Introducing the ARTist Certification Program for Real-Time Voice AI Builders

[pablosalvador10]

I’d like to share something we’ve been building with the community: the ARTist Certification Program.

ARTist stands for Artist + ART (Azure Real-Time Voice Agent Framework).
It’s a structured certification program for engineers and practitioners who are building real-time voice AI agents using the ART Voice Agent Accelerator.

The goal is simple:

• Recognize real, hands-on expertise (not theory)
• Encourage community contributions and knowledge sharing
• Create a trusted network of builders who know how to run voice AI in production

---

What the Program Covers

ARTist certification validates skills across:

• Real-time voice architecture (ACS, Speech, LLMs)
• Orchestration patterns (Voice Live API vs Azure Speech + custom agents)
• Tool integration and multi-agent flows
• Production deployment, observability, and performance
• Security, compliance, and operational readiness

This is not a quiz-based certification. It’s evidence-driven and focused on real systems that actually run.

---

Certification Levels

Level 1: Apprentice

Foundation level for understanding and running the full end-to-end system.
You should be able to deploy locally, explain the architecture, and complete a live voice call.

Level 2: Creator

For builders extending the framework.
This includes creating custom agents, integrating tools/APIs, and contributing back to the community through issues, docs, or discussions.

Level 3: Maestro

Expert level for production leaders.
This covers Azure deployment, observability, performance tuning, security/compliance, and mentoring others in the community.

Each level has a clear technical checklist and required evidence.

---

How Certification Works

1. Complete the checklist for your target level
2. Prepare a small evidence portfolio (screenshots, repo links, demos, logs)
3. Open a comment in this Discussion titled (see below for an example):
   [ARTist Certification] <Level> - <Your Name>
4. Maintainers review and provide feedback
5. Upon approval, you receive:
   • An official ARTist badge
   • Recognition in the Hall of Fame
   • Community visibility for your work

---

Example: What a Certification Submission Looks Like

Below is an example of the kind of evidence-based post someone would share when submitting for Creator (or higher).
Use this as a template: describe what you built, include a diagram/screenshot, link the YAML/config, and show at least one validated test flow.

---

Example Certification Submission (Template)

Sharing an example of a flow I built using the ART framework as part of my certification submission.

This scenario models a real-world insurance voice workflow with mixed B2C and B2B traffic, focused on security-first entry, agent specialization, and controlled handoffs.

Business Overview

This scenario demonstrates a multi-agent insurance voice system that handles both B2B and B2C callers through intelligent routing and specialized agents.

Business Value

Capability	Business Impact
24/7 Automated Service	Reduce call center costs and handle overflow
B2B Subrogation Hotline	Faster inter-company claim resolution
Policy Self-Service	Customers get instant answers without hold times
FNOL Intake	Structured claim collection with fewer errors
Intelligent Routing	Right agent for the right caller type

Agent Architecture

....

(See Example Certification Submission Comment (Creator / Maestro-style)) in the comments for inspiration

[pablosalvador10]

Example Certification Submission Comment (Creator / Maestro-style)

Sharing an example of a flow I built using the ART framework as part of my certification submission.

This scenario models a real-world insurance voice workflow with mixed B2C and B2B traffic, focused on security-first entry, agent specialization, and controlled handoffs.

Business Overview

This scenario demonstrates a multi-agent insurance voice system that handles both B2B and B2C callers through intelligent routing and specialized agents.

Business Value

Capability	Business Impact
24/7 Automated Service	Reduce call center costs, handle overflow
B2B Subrogation Hotline	Faster inter-company claim resolution
Policy Self-Service	Customers get instant answers without hold times
FNOL Intake	Structured claim collection, fewer errors
Intelligent Routing	Right agent for the right caller type

Agent Architecture

                    ┌───────────────┐
                    │   AuthAgent   │  ← Entry Point (Authentication Gate)
                    └───────┬───────┘
                            │
         ┌──────────────────┼──────────────────┐
         │                  │                  │
         ▼                  ▼                  ▼
   ┌───────────┐      ┌───────────┐      ┌───────────┐
   │  Policy   │ ◄──► │   FNOL    │      │   Subro   │
   │  Advisor  │      │   Agent   │      │   Agent   │
   └───────────┘      └───────────┘      └───────────┘
        │                  │                  │
        └────── B2C ───────┘                  │
         (Policyholders)               B2B (CC Reps)

Agent Roles

Agent	Caller Type	Purpose
AuthAgent	All	Greet, identify caller type, authenticate, route
PolicyAdvisor	B2C	Answer policy questions, coverage inquiries
FNOLAgent	B2C	File new insurance claims (accidents, losses)
SubroAgent	B2B	Handle inter-company subrogation inquiries

🎯 Test Scenarios

⭐ Scenario:

Persona: Jennifer Martinez from Contoso Insurance calling about a subrogation demand. This scenario tests the complete B2B workflow with all 6 key inquiries.

Setup

1. Create demo profile: scenario=insurance, role=cc_rep, test_scenario=golden_path
2. Claim number: CLM-2024-1234
3. Caller: Jennifer Martinez, Contoso Insurance

Complete Workflow Script

#	User Question	Expected Response	Tool
1	"I'm calling about claim CLM-2024-1234"	Asks for company + name	—
2	"Jennifer Martinez, Contoso Insurance"	Verifies CC access, hands off	verify_cc_caller → handoff_subro_agent
3	"Is coverage confirmed for this claim?"	"Coverage is confirmed on this claim."	get_coverage_status
4	"Has liability been accepted? What's the range?"	"Liability has been accepted at 80%."	get_liability_decision
5	"Does the demand exceed policy limits?"	"No limits issue. Your demand ($45,000) is within the $100,000 PD limit."	get_pd_policy_limits
6	"Have any payments been made on the PD feature?"	"1 payment totaling $15,000.00."	get_pd_payments
7	"Has my subrogation demand been received? When will it be assigned?"	"Received on Oct 20th for $45,000. Currently under review by Sarah Johnson."	get_subro_demand_status
8	"Can this be rushed due to attorney involvement or statute concerns?"	Agent MUST ask about each criterion before evaluating	See Rush Flow below

Additions to the Framework

• Agent Registry — Centralized agent store with YAML definitions and prompt templates
• Tool Registry — Pluggable tool system with dependency injection
• Handoff Service — Agent-to-agent transfers with context preservation
• Banking Agents — Concierge, AuthAgent, FraudAgent, InvestmentAdvisor

@JinLee794 @pablosalvador10 for review !

[pablosalvador10]

After review from codeowners you are part of the Level 3 (Maestro) ARTist community...congratulations and we can't wait for more contrbutions.