Skip to main content
MIRA (Missing Info Recovery Agent) is Cendra’s AI-powered knowledge discovery system. It solves a critical problem in hospitality AI: when a guest asks a question the AI can’t answer, property managers often reply manually — but that answer never becomes reusable knowledge. The same gap repeats endlessly. MIRA breaks this cycle. It monitors every conversation, detects when managers manually provide answers to questions the AI missed, extracts those answers, and suggests them as knowledge base additions — turning one-time manual replies into permanent AI knowledge.

​
How MIRA Works

  1. Guest asks a question the AI can’t fully answer from the knowledge base
  2. MIRA detects the gap and creates a knowledge candidate
  3. MIRA suggests an answer based on past conversations, property data, and context
  4. You review and approve — one click adds it to your knowledge base
  5. AI improves — next time a guest asks the same question, the AI has the answer

​
MIRA’s Intelligence

MIRA doesn’t just surface unanswered questions — it learns from your team’s behavior:
  • Scans manual replies — when your team answers a question the AI missed, MIRA detects the answer in the conversation
  • Extracts knowledge — identifies the answer and formats it as a knowledge base article
  • Classifies information — distinguishes between stable property knowledge (WiFi password, parking instructions) and temporary booking-specific information
  • Detects conflicts — checks suggested knowledge against existing articles to avoid contradictions
  • Compounds over time — each approved candidate makes your AI smarter, reducing the next cycle of missed questions
The result: your AI’s accuracy improves continuously from your team’s natural workflow — without anyone maintaining dashboards or writing documentation.

​
Knowledge Candidate Fields

Each candidate MIRA surfaces includes:
FieldDescription
QuestionThe guest question that the AI couldn’t fully answer
Suggested AnswerMIRA’s proposed answer based on available context
PropertyWhich property the question relates to
ConfidenceHow confident MIRA is in the suggested answer
SourceWhere the question originated (WhatsApp, email, etc.)
Red FlagsPotential issues with the suggested answer that need human review

​
Reviewing Candidates

Navigate to Agent Hub → Knowledge Candidates to see all pending suggestions.
For each candidate, you can:
  • Approve — adds the question and answer to your knowledge base immediately
  • Reject — dismisses the candidate (MIRA won’t suggest it again)
  • Edit — modify the suggested answer before approving

​
Filtering Candidates

Filter candidates by:
  • Property — see candidates for a specific property
  • Confidence level — focus on high-confidence suggestions first
  • Category — filter by topic type

​
Why MIRA Matters

Without MIRA, improving your AI requires manually identifying knowledge gaps — reading through conversations to find unanswered questions. MIRA automates this discovery process:
  • No knowledge gaps go unnoticed — every unanswered question becomes a candidate
  • AI accuracy improves continuously — each approved candidate makes your AI smarter
  • Human stays in control — MIRA suggests, you decide what goes into the knowledge base
  • Per-property intelligence — candidates are linked to specific properties for targeted improvement

​
MIRA + AI Accuracy

The Knowledge Base page shows an AI Accuracy percentage per property. This measures how often the AI can answer guest questions using existing knowledge. As you approve MIRA candidates, this percentage increases — directly improving your automation rate.