Your policies, product manuals and FAQs, answerable on your website or on WhatsApp. Every answer cites the passage it came from. When the documents don't cover a question, it says so instead of guessing — which is the part most chatbot demos skip.
The assistant on this site's home page is a RAG chatbot built the way yours would be: it answers questions about my work from an index of every article here, shows where each answer came from, and refuses questions its sources don't cover. Ask it something it can't know and watch it decline. The same index is also published as an MCP server, so it can be queried from inside Claude.
Open the live assistant →Starting prices for a standard scope. The final number is fixed in writing before work starts.
from $800
Answer questions from your PDFs, policies, product docs or knowledge base — on your website, WhatsApp, or both.
from $2,000
A CRM built around how you sell, with AI doing the repetitive parts: summarising conversations, drafting replies, flagging leads that went cold.
Smaller pieces of work: $15–$30 an hour. Wiring AI into systems you already run (ERP, internal tools) is scoped and priced on the business services page.
PDFs, Word files, web pages or a knowledge base are split into passages and embedded into a vector index.
Each question is matched to the most relevant passages. Below a similarity threshold, nothing is passed on.
The model answers only from those passages and cites them. No passages, no answer — it says it doesn't know.
A chat widget on your site, a WhatsApp number through the WhatsApp Business API, or both.
A RAG chatbot trained on your own documents starts at $800 for a standard scope: your documents indexed, cited answers, a refusal path, and a website or WhatsApp channel. The fixed price is agreed in writing before work starts. Running costs (model usage and hosting) are estimated up front, because they are what surprises people later.
Yes. Strictly it is not retrained: your documents are split, embedded into a vector index, and the chatbot retrieves the relevant passages for each question and answers only from them. That is retrieval-augmented generation (RAG), and it is why every answer can cite its source.
It answers only from what it retrieved, and when nothing relevant clears the similarity threshold it says it does not know rather than guess. You can try exactly this behaviour on the live assistant on this site, which runs on the same pattern.
Yes. WhatsApp is supported through the official WhatsApp Business Platform API, alongside or instead of a chat widget on your website. WhatsApp charges its own per-conversation fees, which are separate from the build.
From $2,000 for a CRM built around your sales process, with AI summaries, draft replies a person approves, and WhatsApp and email in one place. Larger integrations into systems you already run are quoted separately.
$15–$30 an hour for smaller pieces of work, depending on scope. Fixed-price projects are quoted as a single number before anything starts.
You do. The code lives in your repository from the first commit and runs on infrastructure you control, so you can maintain it yourself or hire anyone else to.
Tell me what documents the chatbot should answer from and where people should reach it. You'll get a fixed price and an estimate of the monthly running cost.