Enterprise ontologySecond Brain · by LapisAI

Ask the record. Get precedent back.

Redwood builds your company’s ontology from its own documents: an open, linked map of its people, clients, projects and know-how. Ask from Claude, Copilot or ChatGPT, and every answer cites the passage it came from, under each person’s own access.

Live ontology · 104,812 notes
Siapa yang pernah menangani proyek data warehouse untuk bank, dan partner apa yang terlibat?
Redwood · Second Brain
Membaca catatan…
Dua proyek data warehouse untuk bank tercatat, pada 2023 dan 2025.
Rina Hartono memimpin modernisasi data warehouse Bank Sentosa Raya pada 2025.1
Microsoft menjadi partner delivery, dengan dua cloud architect hingga go-live November 2025.2
[2] Handover report 2025-11.pdf · hlm. 3 Delivery partner Microsoft Indonesia provided two cloud solution architects from design through go-live.
Dimas Prasetyo menjadi data architect di data mart regulasi Bank Lintas Nusa pada 2023.34
Seluruh perusahaan Ditinjau oleh Sari Wulandari Diperbarui 7 Okt 2026
The ontology People 4,820 Companies 2,316 Projects 1,212 Domains 186 Meetings 21,904 Documents 74,374 English and Bahasa Indonesia · Jakarta region or on-premises · open Markdown
01 Cited answers

Every claim shows its source.

Answers arrive as a card inside the AI tool your people already use. Each sentence ends with a numbered citation; tap it to read the exact passage, then open the note or the original document.

Citations on every claimAnswers, notes and drafts all cite the page they came from, so a careful reader can check any sentence in one tap.
AI stays labelledAI-written content is marked until a named person reviews it, in Redwood, in chat and even in Obsidian.
Only what each person can openRedwood answers from the sources each person already has access to, and never hints that anything else exists.
Cited answer card, light theme, with four numbered sources and a trust line
02 The Second Brain

One ontology of what the company knows, who knows it and how it connects.

One trusted page per client, project, person and domain.Facts are grouped by theme with the year first, each with its citation. Related people and projects sit alongside, and the review label says who checked it and when.
See who worked with whom, and on what.Open the graph from any note and it focuses there, two hops out. At a hundred thousand notes it settles into clusters by type and domain, with a list view one click away.
Agents with real roles, working from a project’s sources.The requirements analyst states its plan, shows each step, and explains any step that fails before you retry or skip it. Drafts stay labelled AI until a named person approves them.
A knowledge team keeps the record honest.Merge duplicates, settle contradictions and review AI-written notes. Every fix says how many notes it will change before it applies.
Entity note page for Bank Sentosa Raya Dark graph in focus mode around Bank Sentosa Raya Agent tab with a plan, tool steps and a failed step offering Retry or Skip Merge review comparing two company notes
03 How it works

From shared folders to a record that adds a ring every year.

01Connect

Point Redwood at SharePoint sites, file shares and meeting transcripts. Each source keeps its original permissions.

18,306documents read
02Compile

Documents become Markdown notes with links and footnotes, reviewed by your knowledge team. The vault opens in Obsidian too.

104,812linked notes
03Ask and draft

People ask in chat or in Redwood. Project agents draft deliverables that stay labelled AI until approved.

212projects on record
04 Governance

Built for the people who answer for the data.

Permission domainsMap directory groups to vaults, and preview exactly what one person’s answers can draw on before you publish a change.
People-note policyChoose which work facts Redwood may keep about people, publish the notice, and handle access and erasure requests with a full log.
Audit to your SIEMEvery question, answer, draft, curator action and admin change is recorded and streamed to the tools your security team already uses.
Models, and where they runPick the model for each job and see where it processes data: Jakarta region, another region, or on-premises for sensitive workspaces.
Yours to takeExport the whole vault as Markdown any time, or sync it to your own Git repository.
Two languages, one recordEvery screen works in English and Bahasa Indonesia, from the answer card to the admin console.
Phone layout of a cited answer in Indonesian, with the source open in a bottom sheet

Readable on a phone, in either language. Dates and numbers switch too: 7 Okt 2026, 18.306.

Like a redwood, it adds a ring every year.

The longer a company runs Redwood, the more of its experience stays findable, cited, and its own.