How Insights Assistant works

Medallia Insights Assistant includes a 4-stage process through which it ingests natural language, evaluates variables, surfaces contextual summaries, and delivers traceable, enterprise-ready data.

System operational flow: Agentic Interpretation> Data Synthesis> Result Traceability> Operational Sharing.

Stage 1: Agentic Interpretation & Selection

This phase represents the initiation of the workflow where conversational user inputs are translated into structured, machine-executable search strategies.

Operational Step-by-Step Execution:

  1. Submit Natural Language Query: The user inputs an open-ended business question using conversational language into the Insights Assistant dialog field (e.g., "Why did complaints spike last month?").

  2. Intent Parsing: The underlying Agentic AI parses the linguistic intent behind the user's prompt to identify key contextual boundaries, filters, and target fields.

  3. Dynamic Tool Matching: The AI programmatically determines and selects the optimal reporting module tools (such as TA Snapshot or Trends Over Time) needed to safely search, locate, and retrieve the required records.

  4. Targeted Data Extraction: The active system agent safely accesses Medallia, Inc.'s defined datasets, querying precisely within the boundaries of the user's role-based permissions to isolate the relevant information.

Stage 2: Data Synthesis & Summarization

Once raw numbers and metrics are isolated, this stage converts complex datasets into digestible executive text summaries.

Operational Step-by-Step Execution:

  1. Transmit Payloads to GenAI: The system agent securely structures the retrieved text variables, exact calculation counts, and data fragments, passing the clean payload over to the generative AI engine layer.

  2. Read and Filter Extracted Data: The GenAI engine processes the filtered subset of information, evaluating the textual distributions, trends, and scores defined in the previous step.

  3. Compile Narrative Summaries: The engine creates a human-readable, narrative text summary that maps key takeaways directly answering the core intent of the user's initial question.

  4. Deploy Operational Review: Frontline supervisors, operational managers, or high-level executives review the final contextual narrative block to instantly decide on localized or global tactical actions.

Stage 3: Traceable Results

This phase addresses accuracy validation, neutralizing the traditional AI "black box" by providing transparent access to source metrics.

Operational Step-by-Step Execution:

  1. Review Reference Links:

    Below the narrative summary block, the system automatically appends direct, inline resource links mapped explicitly to the source modules or charts used in the generation phase.

  2. Execute Multi-Source Audits:

    With just a few clicks, the user can follow the embedded links back to the raw reporting modules to cross-verify specific data variables or sample distributions.

  3. Eliminate Black-Box Ambiguity:

    By revealing the precise "thinking trail" of the agent, the tool builds administrative trust and ensures strict, measurable accuracy targets are maintained.

  4. Accelerate Speed-to-Insight:

    Users bypass traditional manual report filtering rows, verifying automatically generated conclusions in seconds rather than submitting cross-functional data query tickets.

Stage 4: Real-time & Shareable Insights

The closing stage focuses on distributing high-impact insights across complex organizational environments to drive collaborative resolution.

Operational Step-by-Step Execution:

  1. Access Live Operational Status:

    The system guarantees that all answers pull from a complete population of data, providing near real-time insights that accurately reflect the current state of customer and employee experiences.

  2. Initiate Broader Stakeholder Discussions:

    Armed with targeted metrics and root causes, teams can instantly launch data-driven problem-solving sessions without spending time matching offline reports.

  3. ConfigurePresentation Exports:

    Users click the native export panel to seamlessly package single questions, conversation threads, or narrative blocks into cleanly formatted presentation documents in PDF format.

  4. Deploy Findings Universally:

    Share downloaded conversations or summaries directly through corporate alignment channels (e.g., Slack or email updates), providing immediate organizational clarity across distributed cross-functional business units.