Sign in to Soma

All signals in. One coherent response out.
Sign in to access your workspace.

—
Switch role
Query
Create

Soma Intelligence

Ask anything about your knowledge base.

Scope
Create from query
Select a format →
From your query
Run a query in the chat — sources and context will appear here ready to create from.
Output Format
Audience
Depth
Drive: checking…
SharePoint: checking…
Brand voice loaded

Regulatory Review

Review content pieces submitted for approval. Approve or request revision.

Admin dashboard
Manage knowledge sources, team access, and brand configuration.
Documents indexed
—
across active sources
Pending review
—
users & disclosure
Last sync
—
nightly 23:00 UTC
Brand templates
—
Active users
—
across functions
Knowledge sources
Google Drive Required
—
documents indexed
Not connected
SharePoint Optional
—
documents indexed
Company website Optional
Crawl your public website and ingest approved pages automatically as Tier 1 public external documents. No manual upload needed.
Company knowledge pack Recommended
Foundational context injected into every AI response — strategy, products, regulatory framework, brand guidelines, style guide.
Team & access
Team members
—
active users
Disclosure flags
—
Auto-classified. Flagged only when signals actively contradict each other.
Sign-in & security
Configure how your team signs in, and set the maximum document sensitivity that may be shared with the AI provider.
Brand & output
Brand templates Required
Upload your master .pptx and .docx files so every generated output inherits your brand colours, fonts, and layouts.
Content library
—
output formats · audience personas · voice profile
Content output
—
approved pieces
— pending review
— drafts
Maintenance
Re-tag documents
Re-runs AI auto-tagging on all documents. Use after changing the tagger prompt or adding new products. Can be slow on large libraries.
Backfill document tiers
Assigns authority tier (T0–T4), tier weight, disclosure level, and scope to documents that are missing these fields. Run once after upgrading from a pre-Session 10 build, or if documents were imported without tier metadata.
Backfill ingestion method
Tags existing documents with how they entered Soma (Drive connector, SharePoint, manual upload, URL paste, or approved content). Run once after upgrading from a pre-Session 14 build. Only affects documents missing the ingestion_method field — safe to re-run.
Run embargo sweep
Checks all pre-congress and pre-publication documents whose embargo date has passed and promotes them to public. Runs automatically every night at 01:00 UTC — only use this button to trigger it manually ahead of schedule.
Backfill PPTX images
Extracts images from PPTX and DOCX documents that were uploaded before Session 17 and do not yet have entries in the extracted_assets library. Documents with no Storage copy (pre-Session 13) are skipped.
How Soma works
Quick reference for admins

When a document is added to the knowledge base — by upload, connector sync, or URL — Soma reads the full text and uses AI to extract structured metadata. This is called tagging.

Document added
AI reads content
Tags extracted
Saved to KB
What gets extracted
Title
A clean title derived from the document's content, not the filename.
Phase III KEYNOTE-522 results
Content type
The functional category — what kind of document it is.
published_literature
Products
Which products in your portfolio are relevant. Drives query scoping.
Nexavir
Topics / flags
Key themes used to filter the KB: efficacy, safety, biomarker, mechanism…
efficacy · PFS · OS
Function scope
Which teams can see this: scientific, regulatory, medical, marketing, sales.
scientific · medical
Disclosure level
Soma's initial classification of who the document may be shared with.
regulatory_filed
Tags are suggestions — not locked. Edit any field directly on the document card in the knowledge base. Your corrections are applied immediately.

Every document has an authority tier — a trust level that tells Soma how much weight to give its content when answering questions or building content pieces.

Tier definitions
T0
Company core
Mission, strategy, product portfolio, regulatory framework, brand style guide. Always injected into every AI call — never competes with other sources.
5×
T1
Published / MLR-approved
Peer-reviewed papers, approved clinical data, MLR-cleared content pieces. Highest-trust external evidence.
3×
T2
Internal validated
Clinical study reports, regulatory filings, validated field data. Reliable internal evidence, not yet external-facing.
2×
T3
Internal draft
Working documents, field notes, meeting summaries, slide drafts. Useful context, lower certainty.
1.2×
T4
Unvalidated external
Competitor intelligence, raw web content, unreviewed external reports. Included for awareness only — never cited as evidence.
0.6×
How weight affects answers
T0
5×
T1
3×
T2
2×
T3
1.2×
T4
0.6×
Tiers are assigned automatically by Soma. You can override any tier from the document card in the knowledge base. T0 documents are admin-only and locked from deletion.

Disclosure levels control who a document — or the content generated from it — can be shared with. Soma auto-classifies every document on ingest. Regulatory users can override any classification.

Levels — least to most restricted
Public
Freely shareable. Published papers, company website, press releases.
Regulatory filed
Submitted to a health authority. Shareable with HCPs/payers in appropriate context.
Internal only
Internal teams only. Not for external distribution. Default for most uploads.
Pre-congress/pre-pub
Embargoed until a public date. Soma auto-releases at the date you set.
Restricted
Sensitive material. Excluded from AI output unless a regulatory user explicitly approves its use.
Confidential
Board / legal-restricted. Never included in any AI output regardless of settings.
NDA-protected
Partner or third-party NDA material. Isolated from all retrieval by default.
Soma never sends Confidential or NDA-protected content to any external AI provider, regardless of what other settings are configured.
When Soma flags a classification with low confidence it shows a ⚠ on the document card. Regulatory users can review and override from the document detail panel.

Content pieces — presentations, documents, one-pagers — go through a lightweight approval workflow before they can leave Soma. This is the only gate between internal AI drafts and external distribution.

Lifecycle steps
Draft
Soma generates the content piece from the knowledge base. A watermarked draft (.pptx / .docx) is immediately available for internal review.
Author / any role
Submitted for regulatory review
Author sends the piece to the Regulatory queue. Regulatory sees the full draft, all cited sources with tier and disclosure level, and any schema or missing-image warnings.
Author initiates
Regulatory review
Regulatory reads the draft inline — slides or sections rendered with provenance. Can use chat to request targeted edits before deciding.
Regulatory role only
Approved
Watermark removed. Clean file available for download and distribution.
Needs revision
Feedback sent to author. Piece returns to Draft for editing or regeneration.
Approved
Clean file (.pptx / .docx / .pdf) available. Disclosure level locked to the highest level used in cited sources.
Regulatory approves
Auto-ingested to knowledge base
The approved piece is automatically added to the knowledge base as a T1 document — same trust level as published literature. Future content pieces can cite it as approved messaging.
Automatic — no action needed
Only the Regulatory role can approve or reject a piece. Admins can view all pieces but cannot approve them — role and admin are separate permissions in Soma.

Create content

Informed by your knowledge base and brand guidelines.

Evidence topics
Loading sources…
Your generated draft will appear here.
1 · Choose an output format on the left
2 · Set audience, depth & product
3 · Add a key message, then click Generate →

Notifications

Mark all read
No notifications.
Source