Role-Based AI Execution Using Breeze Agents + Copilot Prompt Standards
Use Breeze Studio Agents for Customer & Content Use Cases, and Standardized Copilot Prompts for Sales Execution
When to Use Each Breeze Capability
| Use Case | Tool |
|---|---|
| Website chat, qualification, support | Breeze Studio Agent |
| Content generation | Breeze Studio Agent |
| Deal review, email drafting, next steps | Breeze Copilot |
| Internal sales execution | Copilot + prompt standards |
Requirements
-
Breeze AI enabled
-
Sales, Marketing, or Service Hub Professional+
-
Access to Breeze Studio
-
Defined CRM properties and pipelines
PART 1: Build a Breeze Agent (Persistent, Configured)
Example: AI Customer Qualification Agent
Step 1: Define the Agent’s Role
Agent purpose
-
Qualify inbound leads
-
Ask structured discovery questions
-
Route qualified leads to sales
-
Enforce ICP rules
Agent constraints
-
No pricing promises
-
No contract discussion
-
Route to human when uncertain
Step 2: Create the Agent in Breeze Studio
Path: Breeze Studio → Agents → Create agent
Agent type
-
Customer Agent
Data sources
-
Website pages
-
Knowledge base
-
Approved sales documentation
Rules
-
Ask one question at a time
-
Capture:
-
Company size
-
Use case
-
Timeline
-
-
Stop after qualification criteria met
Step 3: Connect the Agent to CRM
Configure outcomes:
-
Create or update contact
-
Set lifecycle stage
-
Populate qualification properties
-
Assign to sales queue if qualified
This agent is now:
-
Persistent
-
Governed
-
Auditable
PART 2: Create Sales Execution “AI Roles” Using Copilot (Fully Supported)
Copilot does not support saved agents, so we standardise prompts, not assistants.
Step 4: Define a Sales AI Role (Example)
Role: Sales Deal Review Assistant
Responsibilities
-
Summarize deal status
-
Identify risks
-
Recommend next action
-
Draft follow-up emails
This role exists as a prompt standard, not a configuration.
Step 5: Standardised Copilot Prompts (Use Exactly)
Deal Review Prompt
You are a senior sales operations assistant.
Using this deal, associated contacts, companies, and activity history:
1. Summarize the current deal status in 3 bullet points
2. Identify any risks, gaps, or missing information
3. Recommend the single highest-impact next action
Be concise, factual, and specific.
Follow-Up Email Prompt
Draft a professional follow-up email to the primary contact.
Requirements:
- Reference the most recent interaction
- Acknowledge any open questions
- Propose a clear next step
- Keep under 120 words
- Neutral, consultative tone
These prompts:
-
Work from any deal record
-
Produce consistent outputs
-
Are repeatable across reps
Step 6: Enforce Usage with Sales Process (Critical)
To make this operational:
-
Add prompts to:
-
Sales playbooks
-
Internal SOPs
-
-
Train reps to:
-
Use Copilot only with approved prompts
-
-
Managers review:
-
Output quality
-
Adherence to recommendations
-
This is how you get agent-like consistency without agents.
PART 3: Chain Agent → CRM → Copilot (Advanced Pattern)
-
Customer Agent qualifies lead
-
Properties updated in CRM
-
Deal created or assigned
-
Rep opens deal
-
Uses Copilot with standardised prompts
-
Actions taken consistently
This creates a closed-loop AI system without unsupported automation.
What This Hack Deliberately Avoids
-
Autonomous sales agents
-
Pipeline-scoped AI execution
-
AI-triggered workflow actions
-
Non-auditable recommendations
Everything here is supported, governed, and safe.
Common Mistakes to Avoid
-
Treating Copilot like a chatbot
-
Allowing free-form prompts
-
Expecting agents to run inside deals
-
Blending customer agents with sales execution
