Skip to content

Internal AI assistant

An internal assistant that answers from the documents your team already maintains (HR policies, standard operating procedures, product specifications, post-mortems, support history), with a citation back to the source paragraph on every answer. Self-hosted or deployed inside your own cloud tenant. Nothing trains on your data; nothing leaks into a shared vendor account.

Starting from€2,300
Timeline2–5 weeks depending on the number of documents
How we billOne-time, paid in milestones

Deliverables

What you hold in your hand at the end.

The actual files, accounts, videos, and runbooks you walk away with.

  • 01Running assistant in your environment (self-hosted, AWS, Azure, or GCP)
  • 02Connectors to your document sources with incremental sync on a schedule
  • 03Admin dashboard showing top questions, retrieval quality, and flagged answers
  • 04Evaluation harness with a baseline accuracy score and a regression gate
  • 05Training session for your team on how to maintain the knowledge base

What's included

The work itself.

In-scope activities. Everything here is part of the quoted price. No line-item surprises.

  • Document ingestion pipeline: Google Drive, SharePoint, Notion, Confluence, PDFs, tickets
  • Vector store (pgvector, Qdrant, or your choice) deployed in your cloud or in one we manage for you
  • Cited answers: every response links back to the source document
  • Role-based access: engineering sees engineering docs, HR sees HR, etc.
  • Admin panel for reviewing answers, flagging hallucinations, re-indexing
  • Audit log of every question asked and which documents were retrieved

Price tiers

How price and timeline scale.

Same shape of work at every tier. The higher the tier, the more it covers. We agree the right tier on the call.


Add-ons

Extras you can roll into the build.

One-off fees. Add any, all, or none. The calculator shows the running total and you can share a link to your exact configuration.

  • Analytics setup
    GA4 · conversions · dashboards
    +€200
  • Migration from existing site
    Content · URLs · redirects
    +€500
  • Accessibility audit (WCAG)
    Contrast · keyboard · screen-reader · fixes list
    +€300

Not in scope

What this service doesn't include.

Said up front, so nothing surprises anyone at invoice time. Most of these can be scoped in separately. Say the word.

  • Documenting processes that aren't written down yet (separate discovery scope)

  • Public-facing chatbots: see AI chatbot for that

  • Training or fine-tuning a foundation model from scratch


Who this is for

Typical situations that fit.

  • Case 01

    Onboarding new hires without burning senior-engineer time on policy questions

  • Case 02

    Support team looking up product specs and past ticket resolutions in seconds

  • Case 03

    Engineering team searching across Notion, Drive, and GitHub for past decisions


Common questions

Things clients actually ask.

Does my data leave the company?

Only if you want it to. Default deployment runs in your cloud tenant or in one we manage for you. Embeddings and generation can route to your own Azure OpenAI / AWS Bedrock endpoint, so no data crosses into a shared vendor account.

Will it hallucinate?

Less than a raw chatbot, because every answer cites the source it came from. If the source isn't there, the answer isn't either. The eval harness catches regressions before they hit production.

Who can see what?

Role-based access at the retrieval layer. A sales user asking about the current pricing plan gets the answer; the same user asking for the HR grievance policy doesn't see it. Audit log records every query.


Pairs with

Services we tend to ship together.


Want this scoped for your business?

30 minutes, no slide deck. We ask the questions that decide whether the engagement fits, and by end of day you have a written summary and an indicative price. The fixed price comes with the written scope after discovery.