What Are L1, L2, and L3 Interactions?
These tiers reflect more than just the topic of an interaction, they reflect how much context, judgment, and integration are required to reach a resolution.
Level 1 (L1): Routine and Rule-Based
L1 interactions are straightforward, transactional, and well-documented. They require little to no context from previous conversations and no interpretation of policy nuance.
What defines L1:
- One-step tasks with low variability: things such as hours of operation, address of storefronts, or specific FAQs
- Doesn’t require backend system access (or accesses one system in a simple way)
- Resolution can be fully scripted or template-driven
Why AI works well: These interactions can be fully handled by AI agents using simple RAG. There is value in automating these types of interactions and reducing the load on human support.
Level 2 (L2): Contextual and Connected
L2 interactions require a deeper understanding of the customer’s specific situation and the ability to apply logic or rules dynamically. These conversations often involve multi-turn dialogues and decision-making based on backend data.
What defines L2:
- Requires access to and coordination between backend systems (e.g., CRM, claims, scheduling)
- Involves business rules, policies, or calculations based on customer data
- May require AI to explain an outcome or reason through options
- Often includes some ambiguity or decision paths that branch
Why AI needs more than scripts: These are high-value opportunities for AI Agents to execute workflows, retrieve information, and respond with context. But doing so requires integration and reasoning, not just recognition.
Level 3 (L3): Escalated, Emotional, or Expert-Only
L3 interactions are complex, often emotionally charged, and require human judgment, negotiation, or discretion. They can’t be reliably automated (or shouldn’t be), but they can be enhanced with AI.
What defines L3:
- High stakes or emotional sensitivity
- Involves gray areas of policy, edge cases, or exceptions
- May include complaints, legal issues, or escalations
- Requires human empathy or negotiation
Why AI plays a supporting role: AI shouldn’t attempt to resolve these issues, but it should help by gathering context, summarizing the issue, and guiding the human agent for faster and more empathetic resolution.