Resolutions

How Resolutions Work

ResolvCmd produces resolutions by matching ticket context against your connected documentation. This guide explains the process, what the output looks like, and how to improve resolution quality.

The resolution process

When you submit a query (either through the web app or via a ticketing integration), ResolvCmd follows this pipeline:

  1. Context extraction — The ticket subject, description, and metadata are parsed to understand the issue
  2. Hybrid search — Your connected knowledge sources are searched using both semantic (meaning-based) and keyword matching
  3. Re-ranking — Results are re-ranked with a precision model to ensure the most relevant documents rise to the top
  4. Resolution generation — The matched documentation is used to produce structured, step-by-step resolution with source citations
  5. Safety checks — The resolution is checked for completeness, and a confidence level is assigned

Resolution modes

Fast Resolution

  • Uses the top-matching document
  • Produces a concise resolution quickly
  • Best for common, well-documented issues
  • Uses 1 credit

Detailed Resolution

  • Synthesizes information from multiple matching documents
  • Produces a more thorough resolution with additional context
  • Best for complex issues or when you need comprehensive steps
  • Uses 1.5 credits

Enhanced Accuracy

  • Uses a premium model for complex or ambiguous queries
  • Available on Starter plans and above
  • Toggle it on per-resolution when you need higher precision
  • Uses 2x standard credits

Understanding the output

Every resolution includes:

  • Numbered steps — Each step is a specific action. Steps are ordered logically, not just listed.
  • Source citations — Each step shows which document and section it came from. Click the citation to view the original source.
  • Confidence level — HIGH, MEDIUM, or LOW based on how well your documentation covers the topic
  • Notes — Additional context when relevant, such as warnings or alternative approaches

Confidence levels

LevelMeaning
HIGHStrong documentation match. Multiple relevant sources confirm the resolution.
MEDIUMPartial match. Documentation covers the topic but may not address the exact scenario.
LOWWeak match. Limited documentation available. Resolution may require verification.

A LOW confidence resolution isn’t necessarily wrong — it means your documentation has limited coverage of that topic. This is valuable signal for Knowledge Health.

Follow-up actions

After receiving a resolution, you can:

  • Expand resolution — Get additional detail or alternative approaches (0.5 credits)
  • Generate customer reply — Draft a customer-facing response based on the resolution (0.5 credits)
  • Add validation steps — Generate verification steps to confirm the resolution worked (0.5 credits)

Improving resolution quality

Resolution quality depends directly on your documentation:

  • Be specific in your SOPs — Step-by-step procedures produce better resolutions than general overviews
  • Use consistent terminology — Set up your organization dictionary to map internal terms
  • Review Knowledge Health — It shows which topics produce weak resolutions and which docs need improvement
  • Use client-specific documentation — Tag documentation by client or environment for more targeted resolutions
  • Provide feedback — Rate resolutions and add comments. This data feeds into Knowledge Health analysis.