Knowledge Gaps
Identify and resolve unanswered customer questions. Track queries that failed semantic search thresholds, analyze their frequency, and convert them into trained Q&A pairs.
What is a Knowledge Gap?
A Knowledge Gap is recorded when a customer asks a question on WhatsApp, but the SyncFlo AI engine is unable to locate a matching answer in your Knowledge Base with a confidence score above the retrieval threshold.
Instead of guessing or hallucinating, the bot will trigger a fallback response (e.g. asking for clarification or routing to a human agent) and log the query in the Knowledge Gaps console. This highlights exactly what your customers are asking that your AI bot doesn't know yet.
Key Gap Metrics & Attributes
Each identified gap is tracked with the following parameters to help you prioritize your training efforts:
Tracks the number of unique conversations where this question (or semantically similar questions) was asked.
Automatically detected subject category (e.g., Shipping, Billing, Returns, Pricing) based on text analysis.
Categorized as High, Medium, or Low based on question frequency and urgency thresholds.
Timestamp showing the exact date and hour the question was last received from a user on WhatsApp.
The Resolution Workflow
SyncFlo provides a loop to close knowledge gaps. Resolving a gap automatically creates a first-class Q&A pair that feeds back into your AI agent's brain.
Review Gaps List
Navigate to Dashboard > WhatsApp > Knowledge > Gaps. Sort the table by Frequency to find the most common unresolved customer issues.
Draft the Answer
Click the Answer button next to the gap. This opens the resolution dialog showing the raw customer question and affected conversation threads. Type a factual, clear, and action-oriented response.
Save as Q&A Pair
Upon clicking Save Answer, the system writes this new Q&A pair under a built-in source named "Knowledge Gaps". The status of the gap changes to Resolved.
Instant Retraining
The vector database indexes this new Q&A pair instantly. Any customer asking a matching question on WhatsApp will receive the saved answer immediately.
Vector Thresholds & Fallbacks
How does the AI determine if an answer is sufficient or if it constitutes a gap? SyncFlo checks the cosine similarity score between the user's incoming query and your vector database index:
Confidence > 80% (Auto-Reply)
The matching text chunk is highly relevant. SyncFlo automatically extracts the context, generates the answer, and sends it directly back to the customer on WhatsApp.
Confidence < 80% (Fallback + Gap)
The similarity score is too low. The agent refrains from answering to avoid incorrect responses. It sends the designated fallback prompt, logs a new unresolved entry in Knowledge Gaps, and optionally flags a human agent for handoff.
Writing Premium Gap Resolutions
When writing answers to resolve gaps, make sure they are:
- Actionable: Give clear, definitive answers rather than generic or vague summaries.
- Comprehensive: Address sub-queries that are usually tied to this gap (e.g. if shipping prices are asked, include standard delivery times as well).
- Concise: Keep responses brief. WhatsApp messages should be easy to read on mobile screens.