Visual Module Breakdown & Interfaces
Click the tabs below to view detailed functional breakdowns for each core module of the AI Call Recording Multichannel Analyzer Software.
Analyze Sales Call — Ingestion & Dialect Detection
Analyze Sales Call upload portal with language auto-detect, agent picker, and regional assignment.
Audio Upload (mp3, wav, m4a)
Language Auto-Detect
Salesperson Attribution
Region Tagging
Operational Purpose & Capabilities
The Analyze Sales Call portal allows sales reps and managers to upload B2B audio recordings (.wav, .mp3, .m4a). The system transcribes dialogue, translates regional phrasing, and generates evaluation scores with actionable tips.
Input Parameters & Form Controls
- Select Audio File: Native file picker accepting call audio files.
- Language Selection: Configurable dropdown with Auto Detect, Hindi (HI), Gujarati (GU), Tamil (TA), and English.
- Salesperson Attribution: Dynamic agent selector linking the call directly to a rep's profile.
- Region Tagging: Categorizes calls into South Region, North Region, or West Region for territory reporting.
- Analyze Call Action: Submits multipart audio data to
/api/analysis/uploadand starts the AI pipeline.
AO Dashboard — Regional Analytics & Leaderboards
Figure 2: Area Officer (AO) Dashboard showing KPI metric cards, agent leaderboard, and regional breakdowns.
Team Avg Score KPI
Agent Leaderboard
Needs Attention Flag
Regional Breakdowns
Operational Purpose & Executive Metrics
The AO Dashboard is the primary view for sales leaders to monitor team performance, top scores, and regional benchmarks.
Key Performance Indicator (KPI) Widgets
- TOTAL CALLS (32): Cumulative call recordings analyzed across all regional territories.
- TEAM AVG SCORE (53.2 / 100): Rolling aggregate quality score across the entire sales roster.
- TOP SCORE (91 / 100): Highest scoring single sales call, serving as the benchmark standard.
- NEEDS ATTENTION (0 / 100): Lowest scoring call, highlighting reps who require immediate coaching.
Agent Leaderboard & Regional Distribution
The leaderboard ranks reps by average score and grade (e.g. Suresh Patel: 88.8 A+, Ravi Kumar: 81 A+, Priya Sharma: 68 A, Arjun Singh: 59.9 A). The Region Breakdown tracks territory performance: West Region (70.6 avg • 7 calls), North Region (70 avg • 4 calls), and South Region (44.2 avg • 21 calls).
Agent Profile Drawer & Deep-Dive History
Agent Profile slide-over drawer displaying individual rep stats and chronological call history.
Individual Rep Stats
Chronological Call Stream
Deal Stage Tracking
Slide-over UI
Operational Purpose & Drill-Down Analytics
Clicking View ➔ on any rep in the leaderboard opens a slide-over drawer showing individual performance history. This view gives managers full visibility into an agent's calls without navigating away from the dashboard.
Agent Profile Summary Widgets (Example: Suresh Patel)
- TOTAL CALLS (4): Total analyzed interactions attributed to the agent.
- AVG SCORE (88.8): Cumulative mean score across the agent's calls.
- BEST SCORE (91 / 100): Peak performance benchmark achieved by this representative.
Chronological Call Audit Stream
The drawer lists individual calls with IDs, timestamps, deal stages, languages, and scores:
- #d3a4 — 2025-07-06 • Closed • Gujarati (gu) — 90 • A+
- #d3a3 — 2025-07-04 • Intent • Gujarati (gu) — 86 • A+
- #d3a2 — 2025-07-02 • Closed • Gujarati (gu) — 91 • A+
- #d3a1 — 2025-07-01 • Closed • Gujarati (gu) — 88 • A+
Master Call Reports & Funnel Stage
Master Call Reports table with agent filtering, language tags, conversion probabilities, and sentiment.
Consolidated Records
Sentiment Mapping
Conversion Probability
Language Identifiers
Operational Purpose & Audit Schema
The Call Reports table provides a consolidated record of all analyzed calls across the organization. It maps customer sentiment and deal stages to track deal progression from first discovery to close.
Data Columns & Operational Tags
- CALL ID: Unique hash identifier (e.g. #bc85afa0, #ade04ade, #d5a4).
- AGENT & REGION: Sales rep name with regional territory badge (South Region, West Region, North Region).
- DATE & LANG: Call date and language identifier (AUTO, HI, GU, TA).
- SCORE: Numerical quality score paired with grade badge (e.g. 90 • A+, 76 • A+, 45 • A, 20 • B).
- CONVERSION: AI-estimated deal probability (High, Medium, Low).
- STAGE: Pipeline progress (Awareness, Interest, Intent, Closed).
- SENTIMENT: Customer attitude classification (Positive vs. Neutral).
Industry Context & The B2B Sales Quality Blind Spot
Enterprises running distributed B2B sales operations encounter severe quality assurance bottlenecks. Sales representatives and field executives make thousands of outbound commercial pitches every month across diverse geographical territories and languages (including English, Hindi, Gujarati, Tamil, and regional dialects). Due to limited management bandwidth, sales supervisors audit less than 2% of recorded customer interactions. This creates massive revenue leakage, unchecked pitch deviation, and persistent loss of qualified pipeline opportunities.
The Flaws of Manual Sampling Audits
Traditional Quality Assurance (QA) in tele-sales relies on human supervisors listening to occasional 10-minute audio snippets. This legacy approach is fundamentally broken across three vectors:
1. Statistical Irrelevance: Reviewing 1 in 50 calls guarantees that compliance violations, customer objections, and pitch failures in the other 98% remain completely hidden.
2. Subjective Manager Grading: Human reviews depend heavily on individual manager mood, personal bias, and relationship dynamics, creating friction with sales agents.
3. Lagging Actionability: Manual feedback is provided days or weeks after a call occurred, far too late to salvage lost revenue or correct active customer misunderstandings.
The Multilingual Barrier in Indian Commercial Markets
Commercial conversations across India rarely adhere to formal textbook English. Field representatives blend regional idioms and technical terminology into code-switched dialogues (e.g., Hindi mixed with English, Gujarati, or Tamil).
Standard off-the-shelf voice transcription software breaks down when processing these real-world dialects, producing unusable transcripts. AI Call Recording Multichannel Analyzer Software was specifically developed to bridge this gap, transcribing, translating, and evaluating multi-dialect sales interactions.
What Is AI Call Recording Multichannel Analyzer Software?
AI Call Recording Multichannel Analyzer Software is an autonomous conversational analytics and sales quality platform built to provide 100% automated auditing and personalized rep coaching across enterprise commercial organizations.
Strategic Business Vision
The system standardizes sales performance management into an empirical, data-driven science. By analyzing every interaction objectively, sales leaders can eliminate guesswork, calibrate compensation with true closing competence, and duplicate successful sales behaviors across entire regional territories.
Core Technological Capabilities
1
Multilingual Audio Ingestion
Supports high-fidelity transcription and automatic dialect identification across Hindi (HI), Gujarati (GU), Tamil (TA), and English (app.py, analysis.py).
2
Groq LPU Accelerated Inference
Leverages specialized Language Processing Units (LPUs) to evaluate call transcripts in under 3 seconds (services/groq_service.py).
3
9-Metric Objective Scoring Rubric
Evaluates call quality across Greeting, Need Discovery, Value Pitch, Objection Handling, Closing Ability, and compliance standards (services/eval_service.py).
4
Area Officer (AO) Management Hub
Delivers real-time regional leaderboards, team average scores, and conversion funnel distribution tracking (routes/dashboard.py).
5
Automated Agent Coaching
Generates constructive strengths and actionable coaching recommendations tailored to individual sales reps.
Detailed User Personas & Access Control
Organizational Hierarchy
AI Call Recording Multichannel Analyzer Software serves distinct user personas across executive management, regional leadership, and frontline sales.
Persona 1: Area Officer / Regional Sales Manager
Role Code: area_officer
Scope: Regional territory oversight, team leaderboards, rep coaching, and performance calibration.
Core Responsibilities: Monitoring team average scores (e.g. 53.2/100), inspecting individual agent drawers (Suresh Patel, Ravi Kumar), identifying low-scoring calls (0/100), and exporting regional reports (image_391d16.png).
Platform Interfaces: AO Dashboard, Agent Performance Drawer, Daily Report, and Team Excel exports.
Persona 2: Frontline Sales Executive
Role Code: sales_rep
Scope: Outbound commercial dialing, customer discovery, objection handling, and deal closing.
Core Responsibilities: Uploading completed audio recordings, reviewing personal call scores (e.g. 88.8 A+), studying AI coaching recommendations, and monitoring personal leaderboard standing.
Platform Interfaces: Analyze Sales Call upload portal, personal call reports, and feedback cards.
Persona 3: VP of Sales / Chief Commercial Officer
Role Code: executive
Scope: Macro revenue pipeline health, regional performance comparisons, and training curriculum design.
Core Responsibilities: Evaluating regional conversion efficiencies (West: 70.6 avg, North: 70 avg, South: 44.2 avg), tracking pipeline stage velocity (Awareness → Closed), and standardizing company-wide sales playbooks.
Platform Interfaces: Master Call Reports, Regional Breakdown bar charts, and System Trend analytics.
Architecture & Pipeline
5. Full Technology Stack & Inference Pipeline Topology
6. Complete Repository & Codebase Directory Layout
5
Technology Stack & Inference Pipeline
The platform utilizes an asynchronous, high-throughput pipeline coupling a Python FastAPI application with cloud-accelerated Groq LPUs for rapid speech-to-intent analysis.
| Subsystem | Technology / Library | Operational Functionality |
|---|---|---|
| API Framework | FastAPI (Python 3.11) | Asynchronous REST controller exposing endpoints for upload, eval, and reports (app.py). |
| Inference Engine | Groq Cloud SDK | Sub-second LLM execution (Llama-3-70B-Versatile) for transcript scoring (groq_service.py). |
| Speech Pipeline | Whisper Large-v3 | Multilingual audio transcription and English translation with timestamp alignment. |
| Database ORM | SQLAlchemy 2.0 | Relational models managing agents, calls, scores, and regional metrics (models.py). |
| Database Engine | SQLite / PostgreSQL | Persistent storage configured via DATABASE_URL in config.py. |
| Frontend View | Jinja2 Templates | Server-rendered responsive UI with purple brand theme (templates/index.html). |
| Styling Core | Tailwind CSS + Custom CSS | Clean interface with responsive metric cards and modals (static/css/style.css). |
| Client-Side Logic | Vanilla JavaScript (ES6+) | Dynamic DOM updates, asynchronous fetch requests, and drawer toggles (app.js). |
| Batch Automation | Windows Batch Scripts | One-click execution and directory cleanup (start.bat, clean.bat). |
6
Complete Repository & Codebase Directory Layout
The repository is organized into modular services, route controllers, and static assets:
AI-Call-Analyzer-main/
├── README.md # System installation, environment variables, and run guide
├── app.py # FastAPI application factory, middleware, and router mounts
├── config.py # Centralized settings (GROQ_API_KEY, DATABASE_URL, PORT)
├── database.py # SQLAlchemy engine, session maker, and DB init routines
├── models.py # Database models (Agent, CallRecord, ScoreEvaluation)
├── routes/
│ ├── __init__.py # API routers package initializer
│ ├── agents.py # Agent directory, profile drawer, and performance metrics
│ ├── analysis.py # Audio upload handler, Whisper transcription, and scoring trigger
│ ├── dashboard.py # AO metrics, regional distribution, and recent call activity
├── services/
│ ├── __init__.py # Services package initializer
│ ├── eval_service.py # 9-metric qualitative evaluation engine and prompt templates
│ ├── groq_service.py # Groq API client with error handling and retry logic
├── static/
│ ├── css/style.css # Custom UI styles, purple header theme, and grade badges
│ ├── js/app.js # Frontend state management, audio upload AJAX, drawer controls
├── templates/index.html # Single-page dashboard containing all views and modals
├── requirements.txt # Python dependencies (fastapi, uvicorn, groq, sqlalchemy)
├── start.bat # Windows one-click environment launcher
├── clean.bat # Temporary cache, pycache, and audio purge script
├── README.md # System installation, environment variables, and run guide
├── app.py # FastAPI application factory, middleware, and router mounts
├── config.py # Centralized settings (GROQ_API_KEY, DATABASE_URL, PORT)
├── database.py # SQLAlchemy engine, session maker, and DB init routines
├── models.py # Database models (Agent, CallRecord, ScoreEvaluation)
├── routes/
│ ├── __init__.py # API routers package initializer
│ ├── agents.py # Agent directory, profile drawer, and performance metrics
│ ├── analysis.py # Audio upload handler, Whisper transcription, and scoring trigger
│ ├── dashboard.py # AO metrics, regional distribution, and recent call activity
├── services/
│ ├── __init__.py # Services package initializer
│ ├── eval_service.py # 9-metric qualitative evaluation engine and prompt templates
│ ├── groq_service.py # Groq API client with error handling and retry logic
├── static/
│ ├── css/style.css # Custom UI styles, purple header theme, and grade badges
│ ├── js/app.js # Frontend state management, audio upload AJAX, drawer controls
├── templates/index.html # Single-page dashboard containing all views and modals
├── requirements.txt # Python dependencies (fastapi, uvicorn, groq, sqlalchemy)
├── start.bat # Windows one-click environment launcher
├── clean.bat # Temporary cache, pycache, and audio purge script
The 9-Metric Sales Evaluation Rubric Explained
The evaluation engine in services/eval_service.py scores sales transcripts across 9 objective criteria:
| # | Evaluation Metric | Weight | Scoring Criteria & Definition |
|---|---|---|---|
| 1 | Greeting & Rapport | 10% | Professional introduction, clear identification of self and company, warm and respectful tone. |
| 2 | Need Discovery | 15% | Asks open-ended questions to uncover customer pain points, current setup, and budget constraints. |
| 3 | Value Pitch | 15% | Tailors product benefits to the customer's specific needs rather than reciting a generic feature list. |
| 4 | Product Knowledge | 10% | Accurately explains technical specs, pricing tiers, integration requirements, and terms. |
| 5 | Objection Handling | 15% | Listens to hesitation, acknowledges concerns with empathy, and provides persuasive solutions. |
| 6 | Active Listening | 10% | Avoids interrupting, confirms understanding, and addresses customer questions directly. |
| 7 | Closing Ability | 10% | Asks for commitment, secures next steps, books a demo, or clarifies follow-up timeline. |
| 8 | Tone & Professionalism | 10% | Maintains an engaging, confident, and polite demeanor throughout the conversation. |
| 9 | Compliance & Accuracy | 5% | Follows legal disclosure rules, confirms recording consent, and adheres to company guidelines. |
Mathematical Scoring Engine & Letter Grade Curves
Each metric is scored on a 0 to 10 scale by the Groq-powered evaluation model. The overall score is computed as a weighted sum normalized to 100 points:
Final Score = Σ (Si × Wi) × 10
where Si ∈ [0, 10] is the raw metric score and Wi is the normalized metric weight.
Letter Grade Classification Curve
| Score Range | Grade | Performance Classification | Prescribed Operational Action |
|---|---|---|---|
| 85 — 100 | A+ | Benchmark / Master Sales Closer | Archive call as a training template for new sales onboarding. |
| 70 — 84 | A | Proficient / Strong Performer | Targeted feedback on minor discovery or closing improvements. |
| 50 — 69 | B | Developing / Inconsistent Pitch | Assign 1-on-1 manager roleplay on objection handling. |
| 0 — 49 | C / Needs Attention | Critical Underperformance | Immediate manager intervention; audit call for compliance breaches. |
Multilingual Audio Pipeline
To support diverse Indian sales teams, the platform includes a multilingual speech pipeline that handles regional languages and code-switched phrasing.
Language Pipeline Workflow
1. Audio Normalization: Converts uploaded audio to 16kHz mono WAV format to improve speech recognition accuracy.
2. Dialect Detection: When Auto Detect is chosen, Whisper parses the opening 30 seconds of audio to identify the language.
3. Dual Transcription & Translation: Generates both a verbatim native transcript and an aligned English translation.
4. Normalized Scoring: The translated transcript is evaluated against the 9-metric rubric, ensuring consistent scoring across all languages.
Groq LPU Inference Acceleration
Traditional cloud GPUs often take 15 to 30 seconds to evaluate a multi-page sales transcript. AI Call Recording Multichannel Analyzer Software uses Groq Language Processing Units (LPUs) via services/groq_service.py to deliver evaluations in under 3 seconds.
Inference Performance Comparison
| Hardware Pipeline | Token Speed | Analysis Time | User Experience |
|---|---|---|---|
| Standard Cloud GPU (A10G) | 45 tokens/sec | 18.5 seconds | Noticeable UI waiting delay during call analysis. |
| Groq LPU (Llama-3-70B) | 480+ tokens/sec | 2.4 seconds | Near real-time evaluation with immediate dashboard updates. |
Relational Database Schema & Data Models (ERD)
The relational database schema is defined in models.py using SQLAlchemy ORM:
| Table Name | Primary / Foreign Keys | Attributes & Field Types |
|---|---|---|
| agents | id INT PK | name (VARCHAR), email (VARCHAR UNIQUE), region (VARCHAR: South/North/West), created_at (DATETIME). |
| call_records | id INT PK agent_id INT FK | call_hash (VARCHAR UNIQUE), filename (VARCHAR), language (VARCHAR), region (VARCHAR), duration (INT), deal_stage (VARCHAR), sentiment (VARCHAR), created_at (DATETIME). |
| evaluations | id INT PK call_id INT FK | overall_score (FLOAT), grade (VARCHAR), greeting_score (FLOAT), discovery_score (FLOAT), pitch_score (FLOAT), objection_score (FLOAT), closing_score (FLOAT), coaching_notes (TEXT). |
| transcripts | id INT PK call_id INT FK | original_text (TEXT), translated_text (TEXT), detected_language (VARCHAR), turn_count (INT). |
REST API References
16. Call Upload & Pipeline Triggering (routes/analysis.py)
| Endpoint Route | Method | Parameters & Payload | Response Output |
|---|---|---|---|
| /api/analysis/upload | POST | Multipart form: audio_file, agent_id, region, language. | Returns call_id, transcription text, English translation, and evaluation status. |
| /api/analysis/status/{id} | GET | Path parameter: id (Call ID hash). | Returns real-time processing status (transcribing, evaluating, completed). |
Dashboard Analytics & Leaderboards (routes/dashboard.py)
| Endpoint Route | Method | Description | Response Schema |
|---|---|---|---|
| /api/dashboard/kpis | GET | Calculates aggregate KPIs (total calls, team average, top score, lowest score). | JSON object with summary statistics across all analyzed calls. |
| /api/dashboard/leaderboard | GET | Returns ranked list of agents sorted by average evaluation score. | Array of agent records: name, region, calls count, average score, and grade. |
| /api/dashboard/regions | GET | Computes regional averages and call counts (South, North, West). | Array of regional metrics used to render dashboard progress bars. |
Agent Profiles & Trends (routes/agents.py)
| Endpoint Route | Method | Description | Response Schema |
|---|---|---|---|
| /api/agents/list | GET | Fetches all active agents for upload form dropdown menus. | Array of objects with agent_id, name, and region. |
| /api/agents/{id}/history | GET | Fetches historical calls and scores for an agent drawer. | Agent summary stats and chronological call history array. |
| /api/reports/export-excel | GET | Generates downloadable Excel workbook of all call metrics. | Binary Excel file stream formatted for spreadsheet applications. |
Standard Operating Procedures
19. Ingesting & Analyzing a Call
20. Daily Review & Cadence
21. Regional Calibration
19
SOP: Ingesting & Analyzing a B2B Sales Audio Recording
Standard operating procedure for uploading and processing call recordings:
- Open Analyze Call from the left navigation menu (
image_391d37.png). - Click Choose File and select the recording (
.mp3,.wav, or.m4a). - Set language: Keep Auto Detect for mixed dialects, or select a specific language (Hindi, Gujarati, Tamil).
- Select the sales representative's name from the Salesperson dropdown.
- Select the territory from the Region dropdown (South Region, North Region, West Region).
- Click Analyze Call to start transcription and evaluation.
- Once completed, review the generated score, transcript, and coaching recommendations.
20
SOP: Area Officer Daily Review & Rep Coaching Cadence
Daily review workflow for Area Officers to monitor sales performance and provide coaching:
- Open the AO Dashboard (
image_391d16.png) and review yesterday's key metrics. - Check NEEDS ATTENTION to identify any calls that scored below 50 points.
- Review the Agent Leaderboard to see changes in team rankings and performance.
- Click View ➔ on developing reps to review their recent call history and coaching notes.
- Share high-scoring calls (85+ points) with the team as best-practice examples.
21
SOP: Regional Calibration & Team Excel Generation
Procedures for monthly performance reporting and territory calibration:
- On the AO Dashboard, click Team Excel to export all call records (
image_391d16.png). - Click Daily Report to generate an executive summary for regional sales leadership.
- Compare regional averages (West vs. North vs. South) to identify training needs across territories.
- Adjust regional sales targets based on stage progression and conversion rates.
Engineering & Operations
22. Prompt Engineering Schema
23. Deployment Runbook
24. Database Migrations & Backup
22
Prompt Engineering: System Scoring Directives & Schema
The system prompt in services/eval_service.py standardizes LLM output into structured JSON:
{
"scores": {
"greeting": 8.5,
"need_discovery": 9.0,
"value_pitch": 8.0,
"product_knowledge": 9.0,
"objection_handling": 8.5,
"active_listening": 9.0,
"closing_ability": 8.0,
"tone_professionalism": 9.0,
"compliance": 10.0
},
"overall_score": 88.8,
"letter_grade": "A+",
"deal_stage": "Closed",
"sentiment": "Positive",
"conversion_probability": "High",
"strengths": [
"Identified inventory challenges within the first 2 minutes.",
"Handled pricing objections by emphasizing software ROI."
],
"coaching_recommendations": [
"Confirm follow-up onboarding dates more explicitly before closing."
]
}
"scores": {
"greeting": 8.5,
"need_discovery": 9.0,
"value_pitch": 8.0,
"product_knowledge": 9.0,
"objection_handling": 8.5,
"active_listening": 9.0,
"closing_ability": 8.0,
"tone_professionalism": 9.0,
"compliance": 10.0
},
"overall_score": 88.8,
"letter_grade": "A+",
"deal_stage": "Closed",
"sentiment": "Positive",
"conversion_probability": "High",
"strengths": [
"Identified inventory challenges within the first 2 minutes.",
"Handled pricing objections by emphasizing software ROI."
],
"coaching_recommendations": [
"Confirm follow-up onboarding dates more explicitly before closing."
]
}
23
Deployment Runbook: FastAPI, Uvicorn & Batch Setup
Operating procedures for deploying and running the platform in production environments:
| Command Syntax | Operational Function |
|---|---|
start.bat | Windows launch script that activates virtual environment and starts Uvicorn server. |
clean.bat | Purges temporary cache files, __pycache__, and local audio files. |
pip install -r requirements.txt | Installs all required Python libraries (FastAPI, Groq, SQLAlchemy, etc.). |
uvicorn app:app --host 0.0.0.0 --port 8000 --workers 4 | Launches production ASGI server with multi-worker process management. |
24
Database Migration & SQLite/PostgreSQL Backup Runbook
Procedures for regular database maintenance, schema migrations, and backups:
Backup Procedures
- SQLite Environments: Create daily compressed archives of analyzer.db:
copy analyzer.db backups\analyzer_%date%.db - PostgreSQL Environments: Run standard automated database dumps:
pg_dump -U analyzer_user analyzer_db | gzip > backup.sql.gz
Disaster Recovery Runbook
- Stop the application server: Ctrl+C or terminate Uvicorn processes.
- Restore the database file from the most recent backup archive.
- Run schema verification using SQLAlchemy initialization routines.
- Restart the server and verify dashboard metrics load correctly.
Security Architecture
Audio Encryption & Redaction Controls. Audio recordings and transcripts may contain sensitive customer data. The platform includes security controls to protect information in transit and at rest.
Data Protection Controls
PII Redaction: Customer phone numbers, credit card details, and personal identifiers are scrubbed from transcripts before evaluation.
Encrypted API Communication: Groq LPU requests use TLS 1.3 encryption to protect audio data in transit.
Access Control: Dashboard views are role-gated to prevent unauthorized rep access to broader team metrics.
System Troubleshooting
Diagnostic guide for resolving common upload, audio processing, and evaluation errors:
| Symptom | Probable Cause | Remediation Procedure |
|---|---|---|
| Upload fails with HTTP 413 | Audio file size exceeds maximum upload limit. | Increase maximum body size in FastAPI settings or compress audio to 16kHz mono. |
| Groq API returns rate-limit error | API key exceeded tokens-per-minute (TPM) quota. | Implement backoff retry logic or upgrade API tier quota. |
| Garbled regional transcript | Wrong dialect selected on upload. | Select Auto Detect or explicitly set the matching regional dialect dropdown. |
| Dashboard displays 0 calls | Database connection issue or missing seed data. | Check DATABASE_URL in config.py and verify database file permissions. |
Frequently Asked Questions (FAQ)
Common questions regarding AI Call Recording Multichannel Analyzer Software capabilities and configuration:
Q: Can the platform evaluate mixed-language sales calls?
A: Yes. The Whisper Large-v3 engine detects code-switched speech (e.g. Hindi mixed with English) and translates it into clear English for evaluation.
Q: How are deal conversion probabilities calculated?
A: Probabilities are derived from customer sentiment, objection resolution, and agreement to next steps in the transcript.
Q: Can evaluation criteria weights be customized?
A: Yes. Weights can be adjusted in services/eval_service.py to align with company-specific sales methodology.
Operational Glossary & Speech Intelligence Terminology
| Term | Definition in AI Call Recording Multichannel Analyzer Software Context |
|---|---|
| Groq LPU | Language Processing Unit hardware designed for high-speed LLM inference. |
| Whisper Large-v3 | State-of-the-art multilingual speech recognition and translation model. |
| Area Officer (AO) | Regional sales manager responsible for territory performance and rep coaching. |
| B2B Sales Enablement | Automated tooling and feedback designed to help sales teams close deals more effectively. |
| 9-Metric Rubric | Standardized evaluation criteria assessing sales call quality from greeting to close. |
| Deal Stage Progression | Automated tracking of buyer journey stages (Awareness, Interest, Intent, Closed). |
| Turn-Taking Transcript | Time-aligned dialogue record separating agent and customer speaking turns. |
| Conversion Probability | AI-estimated likelihood of deal closure based on conversational sentiment. |
AI Call Recording Multichannel Analyzer Software
B2B Sales Call Intelligence & Speech Analytics ERP