AI services · Knowledge assistant

Free pilot on your knowledge base. Then your team can ask it anything.

Connect a sample of your Confluence, Drive, Notion or SharePoint. We index it, ground it, and show you what a cited, permission-aware answer looks like on questions your team asks every day. At no cost. If it holds up, we roll it into Slack, Teams or a web widget and keep it accurate.

Answers with citations Respects permissions Your data, your models
Sample pilot: 2,400 documents
Documents indexed 2,400
Answer accuracy 96%
Answers with a citation 100%
Time to find an answer minutes to seconds
Hours saved per employee / week 4+
Grounded with citations Permission aware Local or cloud models Always up to date Audit-ready
Why this exists

Your team already has the answers. They just cannot find them.

Every company past a certain size has the same problem. The answer to almost any operational question exists somewhere in the company's Confluence, Drive or shared inbox. Finding it requires knowing the right person to ping, the right search terms for the wiki, or the right folder in the shared drive. Most of the time people give up, ask a colleague who is also guessing, or make a decision without the information.

A knowledge assistant solves the retrieval problem without rebuilding your information architecture. It reads your existing sources, answers questions with a citation to the exact page and paragraph, and respects the permissions that already govern who should see what. You do not get a generic AI that hallucinates policies. You get a grounded answer from your actual documents, every time.

1/5
Knowledge workers spend up to a fifth of the week just looking for information that already exists inside their organisation. That time is recoverable, and the assistant turns it back into productive work.
How it works

Four steps. The pilot result comes back in days.

No platform to buy, no documents to reformat, no model to train from scratch. The pilot connects to a sample of your existing sources.

1

Connect a sample of your sources

One or two of the sources where your team stores knowledge: a Confluence space, a Drive folder, a Notion database. You keep full control over which content is in scope.

~1 day to connect
2

We index and ground

We chunk the content, embed it into a vector index, and map every chunk back to its source document and page so every answer carries a citation your team can check.

A few days
3

Accuracy review

You test the assistant on the questions your team actually asks. We show you answer accuracy, citation rate, and where the gaps are. You see the numbers before committing to anything.

30 minutes
4

Roll out into Slack or Teams

If the pilot holds up, we connect all your sources, wire in your permission model, and surface the assistant where your team already works.

Scoped after the pilot
What the engagement covers

Grounded answers, permission-aware retrieval, always current.

Every answer traces back to a source your team can verify. The assistant only shows what the person asking is allowed to see. And the index stays fresh as your knowledge changes.

Grounded answers with citations

The assistant retrieves the relevant passages first, then composes its answer from them. Every answer includes a link to the exact page and paragraph it came from, so your team can verify it rather than trust it blindly.

Permission-aware retrieval

The retrieval step checks permissions before surfacing a passage. A new hire asking about compensation policy does not see the board's restricted documents. Permissions are inherited from your existing source systems, not reinvented.

Always current

The index syncs with your sources on a schedule you control. When a document is updated or a new page published, it is re-indexed within minutes, so answers reflect the latest version of your knowledge, not a stale snapshot.

Want the depth? How retrieval-augmented generation works, why citations beat a raw chatbot, how permissions are enforced, and how the index stays accurate are all broken down in the complete Knowledge Assistant guide.

Sample answer

This is what your team sees.

A plain language question. A grounded answer with citations to the exact source documents and pages. No need to know which Confluence space to search, no need to ping a colleague, no need to guess.

The card on the right is illustrative. Your pilot is built from your own sources.

Knowledge assistant · Slack
Grounded answer 3 citations 0.4s retrieval
Employee question
What is the process for requesting access to the production database?
Grounded answer with citations
Submit a request via the internal portal under Security > Access Requests, with your manager and the data owner copied. The data owner has 48 hours to approve or deny. Approved requests are provisioned with read-only access by default; write access requires a second sign-off from the Engineering Lead.

[Onboarding policy, p.4]   [Security handbook, section 3.2]   [Engineering runbook, Access chapter]
Where it pays off

Every team that runs on internal knowledge.

The pattern is always the same: a question that someone in the organisation already knows the answer to, trapped in a document nobody can find fast enough.

TeamThey askThey get
New hiresHow do I do X during onboarding?A grounded answer from the onboarding docs, not a Slack ping
SupportWhat is the fix for this known issue?The current knowledge article, cited to the exact section
SalesDoes the product do Z? What does the contract say about Y?An accurate answer with the source document linked
ComplianceWhat is our policy on W?The exact policy clause, with the document and revision date
EngineeringWhere is the runbook for service V?The right page in Confluence, with the relevant section highlighted
Engagement tiers

From a single pilot to a running knowledge layer.

The pilot is always free. After that you pay for the build and the ongoing accuracy work. The exact scope is set after the pilot, because it depends on your sources, your permission model, and how many teams you want to reach.

Pilot

Prove it

Free · one or two sources
  • Index a sample of your existing knowledge
  • Measured answer accuracy on real questions
  • Citations on every answer
  • An estimate of hours saved per person per week
Build

Ship the assistant

Talk to us · fixed build
  • Everything in Pilot
  • Full source index with permission enforcement
  • Slack or Teams integration
  • Web widget option
  • Accuracy and citation rate dashboard
Operate

Keep it accurate

Talk to us · monthly
  • Everything in Build
  • Ongoing index sync as your knowledge changes
  • Accuracy monitoring and gap reporting
  • Model and retrieval upkeep
  • Monthly answer quality report
How pricing works. The pilot is free and you keep the accuracy numbers. The build is a fixed scope. Ongoing operation is priced per month based on the number of sources, the index size, and how many teams use the assistant. We set the numbers after the pilot, because the right scope depends on your sources and your accuracy bar.
Proof

We ship working software, not slideware.

The knowledge assistant is a development and data engineering engagement, which is the work we are built for. A few that map closely:

Request

Get a free pilot on your knowledge base.

Tell us your work email and where your team's knowledge currently lives. We reply within 24 hours to set up a sample run. The pilot is free and you keep the accuracy numbers.

  • Built on your own sources, not a canned demo.
  • Grounded citations on every answer, measured accuracy.
  • Permission-aware from day one, respects your existing access model.
  • We build it, we keep it current. No platform licence to buy.

Prefer a longer brief? Use our form

Get a free knowledge assistant pilot

Two fields. We do the rest.

No credit card. No commitment. We reply within 24 hours with next steps.

FAQ

Questions we get every week.

How is this different from a chatbot? +
A generic chatbot answers from its training data. A knowledge assistant retrieves passages from your specific documents, then grounds its answer to those passages with a citation you can click. It cannot make up an answer it cannot source, which is what makes it usable for support, compliance and operations work where precision matters.
Does it leak information across permissions? +
No. The retrieval step is permission-aware: the assistant only draws from documents the person asking is allowed to read. A new hire asking about budgets does not see the finance team's restricted files. Permissions are inherited from your existing source systems, not rebuilt from scratch.
Where do the models run? +
Your choice. For teams with strict data residency requirements we run the embedding and language models inside your own infrastructure, so your documents never leave your environment. For teams that prioritise speed and simplicity we use a hosted cloud model. The data path is agreed and written into the engagement before any document moves.
How does it stay current as our knowledge changes? +
The index syncs with your sources on a schedule you control, typically daily or near-real-time via webhooks. When a document is updated or a new page is published, it is re-indexed within minutes. The assistant answers from the latest version of your knowledge, not a snapshot from six months ago.
What sources can it connect to? +
Confluence, Google Drive, Notion, SharePoint, Jira, Zendesk, GitHub, databases, PDFs, internal web pages, and most sources that expose an API or a file export. The pilot typically starts with one or two sources, then expands once accuracy is confirmed.
What does it cost? +
The pilot is free. After that, pricing depends on the number of sources, the volume of documents indexed, the channels the assistant surfaces in, and whether you need us to operate and keep it accurate over time. We quote a fixed scope after the pilot, because the right build depends on your sources and your accuracy bar.

Give your team the answer before they finish typing the question.

A free pilot on your real knowledge base. Grounded citations. Permissions respected. Then we ship it into Slack or Teams and keep it accurate.

Get a free pilot
Free pilot · You keep the result · 24h reply

Related reading

Related reading

Related reading

Get a free pilot