Skip to main content

​
How Aura AI Works

Aura AI operates by continuously learning and drawing information from various sources to provide accurate and relevant replies during a chat session.

​
1. Knowledge & Context Sources

Aura AI’s intelligence is fueled by the data you provide. It leverages multiple sources to understand and respond to customer queries:
Aura learns from the ongoing conversation, retaining context from past interactions within the current session. This allows it to personalize responses and avoid asking redundant questions.
Define direct question-answer examples for common queries. This provides a precise and reliable way for Aura to respond to frequently asked questions.
Aura integrates key details from your Emplorium workspace settings, such as your business name, timezone, general contact rules, and language preferences, to provide contextually accurate information.
Aura can access all data fields collected via forms in Flows or populated manually by agents. This allows it to reference specific customer information (e.g., “What’s my order status?”) if integrated with external systems.
Configure Aura to run external API queries when specific keywords or intents are detected (e.g., “order status,” “shipping update”). This allows Aura to fetch real-time information from your backend systems.
Learn more about setting up these API queries in Data Requests GET
Aura can be configured to trigger API calls to update external records or systems. These can be initiated by user actions or Aura’s interpretation of the conversation (e.g., “update my contact info”)
Learn more about setting up these API updates in Data Updates POST/PUT

​
2. Aura AI in Action During a Chat

When Aura is configured through a Flow** **it’ll be assigned to the visitor every time the flow condition is met. For exmaple:
1

User Message Received (Flow Condition)

A user sends a message through the Emplorium Chat Widget.
2

Aura Matches Intent / Keyword

Aura processes the message to understand the user’s intent or detect specific keywords:
  • If a direct match is found in your Q&A pairs, Aura provides the defined answer_._
  • If a match is found with a keyword-based data request, Aura fetches the external information (e.g., order details).
  • If the query is ambiguous, unknown, or not a direct match, Aura applies its fallback logic.
3

Contextual Memory

Throughout the session, Aura retains context from previously submitted fields, referenced articles, and the message trail, ensuring continuity in the conversation.
4

Escalation or Follow-Up

If Aura cannot confidently answer after multiple attempts or if specific escalation rules are met, it will tag the message with #fallback and follow configured escalation rules.

​
Configuring Aura AI

You can personalize Aura AI’s behavior, tone, and escalation rules to align with your brand and support strategy.

​
General Settings & Persona

  • Tone & Style: Define Aura’s conversational tone (e.g., formal, friendly, empathetic).
  • Guardrails: Set boundaries and content restrictions to ensure Aura’s responses remain appropriate and on-brand.
  • Follow-ups: Configure Aura to ask follow-up questions to clarify intent or gather more information.
  • Memory: Ensure Aura maintains context throughout the conversation to provide more personalized and relevant responses.
  • Adherence to Content Restrictions & Safe Mode: Aura is designed to respect defined content boundaries and can operate in a “safe mode” to minimize unapproved responses.

​
Fallback & Escalation Handling

Crucially, Aura AI includes robust mechanisms for when it cannot confidently respond, ensuring customers are never left without assistance.
1

Unanswered Count Tracking

You can configure Aura to track the number of #fallback-tagged messages that occur within a single session.
2

Agent Escalation

After a configurable number (X) of fallbacks, Aura can be set to automatically prompt the user with options like: _“That didn’t help? Let me connect you with a person.” _ “Would you like to talk to a support agent?” This can seamlessly integrate with the Escalate to Agent block in your Flows.
Aura Configuration Pn

​
The Aura AI Playground

The Playground tab in Aura AI configuration (or a dedicated Playground app) allows your team to test Aura’s responses and behavior in a sandbox environment before deploying it live in a Flow or directly to your widget.
  • Test Responses: Input queries and see how Aura replies based on its current knowledge base.
  • Debug Data Requests: Verify that Aura correctly triggers and processes external API calls.
  • Simulate Scenarios: Test different conversation paths and escalation scenarios.
Playground Default State Pn

​
Key Benefits of Aura AI

  • 24/7 Availability: Provide instant support around the clock, even outside business hours.
  • Reduced Agent Workload: Automate responses to common questions, freeing up human agents for more complex issues.
  • Consistent Information: Ensure customers receive accurate and consistent answers every time.
  • Faster Resolution Times: Many queries can be resolved instantly by Aura, leading to higher customer satisfaction.
  • Scalability: Effortlessly handle increased conversation volumes without needing to proportionally increase your human support team.
  • Data-Driven Improvement: Aura continuously learns from interactions, and its fallback metrics provide valuable insights for refining your knowledge base.

​
Best Practices for Deploying Aura AI

Maximize Aura AI’s effectiveness with these strategic approaches:
1

Start with High-Volume FAQs

Begin by training Aura on your most frequently asked questions. This provides immediate value and reduces agent burden.
2

Provide Diverse Knowledge Sources

Utilize a mix of Q&A pairs for precision, and links/articles for broader contextual understanding.
3

Clearly Define Fallback Rules

Ensure a clear path to a human agent when Aura can’t help. A frustrated customer is worse than no automation.
4

Iterate & Monitor

Regularly review Aura’s performance, especially its #fallback rates. Use these insights to identify gaps in its knowledge and continuously refine its training data.
5

Test in Playground Religiously

Before pushing any major changes to Aura’s knowledge base or integrations, thoroughly test in the Playground to prevent unexpected behavior in live chat.
6

Educate Your Agents

Train your human agents on how Aura works, when it escalates, and how they can leverage its assistance (e.g., using suggested replies).

​
Next Steps

⌘I