Skip to content
Sortup
Book a call

14 days · Knowledge Assistant

Your team's documents, answered with citations, inside the tools they already use.

For teams that answer the same questions from PDFs, Notion, Drive or Confluence every week.

Book a 20-minute fit callGet a free AI Teardown

Start to finish

14days

  • Cited answers
  • 50-question eval
  • Slack, Teams or web

The system

Knowledge Assistant, end to end.

  • Already yours
  • We build
How the Knowledge Assistant works, in 5 steps
  1. 01

    Your sources

    PDFs, Notion, Drive, Confluence

  2. 02

    Ingestion

    Chunking, permissions

  3. 03

    Retrieval

    pgvector, filters

  4. 04

    Eval gate

    Your agreed question set

  5. 05

    Answer

    With its source passage

Builds layers

  • 02Data and retrieval
  • 04Evals and guardrails
  • 05The product
All five layers

What you get

Six deliverables you can check.

  • Ingests PDFs, Notion, Drive and Confluence, respecting permissions
  • Every answer cites the passage it came from
  • Tested against 50 of your real questions before launch
  • Runs in Slack, Teams or a web widget
  • Usage, cost and failure dashboard
  • Documented, and you own the code

How it works

14 days, four stages.

Knowledge Assistant timeline: Sources and eval set, days 1–2; Ingestion and retrieval, days 3–7; Interface and guardrails, days 8–11; Final eval and handover, days 12–14.

  1. 01Days 1–2

    Sources and eval set

  2. 02Days 3–7

    Ingestion and retrieval

  3. 03Days 8–11

    Interface and guardrails

  4. 04Days 12–14

    Final eval and handover

What it is, and what it is not.

It is

  • Retrieval over your own documents, running in your own accounts
  • An answer that says so when your sources do not cover the question

It is not

  • A chatbot widget with a long prompt
  • A fine-tuned model or a new source of truth
  • A black box: every answer is traced, every eval run saved

AI work

Similar systems, already shipped.

Built by Sortup or one of our co-founders, as labelled. Not results of this offer.

All client work
The team delivered a complex super app with e-wallet, e-commerce, logistics integrations, and a dynamic commission engine. Their communication was clear, execution was reliable, and they consistently provided solution-driven results throughout the project.
Simon SprouleCo-founder, Gather

Scope

What changes the scope.

  • Number of sources and documents
  • The permission model
  • Languages to cover
  • Slack, Teams, web or all three
  • Hosting and data residency

FAQ

Questions about this one.

What if our documents are messy?

Most are. We sample your sources on days 1–2, so if something needs cleaning before it can answer well, you hear it then, not on day 14.

Which model does it use?

The one you choose: OpenAI, Anthropic, Google or open weights. You hold the API account; we build everything around it.

What does "tested on 50 real questions" mean?

Your team writes 50 questions they actually ask, with the answer they expect. We run the assistant against that set by day 7 and again before launch, and share every result, including the misses.

What happens after day 14?

You own it. Run it yourselves with the handover docs, or keep us on for more sources or a second assistant.

Is the Knowledge Assistant the right first step?

No pitch deck. An engineer, and a straight answer.

Book a 20-minute fit callGet a free AI Teardown
Hours
We work yours, in the GCC, the UK and EU or the US
NDA
We sign yours before the first call
IP
You own the code from the first commit