Node AI.02

AI assistants and chatbots

Assistants that answer from your documents and data, and say when they do not know.

Start a project Take the AI readiness scan

RAG-based AI assistants grounded in your documents and data, for customers or internal teams, on your website, WhatsApp, Slack or Microsoft Teams.

The problem

Customers ask the same questions every day, and the answers are buried in policy PDFs, product sheets and old email threads. Internal teams spend hours searching shared drives for a clause or a number.

A general chatbot does not know your business, and an assistant that makes up answers is worse than none. The answers have to come from your own sources, with the source shown.

What we build

Customer-facing assistants for your website and WhatsApp that answer product, policy and order questions
Internal assistants in Slack or Microsoft Teams that search across documents, wikis and databases
Retrieval pipelines that keep the assistant in sync as documents change
Citations on every answer, and a clear "I do not know" with a handoff to a person when the sources do not cover the question
Conversation analytics showing what people ask, what was answered and where the gaps are

How it works

Documents and data are cleaned, split and indexed. Each question retrieves the most relevant passages, the model answers only from them, and the answer is returned with its sources.

Step 1Collect and clean sourcesWe agree which documents and systems the assistant may use, remove duplicates and outdated versions, and set access rules.
Step 2Index and retrieveContent is chunked and embedded into a vector index, with metadata so answers can be filtered by product, region or user role.
Step 3Answer with guardrailsPrompts restrict the model to retrieved content. Off-topic and unsafe questions are declined politely.
Step 4Test on real questionsWe build a test set from your actual customer or staff questions and measure answer accuracy before launch.
Step 5Launch on your channelsWeb widget, WhatsApp Business, Slack or Teams, with handoff to your support team.

Deliverables

Assistant deployed on the agreed channels
Ingestion pipeline that refreshes the index on a schedule or on change
Test set and accuracy report
Admin view of conversations and unanswered questions
Handoff flow to your team

Typical stack

Python
LLM and agent frameworks
Next.js
Google Cloud

We choose tools to fit your environment; this is where we usually start.

Ways to work with us

Build and HandoverA defined system, delivered, documented and handed to your team.Engagement Managed AI OpsOngoing monitoring, evaluation and improvement of your AI systems.Engagement Forward Deployed EngineersOur engineers inside your team until it runs in production.Engagement

Questions

Can it work on WhatsApp?

Yes, through the WhatsApp Business Platform. You will need a verified business account; we help set it up.

What stops it from making things up?

The model is instructed to answer only from retrieved passages and to say when they do not cover the question. We measure this on a test set before launch and keep monitoring it afterwards.

Can different users see different answers?

Yes. Documents carry access metadata, so an internal assistant can respect the same permissions as your file system or intranet.

Which languages are supported?

Modern models handle English and major Indian languages well. We test the languages your customers actually use before launch.

Adjacent nodes

Part of AI & Agents.

Voice AI agentsVoice agents that pick up on the first ring and book the slot.AI.05 AI voice receptionistEvery call answered, booked or routed, day and night.AI.04 AI data assistantAsk your business data a question in plain English and get the answer with a chart.AI.03 AI-ready data foundationsClean, modelled, governed data, so AI systems give answers you can use.DATA.02

Talk to us about aI assistants and chatbots for your business.

Send a short brief. We reply by email, usually with questions before we propose anything.