For teams whose pipeline lives in spreadsheets.

Each path solves a different starting friction. They're fully interoperable — start in one, switch to another, combine them. Records land in the same place either way.

Three record types — the same three that anchor every other CRM source we support:

• Accounts — companies in your active pipeline. Fields: name, domain, industry, headcount band, ARR band, HQ, owner, tier, notes.
• Contacts — people at your accounts. Fields: name, email, phone, title, associated account, buying role (champion / economic buyer / technical evaluator / operational buyer / gatekeeper), LinkedIn URL, notes.
• Opportunities — active deals. Fields: name, associated account, stage, amount, close date, probability, owner, forecast category, champion contact, notes.

You don't need to fill every field. Management OS uses what you provide and gracefully degrades when fields are missing — a deal with no amount still appears in the Brief; it just doesn't contribute to the forecast roll-up.

Every ingest path lands in a preview pane before anything writes to your workspace. You see:

• Per-record confidence scores (high / medium / low)
• The source snippet each record was extracted from
• Counts by record type (5 accounts · 12 contacts · 23 opportunities)
• Warnings and errors before you commit

Low-confidence records are flagged with an amber pill so you can scan them first. Nothing commits until you click the explicit Commit N records → button. Path 4 — templates — bypasses the Claude column-mapper because the column structure is already canonical; you still see the preview before commit.

Same source (idempotent): Re-uploading the same CSV updates existing records in place rather than creating duplicates. Each record gets a deterministic external ID (ssht_{type}_{12-hex}) derived from its content, so a second upload of the same row finds and upserts the original.

Cross-source (surfaced for review): If you later connect HubSpot and your CSV-sourced "Acme Corp" matches a HubSpot "Acme Corporation," Management OS scores the match (normalized name similarity + domain overlap + Levenshtein distance) and surfaces high-confidence matches as merge candidates. Low-confidence matches don't merge — they remain as separate records until you confirm.

We never auto-merge across sources. The leader confirms every merge.

Application database: Supabase Postgres (Management OS's production database, US East). Encrypted at rest (AES-256), encrypted in transit (TLS 1.3). Workspace-scoped row-level security (RLS) means one customer's data is never visible to another.

Source provider tag: Records ingested via this surface are tagged source_provider = "spreadsheets_csv" in sense_records. This lets the Brief generators distinguish CSV-sourced data from CRM-sourced data when relevant — for example, forecast-accuracy reads weight CRM data higher because CRM data has stage transitions captured natively.

Inference layer: Anthropic API (Claude models). The extraction step (conversational text → structured records, CSV columns → field mapping) runs through Anthropic. Anthropic does not train on customer data per their API terms.

Original input retention: We retain the raw input (conversational text or CSV content) alongside the extracted records so you can re-run the extraction if you want to revise the mapping later. Retention follows the same schedule as all workspace data: retained while active, 30 days after cancellation, then permanently deleted.

1. Sign up at mgmt-os.ai and complete onboarding through the CRM step
2. On the CRM picker, choose Spreadsheets / Manual
3. Pick a starting path: Just tell me (conversational), Upload CSV, Google Sheets (v1.1), or Templates
4. Provide your data. Management OS extracts records and shows you a preview
5. Review the preview — confidence scores flagged in amber if low. Edit input and re-extract if needed
6. Click Commit N records → to write to your workspace
7. Your first Weekly Brief will draw on this data within the hour

Management OS has Apollo and Apify enrichment available for other surfaces, but the Spreadsheets ingest path does not auto-enrich your records. Three reasons:

• Auto-enrichment from external data sources risks surfacing inaccurate or stale third-party data inside what you treat as your source of truth
• Enrichment consumed silently means you pay third-party data costs without explicit consent
• The first Brief is more valuable when grounded in what you actually know — not in fields the system guessed

When you're ready to enrich (e.g., fill missing phone numbers on contacts), Settings → Enrichment lets you opt in per record type. Cost and source provider are shown before any external call fires.

The Spreadsheets path is a bridge, not a destination. When your team is ready for a real CRM, Management OS exports your accumulated data into the native import format for HubSpot, Salesforce, or Pipedrive — column names mapped to each CRM's expected import schema, ready for their bulk-import wizard. Ships in v1.1.

On graduation, your historical Briefs, Account Plans, and Coaching Briefs remain in Management OS. Going forward, data flows from your new CRM via OAuth; your previous Spreadsheets-sourced records remain queryable with the source_provider tag so you can audit the transition.

1. Open Management OS → Settings → Integrations
2. Find Spreadsheets / Manual in the connected integrations list
3. Click Disconnect — this stops the surface from accepting new data
4. To delete the historical records, use Settings → Data → Delete spreadsheets-sourced records

To remove all historical Briefs and data after disconnect, email privacy@mgmt-os.ai with your workspace identifier. We process deletion requests within 30 days.

Security incident reporting: security@mgmt-os.ai
Privacy / data subject requests: privacy@mgmt-os.ai
Integration support: support@mgmt-os.ai

For general questions about Management OS, see our Support page or Manifesto.