The terms that define modern, AI-native customer support. Written for the people building it.
Core Metrics
The numbers that tell you whether your AI agent is actually working.
- What is Resolution rate? Resolution rate The share of conversations an AI agent fully closes on its own. The most honest measure of whether your AI actually solves problems.
- What is Involvement rate? Involvement rate How much of your inbound support volume the AI agent touches, whatever the outcome. A measure of footprint.
- What is Automation rate? Automation rate Resolution rate times involvement rate. The single number that ties AI impact to cost and capacity.
- What is Deflection rate? Deflection rate Share of support requests resolved without a human. Useful when measured honestly; misleading when silence counts as success.
- What is Containment rate? Containment rate The share of conversations kept inside the automated experience. Directional at best, misleading as a success measure.
- What is CSAT (Customer Satisfaction Score)? CSAT (Customer Satisfaction Score) A survey score for how satisfied customers say they are. Useful but partial, covering a sliver of conversations.
- What is AHT (Average Handle Time)? AHT (Average Handle Time) The average time a human spends on a conversation. A legacy metric that rises, by design, once AI takes the easy work.
- What is FCR (First Contact Resolution)? FCR (First Contact Resolution) Issues resolved in a single interaction. Loses meaning once AI handles the frontline.
- What is Cost per resolution? Cost per resolution The full cost of actually solving one conversation. A truer lens than raw unit cost.
Economics
How the math of support changes when AI carries the volume.
- What is Fully loaded cost per conversation? Fully loaded cost per conversation The real cost of a human-handled conversation, beyond salary. Most teams underestimate it.
- What is Break-even threshold? Break-even threshold The automation rate at which an AI agent's value exceeds its total cost. Every resolution past it is high yield.
- What is total cost of ownership (TCO)? Total cost of ownership (TCO) The complete lifetime cost of running an AI agent, not just the per-resolution price. What real ROI hinges on.
- What is Budget reallocation? Budget reallocation Realigning support spend to an AI-first model instead of cutting it. From cost savings to redeployment.
- What is Support cost decoupling? Support cost decoupling Breaking the old link between business growth and support headcount growth.
Roles and Organization
The new shape of a support team built around AI.
- What is AI operations lead? AI operations lead The owner of day-to-day AI agent performance: quality, tuning, and iteration.
- What is Knowledge manager? Knowledge manager The owner of the content an AI agent depends on. An agent is only as good as its inputs.
- What is Conversation designer? Conversation designer The owner of how the AI agent communicates: voice, structure, and handoff logic.
- What is Support automation specialist? Support automation specialist The builder of the workflows and actions an AI agent can execute, not just answer.
- What is Player-coach model? Player-coach model A leadership model for AI-first support: part strategist, part hands-on operator.
- What is AI-first organization? AI-first organization A support org built around AI as the core of the operation, not an add-on.
- What is Support as a product? Support as a product Running support with the rigor a product team brings to its core product.
Experience and Architecture
How an AI-native support experience is designed to feel.
- What is AI as first responder? AI as first responder Making the AI agent the default entry point for every conversation, and the end point for most.
- What is Hybrid by design? Hybrid by design AI and humans working as one system from the start, not stitched together after.
- What is Invisible handoff? Invisible handoff A transition between AI and human the customer never notices. Context carries over intact.
- What is Proactive support? Proactive support Using AI to anticipate needs and offer help before a problem is raised.
- What is AI copilot? AI copilot AI that makes human agents faster: summaries, categorization, suggested answers.
- What is Content as infrastructure? Content as infrastructure Treating knowledge like critical infrastructure, owned, structured, and versioned.
- What is Ingestion-ready content? Ingestion-ready content Knowledge structured so an AI agent can reliably consume and use it.
- What is Knowledge base? Knowledge base The repository of articles and docs that fuels an AI agent's answers.
Process and Evaluation
How to test, govern, and keep improving an AI agent.
- What is Continuous improvement loop? Continuous improvement loop A structured cycle: train, test, deploy, analyze, repeat. Systems that learn by default.
- What is Lightweight governance? Lightweight governance A low-friction way to keep iterating on an AI agent without bureaucracy.
- What is a proof of concept (POC)? Proof of concept (POC) A focused evaluation of one strongest-fit agent in a realistic setting, not a multi-vendor bake-off.
- What is sandbox testing vs. live testing? Sandbox testing vs. live testing Two environments for evaluating an agent: controlled sandbox and real production traffic.
- What is Human involvement rate? Human involvement rate How often humans step in, and whether those moments are genuinely high value.
Vendors and Alternatives
Who else builds AI support agents, and where SupportWire fits.
35 terms and counting. Updated June 2026.
Know the words. Run the playbook.
An AI agent that earns these numbers.
Resolution rate over deflection. Hybrid by design. SupportWire is built around the model this dossier describes.