Put your data on the record.

Find My Data is customer-hosted data-ownership and sensitive-data intelligence software. It finds data that may belong to a business domain — across Microsoft 365, Google Drive, Box, Azure storage, on-prem file shares, and SQL databases — and asks the people who understand it. AI drafts the record in pencil; your people confirm it in ink. Evidence behind every entry, and nothing leaves your deployment.

FREE FOR TWO USERS · CUSTOMER-HOSTED · SINGLE-TENANT · YOUR AZURE, YOUR KEYS, YOUR RECORD

Plate I. — As proposed (pencil) and as confirmed (ink).

1 confirmation · 12 identical copies updated · 0 files moved

Fol. 2 · The four entries

Find it. Confirm it. Connect it. Govern it.

A closed loop that puts the people who understand the data — not only security engineers — at the center of the record.

No. 001

Find it

Incremental, resumable scans across your estate — Microsoft 365, cloud storage, on-prem file shares, and databases — combine policy-grounded AI, detectors, metadata, permissions, and exact and near-duplicate fingerprints to surface candidate matches — every score carrying its evidence.

No. 002

Confirm it

Data-domain owners answer one honest question — is this yours? — with the evidence in front of them. AI proposes; authorized humans decide. This is the one entry a machine cannot make, and every decision is recorded, reversible, and propagates to identical copies.

No. 003

Connect it

Confirmed data links to duplicates, exposure, sensitivity labels, owners, and risk through a typed relationship graph — with the reason for every connection written beside it.

No. 004

Govern it

Propose a Purview label or remediation, approve it under separation of duties, execute idempotently, verify against the source, and keep a tamper-evident audit trail. No autonomous action — ever.

Fol. 3 · Two hands keep this book

One record, two kinds of keeper

Double-entry governance: the people who understand the data write it, and the people accountable for it can prove it.

For data-domain owners

You'll be asked one honest question.

Is this yours? — with redacted evidence, confidence, and reach in front of you. Plain speech, no security jargon, keyboard-fast once you've earned it.

  • Review candidate matches with the evidence and its confidence beside them
  • See where your data lives — outliers, duplicates, exposure, and reach
  • Confirm, correct, or reject; one decision can settle twelve identical copies
  • Ask plain-language questions, scoped to exactly what you own
For security, privacy & GRC

You'll get a defensible record.

Not a dashboard of guesses — a ledger of who confirmed what, on which evidence, under which authority, and what happened next.

  • Ground the taxonomy in your own policy, with citations
  • Map information types to regulations and sensitivity labels
  • Approve material classifications and remediation under separation of duties
  • Every finding, answer, and action exposes its evidence and provenance
Fol. 4 · Standing rules

Nothing asserted without evidence

Find My Data installs inside your Azure environment. The vendor plane handles licensing and coarse fleet health — and nothing else.

§ 1

Content stays in your boundary

Files, excerpts, paths, fingerprints, embeddings, identities, permissions, findings, and Analyst conversations never leave your environment to license or support the product.

§ 2

Evidence before assertion

Every classification, relationship, finding, and AI answer exposes its evidence, confidence, provenance, and version. Nothing is a black box.

§ 3

Human authority

AI proposes; authorized people confirm ownership, material classifications, regulatory mappings, model releases, and remediation.

§ 4

Least privilege & identity

Microsoft Entra sign-in, deny-by-default capabilities scoped to business domains, separate read and remediation identities, and a per-tenant hash-chained audit trail.

§ 5

Minimized, purpose-flagged evidence

Only bounded excerpts and derived features are retained, purpose-checked on read. Detected identifiers are redacted from stored previews.

§ 6

Governed AI

Provider-neutral gateway on your own account — Azure OpenAI or Anthropic — with a deterministic default, versioned templates, untrusted-content delimiting, per-task cost visibility, and no autonomous mutation.

Plate IV. — The boundary, as enforced. One line crosses; it carries no content.
Fol. 5 · For the avoidance of doubt

Corrections are entered, never erased

What this product is not, struck through in the record the way a ledger corrects itself — visibly.

Entered in errorThe correction
A shared content SaaS — an "app" holding your files— struck. see correction
Customer-hosted enterprise software, installed in your own Azure tenancy
A consumer device or people locator— struck. see correction
A way for enterprises to find, confirm, and govern business data in Microsoft 365
An autonomous AI that acts without approval— struck. see correction
Evidence-backed and human-authorized end to end — AI proposes, people decide
Fol. 6 · The estate as surveyed

The whole estate, honestly staged

Microsoft 365, Google Drive, Box, Azure storage, on-prem file shares, and SQL databases — one pipeline, one record. Where a live path still waits on a Microsoft approval, the record says so plainly — the same rule the product itself lives by.

Plate VI. — The estate as surveyed. Every wing keeps one record.

One pipeline for every source

Files, mail, chat, blobs, file shares, and database tables flow the same classification, evidence, owner-review, and findings pipeline — and duplicate detection works across sources, so the same contract is one record whether it lives in SharePoint, Drive, or a share.

Explainable intelligence

Detectors, keyphrases, exact and near-duplicate fingerprints, semantic similarity, and a calibrated ensemble — every score arrives with its evidence, and re-scoring never re-touches the source.

Measured scale

One million assets measured — not extrapolated — on the zero-dependency SQLite default; an optional PostgreSQL backend is load-validated to two million with pgvector semantic search.

Deploys in your environment

Single-tenant in your Azure (managed identity, Key Vault, container + Bicep), Microsoft Entra sign-in, least-privilege Sites.Selected scanning, and your own AI provider account — Azure OpenAI or Anthropic. See the documentation for exact capability states — nothing claims to be live until it is.

Fol. 7 · The tariff

Two tiers. Every feature in both.

The platform is never metered by feature — only by people. Your first two users are always free, on either tier; each user beyond them is $49 a month.

Free — the first two keepers

$0 — up to two users, for as long as you like.

The complete platform, not a trial: every connector, every feature, full scale. Install it, survey the estate, show the record to the organization — on your own clock, with no card on file.

  • Everything in the product — no feature gates, no evaluation deadline
  • Customer-hosted in your own tenancy, like every deployment
  • Built for the land phase: install, scan, demonstrate, secure funding
Paid — when the register grows

$49 per user / month — beyond the first two.

Enroll the data domain owners, auditors, and executives who belong in front of the record. The first two users stay free on this tier too — you pay only for each person beyond them.

  • The same platform as Free — the tiers differ only in people
  • Only active users count: 90 days without a sign-in and a user rolls off
  • Adjust the count any time; the deployment's own reported number trues it up

TARIFF, WORKED · 10 USERS → FIRST 2 FREE + 8 × $49 = $392 / MONTH · FULL SCHEDULE

Fol. 8 · Rule off the day

The day's entries are made.
Open your own register.

Create a verified business account, receive a signed entitlement — free for your first two users, for as long as you need — and install Find My Data in your own environment, where your record belongs.

On the demonstration record — the demo environment is a work of fiction, kept truthfully: fictional synthetic data ("Meridian Grove Holdings"), clearly identified as a demonstration, never a customer's installed environment, and holding no real data.