Faster Compliance Checks with AI-Powered Call
Analysis
90% reduction in compliance review time, with
automation that checks 4,000+ calls a month.
INDUSTRY
Medicare compliance SaaS
(Healthcare)
Service
Craft New
Software
Engagement
90 days to
launch
Team
Lean team of
3
ON THIS PAGE
The Outcome – Potential Impact
What this could reach once it ships
90
Days from idea to product launch
4000+
Calls processed monthly
7
Months to a path to profitability
THE CHALLENGE
Every sales call to a senior had to be checked by hand
Medicare Advantage agencies must make sure their agents’ sales calls follow strict
regulations. When seniors (65+) call to enroll in a plan, the conversation has to follow
specific rules that protect this vulnerable group. The rules also change by jurisdiction and
product, which makes reviewing recorded calls very labor-intensive.
The traditional compliance process was manual and slow:
Compliance officers personally reviewed each recorded call
They checked every call against a list of required disclosures and protocols
Calls often ran 1–2 hours, so each review took a long time
Manual review capped how many calls could be checked
Agencies had no scalable way to monitor agent performance
Our client set out to help Medicare Advantage agencies thrive in this complex regulatory environment, and came to us with a vision: automate compliance and quality analysis with AI.
THE APPROACH
Transcribe, score, verify
The industry needed a tool that could analyze every conversation, verify
compliance, score each call objectively, surface coaching insights and scale with call
volume. We built it on Incubyte’s deploy-readyAI framework.
01
A solid technical foundation
We built the web app with Ruby on Rails and React,
hosted on AWS, so call recordings, transcription and scoring could scale reliably.
02
Accurate transcription
The AI framework uses Deepgram’s speech-to-text to
transcribe each recording and build the dataset for analysis.
03
Consistent scoring
We defined a fixed set of question categories that every
transcript is evaluated against, so scoring stays consistent and relevant.
04
AI response categorization
A tailored prompt gives Anthropic’s Claude clear
instructions and context to analyze and categorize each response.
05
Verified results
AI responses are checked against predefined logic and business
rules, then shown as clear, actionable scores in the UI.
06
Lower running costs
We replaced AWS Transfer Family with an open-source SFTP
server on AWS, cutting the processing cost of incoming data from $250 to $5
THE OUTCOME
A lean team, a profitable product
A team of three, working test-first with a quality-first mindset, delivered a high-quality
product on time. The client was on a path to profitability within 7 months.
Infrastructure costs down 60%, from $500 to $200 a month, with simpler deployments.
4 months from project start to onboarding the first customer.
Client onboarding in hours.
4,000+ calls processed monthly, with headroom to scale to 30,000.
What began as a tool for insurance is now a platform that can analyze sales calls, customer
support and other regulated conversations across industries.
STACK
Tech
Ruby on Rails
React
AWS
Deepgram
Anthropic Claude
Open-source SFTP
TDD
Team
Akshay Vadher
Software Craftsperson – Tech Advisor
Sanket Dhokte
Technical Product Craftsperson
Yogendra Jaiswal
Software Craftsperson
Weighing a rewrite?
There’s a third path. Most of our work begins with a thirty-minute conversation.