The pain

You're a solo coach / consultant / SaaS founder. You answer the same questions every week:


Each one is 2 minutes to answer. You get 30 a week. That's an hour. But the real cost is the interruption — these messages land mid-flow, mid-call, mid-everything.

Who has it

Anyone with a website, a product, or a service that has surface area. Coaches with programs. SaaS founders with pricing pages. Consultants with packages. The signal is "I keep writing the same answer."

The fix

A custom RAG (retrieval-augmented generation) bot trained on your own content — not a generic ChatGPT plugin. Lives on your site. Answers in your voice. Escalates the 5% it can't handle to your inbox.

Setup (1 day, mostly waiting)

  • Gather your knowledge base. FAQ doc, blog posts, sales page copy, transcripts of past sales calls, any onboarding emails. Aim for 20–60k words.
  • Embed it. Pinecone (free tier) or any vector DB. Chunk into ~500-token pieces.
  • Build the bot. Small Flask or Next.js app. The prompt is the leverage:

    • > You are [your] business assistant. Answer using ONLY the provided context. Match [your] voice — direct, plain English, no buzzwords. If the context doesn't cover the question, say "I don't have that info — let me ask [name] directly" and capture the question for review. Never make up specifics like prices, dates, or guarantees.

  • Deploy. Vercel/Render/your VPS. Embed on your site.
  • Set up the escalation feed. Unanswered questions → your inbox once a day.

    • Running it (5 min, daily)

    Read the daily escalation digest. The questions it couldn't answer are the gaps in your knowledge base. Add the answer to your knowledge base. Re-embed. Done.

    What it looks like running

    Visitor lands on your pricing page at 11pm. Asks: "Can I pause the program if I have a busy month?"

    Bot replies in your voice: "Yes — you can pause for up to 4 weeks per year, no questions asked. Most people pause around the holidays. Want me to walk you through how that works?"

    You wake up. The bot tells you it handled 14 questions overnight, plus one it escalated: "Do you work with teams of 3+ people?"

    You answer that in 90 seconds, add it to the knowledge base, and tomorrow the bot handles it too.

    Why this works

    This is a Assistant-tier workflow — AI as Remote Assistant. The buyer is asking the heaviest trust question yet: "Can AI talk to my customers without me watching?"

    It works specifically because of the constraint "answer only from the provided context, never invent." That's what makes the trust transaction safe. A generic chatbot will hallucinate prices. A RAG bot anchored to your knowledge base will not.

    The compound benefit: every escalated question makes the bot smarter. Within 8 weeks, it handles 90%+ of inbound and you only see the genuinely interesting questions.