Back to browse
AI that replaces the first 15 minutes of every client call

AI that replaces the first 15 minutes of every client call

by vira28·Feb 12, 2026·3 points·0 comments

AI Analysis

●●SolidSolve My ProblemNiche GemShip It

Voice-cloned lead qualifier for service pros, but still competing against generic AI agents.

Strengths
  • Real insight: first 10-15 minutes of calls are templated—targeted problem worth solving.
  • Custom RAG + structured knowledge graph instead of off-the-shelf LLM wrappers.
  • Multi-persona support and doc ingestion (websites, PDFs, transcripts) reduces setup friction.
Weaknesses
  • Voice agents for lead gen already exist (Gong, Chorus, custom Twilio+LLM stacks); voice cloning adds polish but not a new category.
  • Landing page heavy on testimonials, light on technical differentiation or concrete metrics (conversion lift, call accuracy).
Category
Target Audience

Service professionals: consultants, CPAs, advisors, agencies, and real estate professionals seeking lead qualification automation.

Similar To

Gong · Chorus · Synthesia

Post Description

Hi HN — I’m Viggy, founder of MyClone.

We originally built a marketplace connecting startup advisors with founders. After ~6 months, we had hundreds of recorded discovery calls and transcripts.

We noticed something interesting: The first 10 minutes of most advisory calls were almost identical. • What do you do? • Who is this for? • Pricing? • What information do you need from me? • Am I a good fit?

So we pivoted.

Today, MyClone lets service professionals (consultants, CPAs, etc.) deploy an AI voice agent that: • Answers FAQs in their voice • Runs structured intake conversations • Qualifies leads • Routes good prospects to a human

Technical details

Under the hood: • Multi-source ingestion (websites, docs, PDFs, media transcripts) • Structured knowledge graph + chunked semantic retrieval • Custom RAG pipeline (not using off-the-shelf frameworks) • Prompt templates generated dynamically based on objective (FAQ vs intake vs qualification) • Guardrails to prevent hallucination beyond uploaded corpus

We learned quickly that generic “chat with your docs” doesn’t work for client-facing scenarios. The hard part is: • Controlling tone • Avoiding overconfidence • Structuring conversations instead of answering open-ended questions • Handling partial / messy user inputs • Maintaining low latency in voice mode

Still early, but we’re seeing adoption from solo and small firms who want a 24/7 “AI associate” for first-touch conversations.

We do share our technical learnings in our blog - https://www.myclone.is/blog Happy to answer technical questions, RAG architecture details, failure modes, or things that broke during the pivot.

— Viggy

Similar Projects