BETSOL AI Labs

Live demonstrations of AI-powered solutions for our customers.
AI assistance, augmentation, and even complete automation.

Related blog
Learn and play chess
Ironsight
Healthcare clinicians portal
HRMS
Algis
Smart jira
Specialist Guide

Demo lab projects

Greenfield without 3rd-party integrations chess.betsol.com

AI-accelerated app development

A secure, high-performance application built entirely through AI-guided development as a POC for taking AI automation to the extreme. Features adaptive AI opponents, personalized learning paths, and an AI-authored content management system.

Key highlights

Built from concept to deployment in 2 weeks
AI wrote 100% of the code under expert guidance
Includes 50+ AI-generated lessons and tutorials
Enterprise-ready with full security and performance optimization

Try now More

Features

Full-featured web-based Chess Trainer, including modules, lessons, demos, exercises, analytics, and adaptive free play against AI.
Complete content management system (like WordPress) built from the ground up to allow for agentic AI to write all of the marketing content, while meeting all SEO best practices.

Human work

Brainstormed the initial idea with AI
Structured the AI workflow
Performed user acceptance testing for most user stories

AI work

Brainstormed the initial idea with the human
Wrote the PRD (Product Requirements Document)
Research and created the chess curriculum
Created the style guide
Broke down PRD into development phases → Sprints → small, testable User Stories
Wrote prompts for each User Story
Executed the Sprints, automatically pulling from backlog
Tested and debugged
Ensured enterprise readiness by ending each sprint with performance and security User Stories
Managed CI/CD, build, and production deployments
Created the content plan
Wrote all marketing content and created all images

Agentic Tech & Workflows

Claude
Modern web frameworks
Automated testing
CI/CD pipelines

Greenfield with 3rd-party integrationsBETSOL

IronSight: Detecting Shadow IT & Shadow AI

IronSight is a security innovation designed to provide real-time visibility and control over enterprise Windows endpoints. Built as an AI-driven monitoring platform, it helps identify risky domain access, shadow IT usage, and emerging shadow AI activity including unauthorized AI tools and services, enabling proactive risk management, stronger data protection, and adaptive enterprise governance.

Key highlights

Real-time DNS-level monitoring across enterprise endpoints
Instant blocking of unauthorized domains and AI services
Centralized visibility into applications, AI usage, and device activity
Comprehensive audit trails supporting compliance and security analysis

Capabilities:

Captures all DNS queries at the kernel level, including traffic that bypasses traditional monitoring tools
Enforces security policies instantly using NRPT rules and Windows Firewall integration
Terminates active TCP connections to prevent data exfiltration to unauthorized services
Aggregates installed software by publisher through registry-based inventory collection
Generates AI-assisted risk insights for domains and applications with cached intelligence
Stores authoritative DNS responses and full historical activity for audit and investigation

Agentic Tech & Workflows

Claude Code
Python
Rust
ClickHouse
Windows ETW
Modern web frameworks

Greenfield without 3rd-party integrations U.S. Healthcare Company

AI-Enhanced Clinicians Tracking System

A comprehensive provider management and clinician operations portal built using secure agentic AI and augmented engineering workflows. Designed for healthcare environments, the platform delivers real-time clinician availability, facility scheduling, and advanced provider search while replacing fragmented legacy tooling. The system combines AI-assisted development with human validation to achieve faster delivery without compromising healthcare-grade security, auditability, or data integrity.

Key highlights

Real-time clinician availability and facility scheduling visibility
High-performance provider directory with sub-100ms search response
Healthcare-grade security with Azure AD SSO, MFA, and RBAC enforcement
Scalable serverless architecture with fully auditable clinician and provider workflows

Capabilities

Centralized clinician resource management and provider directory
Advanced provider search via optimized Azure SQL views (including vwProviderSearchEnhanced)
Secure role-based access control aligned with Azure AD groups
Parameterized data access with stored procedures and strict read/write separation
Centralized audit logging and privacy-aware error handling
Automated Swagger/OpenAPI documentation and developer artifacts
Expanded unit and regression coverage through AI-assisted test generation
Scalable serverless backend powered by Go-based Azure Functions

Agentic Tech & Workflows

Claude
Microsoft Azure
Azure SQL
Swagger / OpenAPI
Go (Azure Functions)

Brownfield with 3rd-party integrations BETSOL

AI-Powered HRMS Platform

A full-featured HR management system with employee lifecycle management, digital onboarding workflows, document management, Azure AD integration, and real-time compliance tracking.

Key highlights:

Hierarchical organization structure with multi-level reporting
Real-time Azure AD synchronization every 15 minutes
Automated employee onboarding and offboarding
GDPR and SOC 2 compliance built-in

Read the case study More

Secure document storage with Azure Blob integration
Role-based access controls with granular permissions
Automated retention policies and compliance tracking
Version control and audit trails for all documents
Customizable task templates for different roles
Multi-department coordination (IT, HR, Finance, Manager)
Progress tracking and automated notifications
Digital document collection and e-signatures
Immutable audit logs for all system actions
AES-256 encryption at rest, TLS 1.3 in transit
Role-based access control with Azure AD SSO
Backend: Python 3.11+, FastAPI, SQLAlchemy 2.0 (async), Pydantic, Celery, PyJWT
Database: MySQL 8.0+ with complex relational schema, stored procedures, and triggers
Cloud: Microsoft Azure (App Service, Blob Storage, Azure AD, Key Vault, Monitor)
Security: OAuth 2.0, SAML SSO, JWT tokens, field-level encryption
Integration: Microsoft Graph API, Office 365, automated AD sync with conflict resolution

Agentic Tech & Workflows

Claude
Python
MySQL
Microsoft Azure

Greenfield with integrations BETSOL

Domain-Agnostic Monitoring & Remediation Framework

AIgis is a domain-agnostic framework for building intelligent monitoring and remediation pipelines. It defines collectors to gather signals from any domain, apply deterministic rules for threshold-based evaluation, and optionally invoke AI models or enterprise LLMs to explain anomalies and generate structured remediation actions. All of this is gated behind human approval before execution.

Key highlights

Domain-agnostic collector protocol: plug in any signal source, get typed outputs
Deterministic rules engine with cascading WARN/CRITICAL severity levels
AI-powered explanation layer for anomaly explanation and confidence-scored action suggestions
Human-in-the-loop approval gate, no autonomous execution
Run history is fed back to the LLM for cross-cycle continuity and pattern detection
Web dashboard + REST API for config, history, and real-time execution streaming

Agentic Tech & Workflows

Python
FastAPI
React
Enterprise LLM APIs
Observability frameworks

Greenfield without integrations BETSOL

DeliveryIQ

DeliveryIQ provides executive-level insights into the health of enterprise-wide delivery, surfacing issues customers may face before they become systemic. DeliveryIQ uses AI to turn raw professional-services artifacts, on-the-job feedback, status updates, emails, and delivery notes, into structured, per-account health scores, trend signals, and QBR-ready reports. Every artifact is anonymized before any AI call, and every AI-generated insight is reviewed and approved by a delivery lead before it reaches an executive, making delivery intelligence consistent, auditable, and compliant by design.

Key highlights:

Privacy Shield masks names, emails, phones, and system IDs into typed placeholders before every AI call, so the model never sees raw PII
Human-in-the-loop approval gate ensures no insight reaches an executive without explicit delivery-lead sign-off, per finding, reversible, and logged
Full audit trail logs every state change atomically, from ingestion to executive view, answering "who saw what, and when" for any account at any time
845+ tests passing across API, Web, and Privacy Shield

Capabilities:

Contributors submit raw artifacts, feedback, status updates, or emails
Privacy Shield runs automatically, identifying and replacing all PII with typed placeholders before any AI call
Delivery leads approve or reject each AI-generated insight individually before it escalates
Approved insights surface to executives as health scores, trend signals, and QBR reports
30+ architectural decisions tracked in a versioned decision log
Supports 3 AI provider modes: Mock, Claude (Sonnet/Haiku), and Local LLM
Zod provides a single source of truth for end-to-end type safety
Roadmap: an admin-configurable agent layer will connect directly to external sources (Jira, CRMs, GitHub) to gather signals automatically through the same Privacy Shield pipeline

Agentic Tech & Workflows

Claude (Sonnet/Haiku)
Local LLM
API & Web app
Zod

Greenfield with integrations BETSOL

HR Assistant: Agentic AI (RAG Architecture)

An AI-powered HR assistant built natively into Microsoft Teams that gives employees instant, accurate answers to natural-language questions using cited HR policies. It also handles Jira Service Management requests, including ticket creation, SLA checks, and approvals, without leaving the chat. By bringing HR knowledge and IT support into one conversational interface, it replaces scattered policy documents and cross-app workflows with a faster, simpler employee experience.

Key highlights

Reduces the operational load of repetitive HR queries across HR teams and employees, cutting time lost to manual policy searches.
Multi-tenant SaaS-ready, one deployment serves unlimited organizations with strict data isolation
Zero-secret production, Managed Identity authentication removes credential rotation overhead and secret-leak risk
Zero misrouted tickets, the JSM plugin auto-classifies and routes service requests to the correct queue
Compliance built-in, full redacted audit trail in Application Insights with no raw PII ever reaching logs or LLMs

Capabilities:

Ingests HR policy PDFs, deduplicates them using SHA-256, generates embeddings with Azure OpenAI, and indexes content securely by tenant.
Every query passes through jailbreak detection and PII/PHI redaction. If either security layer is unavailable, the request is blocked rather than passed to the LLM.
Intelligently routes queries across HR, Jira, JSM, Confluence, and general conversations, then selects GPT-4o or GPT-4o-mini based on query complexity.
Uses Azure AI Search for hybrid vector and keyword retrieval with mandatory tenant-level filtering to prevent cross-tenant data access.
Generates cited answers in a conversational tone and delivers them through rich Adaptive Cards in Microsoft Teams.
Fail-closed security architecture with India-specific detection for sensitive identifiers including Aadhaar, PAN, voter ID, UAN, and IFSC.
Supports HR policies, Jira issues, JSM requests, and Confluence knowledge through a unified routing layer and domain-specific plugins.
Framework-agnostic RAG core supports Teams and web today, with the flexibility to extend to Slack and email without duplicating core logic.
Uses per-user delegated Atlassian OAuth with encrypted token storage, allowing the bot to act on behalf of the employee rather than through a shared service account.

Agentic Tech & Workflows

Azure OpenAI (GPT-4o / GPT-4o-mini)
Semantic Kernel
Azure AI Search
Azure Functions v2
M365 Agents SDK
Entra ID
Jira Cloud
Confluence

Project Complexity Categories

Understanding how project type impacts AI-driven development efficiency

AI IMPACT: HIGH

Greenfield

Without Integrations

Built from the ground up with zero external dependencies.

Highest AI acceleration potential
Minimal integration complexity
Maximum time and cost savings

AI IMPACT: MEDIUM

Greenfield

With Integrations

Net-new solution incorporating external platforms, APIs, or data sources.

AI still drives efficiency gains
Integration design adds scope
Dependency management increases complexity

AI IMPACT: LOWER

Brownfield

Legacy Modernization

Migrating or modernizing existing systems with undocumented or partially documented business logic.

Extensive discovery required
Tribal knowledge dependencies
Highest project risk level

BETSOL Copyright © 2026. ® All Rights Reserved.