
That on-premises office server or shared drive full of old folders may be worth real money. You can sell shared drive data, de-identified, to AI builders who need years of genuine business documents, estimates, forms and job files to train models.
Plenty of established businesses never moved fully to the cloud. The S: drive, the NAS in the closet, the Windows file server that has quietly collected twenty years of job folders: these are some of the least-seen document archives anywhere, which is exactly why they interest model builders.
Why model builders care about on-prem archives
Cloud platforms have made some business content easier to find and license. Local servers are the opposite: walled off, rarely exported, and often older than any SaaS account the company holds. For a team training operational models, that scarcity is the selling point.
Buyers use shared drive corpora to:
- teach document agents how real estimates, bids and work orders are structured;
- train classifiers on messy, inconsistent naming the way humans actually save files;
- model long-running processes across many years and many versions of the same form;
- improve retrieval systems that must find the right file in a cluttered hierarchy.
Older, specialized archives from fields like home services and trades are especially hard to source elsewhere.
Privacy: what stays behind
We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.
On a local server the scoping step matters more than anywhere else, because old drives tend to mix everything together. Before any processing, you and your IT contact mark which top-level folders are in scope. Payroll, HR, medical, customer personal files and anything you are unsure about stay out.
Inside the folders that remain, identifiers such as names, addresses, phone numbers, account numbers and signatures are removed or replaced with consistent tokens. Scanned images with personal details are excluded unless they can be reliably cleaned. The structure and the business language remain; the people do not. For contractual or regulatory questions about your specific records, talk to your attorney.
Header sample for a file server
A header sample is a short list of field names describing your inventory. It contains no records and no file contents. For a shared drive, the columns typically come from a simple directory listing:
- relative_path_depth
- top_level_folder_type
- file_name_pattern
- file_extension
- size_kb
- date_created
- date_modified
- owner_role
- is_scanned_image
- duplicate_group_id
Your IT contact can produce this from a basic scripted listing. The header sample guide explains what to include and what to leave out.
What pushes the price up
- Age: a server running since the early 2000s holds a historical arc that newer cloud accounts cannot match.
- Working documents over media: estimates, forms and spreadsheets are worth more than photo dumps and installers.
- Consistent use: folders organized by job or client, used daily, tell a coherent story.
- Niche industry: trades, construction, manufacturing and professional services archives are scarce.
- Volume of contributors: more staff means more variety in writing and process.
According to founder Jeff Cline, small companies have gotten over $100,000 for their data, and multiple buyers bid so the highest offer is visible to you. Results vary; not a guarantee.
The process, start to finish
- Tell us the basics. Size, oldest folder date, number of users and your industry go into the bid request form.
- Provide field names only. Your IT contact exports a header sample of the file inventory.
- Watch buyers compete. Offers come in and you see them side by side.
- Scope, scrub, sign, get paid. In-scope folders are de-identified on your side, the license is signed, data is delivered, and payment follows.
In Jeff Cline's experience, bids return within 72 hours when the data is correct and a header sample is provided, and a typical deal closes in about 90 days. Results vary; not a guarantee.
The clock is running
AI labs are buying unique operational data now because they still need it. Once models have learned enough from enough archives, an additional file server teaches them less and prices soften. An old drive sitting in a closet earns nothing; licensed while demand is high, it might fund the next upgrade. Get a free valuation before the window narrows.
How to request a bid for your Office server / shared drive data
- Ask your server manager for size. Get the total size of the shared drive and the date of the oldest top-level folder.
- Note users and industry. Estimate how many people save files there and what line of business the folders cover.
- Fill in the bid form. Submit tool, years, users, storage and industry on the get-a-bid page.
- Export a header sample. Have IT run a directory listing and send only the column names, no records.
- Pick the best offer. Compare competing bids, then scope, de-identify, license and deliver.
Frequently asked questions
Is an old office server really worth anything to AI companies?
Often, yes. Local archives are rarely available to model builders, so long-running folders of estimates, forms and job files can be scarce training material. Value depends on age, content type and industry.
We don't have an IT department. Can we still do this?
Yes. You need someone who can report total size and the oldest folder date, and later run a simple directory listing. Many small offices use their outside IT provider or a tech-comfortable staff member.
What about HR and customer files on the same drive?
They are excluded. We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control, and sensitive folders are scoped out before processing begins.
Do we have to move the server or give anyone remote access?
No physical move is needed. Processing works on a copy of the in-scope folders, and you control what leaves the building after de-identification.
How long does a sale take?
According to founder Jeff Cline, deals typically close in about 90 days, with first bids often inside 72 hours once a correct header sample is in hand. Results vary; not a guarantee.
Do we lose ownership of our files?
No. You license a de-identified copy under a written agreement; your originals stay on your server. Have your attorney review license terms before signing.
Updated November 19, 2026.
OpenDataPolicies.org connects businesses with buyers of de-identified operational data. We are not a law firm, broker-dealer or investment adviser, and nothing on this site is legal, tax or financial advice. Statements about amounts, timing and outcomes are Jeff Cline's statements based on past transactions. Results vary; nothing here is a guarantee of any bid, price, buyer or closing. Every deal depends on the data, its quality, its rights and buyer demand at the time.