AI for the whole organisation. Scoped to each department.

Why we built it

Organisations want their people to use AI, and their compliance and IT teams want to know exactly what it saw, what it wrote and who approved it. General assistants answer the first need and ignore the second. HOOM was built for both: the assistant staff actually use, inside the boundaries the organisation actually needs.

What it does

Draft from templates, in the department’s voice.

Each department gets task templates — letters, memos, research briefs, announcements, market commentary. Fill a short form, watch the draft stream in, refine it in one click. Drafts follow the department’s configured tone and can pull passages from its own library so they reflect real internal material.

Ask questions of your own documents.

Upload PDF, Word and Excel files into a department library and ask in plain language. Answers carry citation chips that reveal the exact source passage; if the library does not cover the question, HOOM says so rather than inventing an answer.

Compliance review before anything leaves.

Regulated templates are badged up front. Authors submit to a reviewer queue; reviewers approve (signed with their name) or request changes; approval unlocks export. Editing an approved document sends it back for review, and nobody can approve their own work.

Files with explicit AI permission.

A private drive with sharing along the org structure, plus a deliberate “Share with AI” grant. The assistant only uses files that are both shared with AI and visible to the person asking, and a single page shows everything the AI can see, with one-click revoke.

Email that is drafted, never sent, by AI.

The assistant fills approved organisation email templates only, addresses internal recipients only, and always produces a draft for human approval.

Inside Outlook.

Run templates and pull up drafts from a task pane in Outlook on web, Windows and Mac, signed in silently through the existing Microsoft session.

Access the way IT expects.

Passwordless email sign-in or single sign-on (Microsoft Entra, Google Workspace, Okta, any OIDC provider), automatic provisioning from approved domains, and instant revocation that ends every active session.

Audit, retention and erasure.

A full audit log across sign-ins, settings, every task run, review, export, upload, chat and AI-share grant, with monthly tamper-evident archives. Separate retention periods for documents and audit records, complete purge on expiry, and one-click right-to-erasure for an individual.

Data protection by default.

Inputs that look like NIC, phone or account numbers trigger a warning or a hard block before anything reaches the AI, per policy. API keys are encrypted at rest and never written to the audit trail. Reporting is aggregate only — no per-person leaderboards.

Administration without a consultant.

Guided setup wizards for SSO, email and AI provider with test-connection checks; a template editor with test runs and health checks; branding with an accessibility contrast check; quotas and rate limits; a system-email designer; and an outbound email log with the exact body delivered.

Choice of AI provider.

Anthropic or OpenAI, with tunable retrieval behaviour. [CONTENT REQUIRED: whether an on-premise or sovereign deployment option is offered]

Who it is for

Stockbroking, banking, insurance and other regulated organisations where AI-assisted work must be reviewed, attributable and auditable — and any organisation that wants departments to use AI without sharing everything with everyone.

Built the way we build everything

Time-saved reporting on every run. Crash recovery for interrupted drafts. Clear error messages that name the reason and the next action. Light and dark themes in the organisation’s colours. The details are the product.

Status

[CONTENT REQUIRED: live / pilot / available on request; first deployments; pricing model]