Use caseData privacy, banking
4.1 million files, six sources, and a data map nobody trusts. ID scans and doctor's notes sit in old folders any employee can open.
Every source is read in place, nothing moves during the scanEach finding points to a folder, a file and a pageEvery fix has an owner and is checked by the next scan
4.1 M files read where they sit, 0 moved. Each finding keeps its location, its category and the rights on its folder.
source · 6 sources, 3 nights of scheduled wavesSensitive data, wide rights · its owner decides the fix
14,800 driver's license images sit in a 2016 scan folder that every employee can read. Nobody put them there to cause harm: a scan saved to a share, a spreadsheet exported for one campaign, a doctor's note filed next to a vacation planner. The privacy officer cannot set priorities without a map, and the IT director cannot lock a share without knowing what it holds.
This inventory reads the files themselves. The operations engine works through every source in place, the AI identifies personal and confidential information, and the report becomes the work list. The privacy officer checks the findings on a sample, and each data owner decides what happens to their own data.
One screen counts, the other turns the count into places to fix. Underneath both: the scope is agreed before anything runs, the findings are checked before anyone reports them, and every fix belongs to the person who owns the data. Dana's month follows, starting with the list of sources.
A scan is only useful if its scope is clear. Marcus and Dana agree on the sources and the categories before anything runs.
Marcus Reed, the IT director, lists the sources with Dana Whitfield, the privacy officer: the document repository, three network drives, the SharePoint sites of the retail bank and an old scan archive. Each one is connected for discovery, which analyzes content where it sits.
Dana picks what matters for the bank: account and card numbers, Social Security numbers, images of driver's licenses and passports, and health data, which HR must keep confidential and apart from personnel files. The AI will look for each of them in text, scans and images.
She sets a confidence threshold. A finding below it goes to human review instead of the report.
The scan starts on Friday evening and works through the sources in waves. On the scan archive, Marcus limits the load so the old server stays responsive during the day. He follows throughput, errors and remaining volume while it runs.
The scan is finished. Dana turns a long list of findings into a short list of places that need attention.
The discovery report lists each location with the categories it holds and the rights inherited on its folder. Dana sorts on both. One line stands out: \\fs02\Credit\Scans_old holds 14,800 images of driver's licenses, and every employee can read it.
Before reporting anything, Dana opens 50 findings in the viewer. In a training deck, the account number flagged on slide 12 is the example account printed in the bank's own procedure. She marks it a false positive. Discovery had already listed 340 identical copies of the deck, and the same decision applies to them.
The second priority is an HR team folder in Charlotte. The AI found 212 doctor's notes and three medical reports saved by managers next to the vacation planners. The HR repository already has a restricted class for exactly these documents.
A map earns its keep when each line has an owner and an action. The AI proposes, and each data owner rules on their own files.
For each location, the AI proposes one action: restrict access, move the files to the repository class that fits them, or delete copies that duplicate a filed original. Dana reviews the list, changes one move into a deletion, and assigns each line to its data owner with a due date.
Mei Chen, head of credit operations, approves both tasks for Scans_old. The operations engine moves the 14,800 images into the loans class of the repository, with credit operations' rights and the class retention rule, and restricts the old folder. The reconciliation report compares what was expected with what arrived.
At the end of the month Dana presents the inventory: what was scanned, what was found, what is fixed and what is planned. The figures come straight from the discovery report, which exports as it is, and every line opens down to the folder it describes. From now on, the scan runs every quarter.
Every location in the report gets one of three tags. Dana reads the tag first, because it tells her who should hear about the folder next.
Or only false positives, checked by a person on a sample.
What follows: no action, the location is checked again next quarter.Doctor's notes in a team folder, ID scans outside the loans class. The proper home is named.
What follows: a move is proposed, the data owner approves it.The location holds health or ID data and too many people can open it.
What follows: access is restricted first, then the files are moved or deleted.It moves nothing, deletes nothing and changes no folder right. Every change is approved by the data owner, carried out by the operations engine and written into the report line by line.
The AI reads, identifies and proposes. The plan names who acts on each location, and that holds across four million files.
Counted on the first run of INV-2026-01. A bank with more shares, or older ones, will land on other counts.
Where these numbers come from. They are counted on the scenario described on this page, with example data. They are not a customer result.
Marcus runs the servers, Dana owns the map, and the team heads own the files. A scan this size needs all three on board.
Asked where personal data sits, you can only answer from the data map.
A scan of four million files sounds like a threat to servers you have to keep up.
The cleanup plan for your files was written by someone who has never opened them.
With one share. Pick the network drive your teams have used the longest, and run the scan on it alone. The findings on one share are enough to set the priorities for the rest.
The scan plugs into the network drives, SharePoint sites and repository you run today. Nothing has to be migrated first.
Let's talk
Choose one network drive or one repository folder. We will run the scan with your privacy officer and your IT team, and read the findings together.