Skip to content
HHussain C

02Service

RAG & custom AI chatbots

Chatbots and assistants that answer from your own documents, show where each answer came from, and say so when they don't know.

EmailWhatsAppSend a brief

The short answer

Updated

RAG (retrieval-augmented generation) is how a chatbot answers from your own content. It searches your documents first, then the model answers using only what it found, with a link to the source. I build RAG chatbots for support, sales and internal teams, usually in 3 to 6 weeks for $6k to $20k.

Typical budget
$6k to $20k
Timeline
3 to 6 weeks to production
How it's quoted
You get a fixed quote after the first call. Model, vector database and hosting costs are separate.

A chatbot that knows your business, and admits when it doesn't.

Most chatbot complaints come down to two things. It made something up, or it couldn't find an answer that was sitting in a PDF the whole time. Both are search problems, not model problems.

So most of the work goes into the unglamorous layer: cleaning and splitting your documents, searching them well, checking who's allowed to see what, and testing answers against the questions people actually ask. The model is the easy part.

A good fit if

What you get

How a project runs

  1. Week 1

    Audit

    What content exists, who can see what, and the questions people ask most.

  2. Weeks 2 to 3

    Index and prototype

    Your content indexed and a working chatbot, tried by a few people on your team.

  3. Weeks 4 to 6

    Production

    Sign-in, permissions, the chat widget or integration, monitoring and launch.

  4. Ongoing

    Improve

    Fix the questions it misses and tune the search, each change checked against the test set.

Questions people ask

What is RAG, in plain terms?

The chatbot looks things up before it answers. It searches your content for the relevant passages, hands only those to the model, and the model writes the answer from them, with a link to the source.

Should we use RAG or fine-tuning?

For answering from your own content, RAG almost every time. It stays current when documents change, shows its sources and costs far less. Fine-tuning is for teaching a style or a narrow task, not facts.

Where can the chatbot live?

On your website, inside your app, in Slack or Teams, on WhatsApp, or behind an API that your own product calls.

How do you stop it making things up?

It only answers from the passages it found, it cites them, and it's tested on questions where the right answer is "I don't know". Those tests run before every release.

How much does a RAG chatbot cost?

Usually $6k to $20k to build. One clean source, like a help centre, sits at the low end. Several sources, per-user permissions or scanned documents push it up. Running costs are usually a small monthly bill for the model and the vector database.

How long does it take?

3 to 6 weeks to production. You can try a prototype on your own content after about two weeks.

What do you need from us to start?

Access to the content it should answer from, a list of the questions people ask most, and someone who can tell whether an answer is right.

Further reading

Related services

Contact

Tell me what you're trying to fix.

I read every message myself and usually reply the same day. A rough idea is enough to start.

Or send a short brief

Hours

Your hours work for me.

  • San Francisco

    --:--

  • New York

    --:--

  • London

    --:--

  • Berlin

    --:--

  • Dubai

    --:--

  • Singapore

    --:--

  • Sydney

    --:--

  • Surat

    --:--

    Me

I keep flexible hours and move my day to overlap with yours, whether that's a morning standup in New York or a late release in Sydney.