AI assistants

An assistant that actually knows your business

Ask a question and get an answer from your own documents, with the source. For your team or for your customers, on web chat, WhatsApp or inside your ERP.

At a glance

  • Answers from your own documents (RAG)
  • Internal knowledge assistants and customer support bots
  • Web chat and WhatsApp
  • Private or open-source models for data privacy

The problem we solve

Your knowledge is stuck in folders

Your company already knows the answer. Finding it is the hard part.

ask, don’t search
  • The same questions get answered again and again, by your most senior people.
  • Policies and procedures sit in PDFs that nobody opens.
  • New staff need weeks before they can work on their own.
  • Customers wait for office hours to get a simple answer.
  • Generic chatbots make things up, and you can’t trust what they say.

What's included

What you get

An assistant you can trust, because it shows its sources and knows when to say “I don’t know”.

  • Your documents, indexed

    We load your PDFs, spreadsheets, web pages and database records, and keep them up to date.

  • Answers with sources

    Every answer points to the passage it came from, so people can check it.

  • Access control

    The assistant only uses documents the person asking is allowed to see.

  • Honest limits

    When the documents do not contain the answer, it says so instead of guessing.

  • Hand-over to a human

    For customer bots, tricky questions go to your team with the whole conversation attached.

  • Insight into what people ask

    See common questions and gaps in your documents, so you can improve them.

Capabilities

What we build

The same engine, shaped for different jobs.

  • Document Q&A (RAG)

    Ask questions across thousands of pages, contracts, manuals or regulations. Get answers with citations.

  • Internal knowledge assistant

    An always-available colleague for HR policy, procedures, product details and onboarding.

  • Customer support bot

    Answers common questions at any hour, collects details, and passes complex cases to your team.

  • WhatsApp assistant

    Meet customers where they already are: orders, status, FAQs and appointment requests.

  • Web chat widget

    A branded chat on your website, with your tone of voice and your answers.

  • Assistant inside your ERP

    Ask for stock, sales or balances in plain language, right where your team works.

How it works

From your documents to a cited answer

This approach is called retrieval-augmented generation, or RAG. In plain words: find the right passages first, then write the answer from them.

  1. Your documentsPDFs, spreadsheets, web pages and records.
  2. IndexedSplit into passages and made searchable.
  3. QuestionSomeone asks in everyday language.
  4. FoundThe most relevant passages are retrieved.
  5. AnswerA clear reply, with links to its sources.

Data privacy

Your documents stay yours

Many of our clients can’t send their documents to a public AI service. We build for that.

  • Private or open-source models

    Run on open-source models on your own hardware, for example through Ollama. With a private setup, documents and questions never leave your servers.

  • Access rules

    The assistant only uses documents the person asking is allowed to see, so confidential material stays confidential.

  • Sources you can check

    Every answer shows where it came from. If the documents do not contain the answer, the assistant says so.

Our process

How we build an assistant

We test with real questions before anyone else sees it.

  1. Choose the job

    We pick one clear use: who asks, what they ask and what a good answer looks like.

  2. Gather and prepare

    We collect your documents and data, clean them and decide who can see what.

  3. Build and test

    We build the assistant and test it with real questions from your team, checking every answer.

  4. Launch

    It goes live on the channel you chose, with a human fallback and clear limits.

  5. Improve

    We review what people ask, fix gaps in the documents and tune the answers.

Tech we use

Proven tools, chosen for the job

We pick boring, reliable technology for the foundations and modern tools where they help.

Models

  • Open-source LLMs via Ollama
  • Hosted models where policy allows

Retrieval

  • RAG
  • Document chunking and embeddings
  • Vector search

Application

  • .NET Web API
  • React + Vite
  • WhatsApp and web chat channels

Deployment

  • Your own servers or private cloud
  • Docker

Related work

See it in a real project

Taxfolio’s chat answering what the VAT registration threshold is for a UAE business, with the source filing and a human review badge shown under the answer.
AI assistantTax and compliance

Taxfolio

A tax research agent with a chat on top. It keeps tax rules in one reviewed ledger, and every answer shows the exact filing it came from. It runs on open-source models on local hardware, so data stays private.

Read the case study

FAQ

Questions we often hear

A general chatbot answers from what it learned on the internet. Our assistants answer from your documents and data, show the source, and say when they do not know. That makes them far more useful, and far safer, for business use.

Not sure what you need?

Answer a few questions and get a suggested set of modules and a readiness score. It takes about two minutes.

Take the ERP readiness check

Got an idea?

Let's sketch your ERP together

Book a free call. Tell us how your business runs today. We’ll tell you, plainly, what we would build and what it would take.