The chatbot retainer that prints.
Every business has the same wish: a chatbot that actually answers questions about their business — their products, their policies, their docs — instead of generic nonsense. That's RAG, and modern tools let you build one without a computer-science degree. The model is the same one that makes the automation agency great: a setup fee to build it, then a monthly retainer to host and maintain it — recurring revenue for something a business can't easily take in-house.
01 RAG in plain English
A normal chatbot answers from what the model already knows — which means it knows nothing about this business and will confidently make things up. RAG (retrieval-augmented generation) fixes that: before answering, the bot retrieves the relevant pieces from the client's own documents and answers from those. Feed it their help docs, product manuals, policies, and FAQs, and it becomes an expert on their business specifically — grounded, accurate, and far less likely to hallucinate. That grounding is the entire value you're selling.
02 What you actually sell
Two parts, and the second is the business:
- Setup fee — build the bot, ingest and clean their documents, tune the prompts and guardrails, and deploy it on their website or support tool.
- Monthly retainer — host it, keep the knowledge current as their docs change, monitor answers, and improve it. Their content changes constantly, so this isn't busywork — it's the reason the bot keeps working.
Customer-support bots (deflect repetitive tickets), website FAQ/sales assistants (answer buyers 24/7 and capture leads), and internal knowledge bots (let staff ask the policy/manual instead of bugging a manager). Support deflection has the clearest ROI — "this answers the 40% of tickets that are the same five questions" is a number a business feels instantly.
03 How it's built now
You don't write a vector database from scratch. A whole layer of tools handles the hard parts — you ingest documents, connect a model, set guardrails, and embed a widget. Options range from no-code/low-code chatbot platforms (fastest to ship), to Microsoft Copilot Studio for businesses in the M365 world, to custom builds on frameworks when a client needs something specific. Start with the simplest tool that does the job; move up only when a client's needs force it.
04 Price it — setup + retainer
Drag the inputs to see a realistic 2026 deal shape:
05 The guardrails that keep clients
A client's support bot occasionally gets asked something not in the docs. Best design?
Your move this week
Build a demo bot on your own knowledge base — your résumé, a product's help docs you can find publicly, anything. Get it answering accurately and admitting when it doesn't know. That working demo is your entire sales pitch: show a business "here's yours, trained on your docs," and the setup-plus-retainer close gets easy.
What you can do now
- Explain RAG simply: a bot that answers from the client's own documents, grounded and accurate
- Sell a setup fee (build + ingest + deploy) plus a monthly hosting/maintenance retainer
- Lead with support deflection — the clearest, most felt ROI
- Build with the simplest tool that works (no-code platform, Copilot Studio, or custom)
- Bake in guardrails and a human fallback, keep docs current, and never promise perfection