system is primarily comprised of file-based memory—specifically company_context.md—which is injected into the system prompt of every agent during a pipeline run. By centralizing policies regarding refunds, booking changes, and pricing, the system ensures consistent decision-making across the Triage, Sales, and Customer Service agents.
Sources: app.py:270-277, orchestrator.py:159-166, README.md:92-100
Organizational Structure & Background
BrightPath Training is a London-based leadership and professional development firm established in 2018. The organization operates with a lean team led by founder Sarah Chen, who serves as the final arbiter for high-value decisions and custom quotes.
Personnel and Roles
| Name | Role | Responsibilities |
|---|---|---|
| Sarah Chen | Founder / Lead Trainer | Strategy, final decision on refunds >£500, custom quotes |
| Marcus Webb | Operations Manager | Administration, bookings, and Telegram responses |
| Priya Kapoor | Sales & Partnerships | Managing client relationships and new business |
| Associate Trainers | Freelance | Delivery of training (James Holloway, Rachel Obi) |
Sources: memory/company_context.md:5-18
Key Client Tiers
The company distinguishes between general customers and Annual Partners (VIPs). Annual Partners are high-value accounts that receive preferential treatment regarding booking flexibility and personal account management.
- Nexus Digital: Annual partner since 2022 (Dan Farrow).
- Hartley Legal: Annual partner since 2023 (Fiona Marsh).
- Westbridge Housing: Annual partner since 2021 (Tony Reid).
Sources: memory/company_context.md:20-25, data/crm.json:62-75
Sales Policies & Pricing
The Sales Ops Agent utilizes a strict pricing and qualification framework to handle inbound leads. Policies dictate that agents should focus on ROI framing and avoid quoting custom prices without a discovery call for large groups.
Product Catalog
| Product | Specification | Pricing |
|---|---|---|
| Public Workshop | Half Day | £295 / person |
| Public Workshop | Full Day | £495 / person |
| Corporate (5-15) | Half / Full Day | £2,200 / £3,800 |
| Corporate (16-30) | Half / Full Day | £3,500 / £5,500 |
| Leadership Coaching | 1:1 Session | £350 (Packages from £1,800) |
| Annual Partnership | Full Year | From £12,000 / year |
Sources: agents/sales_agent.md:14-21
Qualification Logic
Agents categorize leads into three tiers: Hot (budget confirmed, <6 weeks timeline), Warm (interest confirmed, timeline unclear), and Cold (early stage, no budget). If a lead mentions more than 30 people, the policy mandates an automatic suggestion of the Annual Training Partnership.
Sources: agents/sales_agent.md:23-32
Customer Service & Operational Policies
The Customer Service Agent is governed by a set of rigid operational policies regarding financial transactions and scheduling. These rules are applied programmatically by the AI to determine if a "policy_applies" or if an "exception_needed" should be flagged to the manager.
Booking and Refund Rules
The following logic defines the company's stance on flexibility:
This diagram illustrates the decision tree used by agents to apply company policy based on timing and customer status.
Sources: agents/customer_service_agent.md:14-30
Escalation Triggers
Policy dictates specific conditions where the AI must stop automated processing and escalate to Sarah Chen:
- Complaints regarding trainer quality or content.
- Refund requests exceeding £500.
- The third complaint from the same customer (triggers priority account review).
- Hostile sentiment detected in triage (sets urgency to "high").
Sources: agents/customer_service_agent.md:28-32, agents/triage_agent.md:37-41, memory/company_context.md:9-10
Brand Voice and Communication Guidelines
The "Tone Guidelines" ensure that all AI-generated drafts remain consistent with BrightPath's professional identity.
- Warmth: Use a professional but warm and solution-focused tone.
- De-escalation: If a customer is hostile, the agent is instructed to stay calm and acknowledge frustration without matching the tone.
- Personalization: Use first names for VIPs and returning clients.
- Target Response Times: Same day for general enquiries; within 2 hours for complaints.
Sources: agents/customer_service_agent.md:37-42, memory/company_context.md:36-39
Data Architecture and Policy Injection
The system injects company policies into the LLM context through a structured retrieval process.
This sequence shows how the policies stored in Markdown files are loaded by the orchestrator to ground the agent's behavior.
Sources: orchestrator.py:159-166, app.py:270-280
Key Data Structures
The system uses JSON schemas to enforce policy adherence in agent outputs.
{
"policy_applies": "boolean",
"policy_used": "string identifying specific policy section",
"exception_needed": "boolean if manager intervention required",
"is_vip": "boolean based on CRM/Memory lookup"
}
Sources: agents/customer_service_agent.md:47-62
Summary
Company Context & Policies act as the "Source of Truth" for the BrightPath AI Operations system. By defining strict rules for pricing, refunds, and escalations within Markdown-based memory, the system enables autonomous agents to handle standard operations while safely identifying edge cases that require human oversight from the company founder.