Your documents, answerable

Marella turns the documents your organisation already has into an asset you can question — privately, with citations, under your rules.

Book a demo Security details

Marella — Policy review

Private

What does our retention policy say about client files after an engagement closes?

Client files must be retained for seven years after the engagement end date, then reviewed for secure disposal. Files under legal hold are excluded until the hold is lifted.

¶ Retention policy · §4.2 ¶ Engagement handbook · p.18

Ask across 4,120 documents…

Source · Retention policy.pdf

Every answer cites the exact passage it came from. One click opens the source.

From scattered documents to answers

Step 1

Ingest

Drives, inboxes and the tools you already run — SharePoint, Slack, Notion and more, even scanned or messy files
Step 2

Understand

Documents parsed into passages, entities and concepts
Step 3

Connect

The knowledge network links what belongs together
Step 4

Answer

Agents respond with citations back to the source

01

Ask, and check the answer's working

Chat runs against your corpus, not the open internet. Citations sit inside every answer and open the source passage beside the conversation, so checking the evidence is part of reading it.

A dedicated validation agent can re-check answers against the source documents themselves, flagging any claim the evidence doesn't support.

Answer

The notice period is 90 days for enterprise agreements signed after March 2024.

¶ MSA v3 · §12.1 ¶ Renewals note

When the documents don't answer, Marella says so instead of guessing.

02

A knowledge network, not a search index

As documents arrive, Marella extracts the people, companies, topics and concepts inside them and links them into a network your team can browse, correct and build on. Curation tools — ontology, consolidation — keep it clean as it grows.

Knowledge network — Entities

1,284 entities · 5,317 links

All PeopleCompaniesDocumentsTopics

AC

Alderline Capital

Company

risksupplier
14
QR

Quarterly risk report

Document

compliance
9
MP

M. Patel

Person

engagements
6
DP

Data protection

Topic

UK GDPRpolicy
21

03

Agents, skills and templates

Give each team an agent that knows its documents, follows its playbook and writes in its format. Skills make capabilities reusable; templates make outputs consistent; scope decides what each agent may touch.

An orchestrator draws on a registry of specialist agents — research across your knowledge base, the web, PubMed and legislation; standards-compliance and audit-readiness checks; calculation, vision and document generation.

Agent Scope: Legal · 1,204 docs

Contract review

Checks agreements against the negotiation playbook, flags deviations with citations.

Agent Scope: Operations · 342 docs

Policy assistant

Answers on current policy versions only; archives are excluded from scope.

Skill Used by 3 agents

Clause extraction

Pulls defined clause types into a structured table with source references.

Template Output format

Board brief

Two-page summary: findings, evidence, open questions — the same shape every time.

04

Access that mirrors your organisation

Departments and roles decide who can see and do what — the same access model your organisation already runs, applied to AI. Every action is recorded in an audit trail your compliance team can read.

Read security & trust

Organisation — Access & audit

Logged

Roles

Compliance
Read + annotate

All departments
Client teams
Read + chat

Own department
Administrators
Configure

Org-wide

Audit trail

09:41 — agent 'Policy QA' answered · 2 citations
09:38 — j.owens viewed Retention policy §4.2
09:31 — dept 'Advisory' scoped to 312 documents

Decision traces show what the AI did and why; the audit log records every security-relevant event.

Every surface, one system

The workbench your teams live in, and the controls your admins need — in one place.

Chats

Conversations with agents across your corpus

Media library

Every document, searchable and versioned

Entities

People, companies, documents and topics

Concepts

The ideas your organisation reasons with

Ontology

Your taxonomy — curated, not imposed

Agents

Configured assistants with scoped access

Skills

Reusable capabilities agents can draw on

Templates

Consistent outputs for repeatable work

Departments

Access scoped to your structure

Decision traces

What the AI decided, why, and what happened

Assets

Generated documents, spreadsheets and reports

Integrations

MCP-based connections to your tools