Insurance document management where the PDF stays attached to its policy.
A policy record without its document is a claim with no evidence behind it. Insoora keeps every uploaded PDF in private storage, attached to the policy it was read from, named after that policy, and reachable only through a request that re-checks who is asking.
- Every PDF kept with its policy record
- Named for the policy, not for the sender
- No public URL, ever
- Duplicates matched by content, not filename
On this page
The problem nobody calls document management
Every policy PDF arrives with a useless name
Policy.pdf, Document(3).pdf, IMG_20250114_WA0007.jpg, 576575767567.pdf. A folder of four hundred of those is a folder you cannot search.
Once a policy is confirmed, Insoora gives its documents a name derived from the policy itself: the policy number, the vehicle registration for a motor policy or the policy type otherwise, and the year — so a motor document is filed as something like 576575767567-GJ06KQ9835-2026.pdf and a health one as HLT-2026-4455667-HEALTH-2026.pdf. A second document on the same policy gets a v2 suffix so two files in one list are distinguishable.
The name is metadata rather than a rewrite. The insurer's bytes are never touched and the stored object is never renamed, because renaming an object means copying it and deleting the original, and a delete is the one operation that can lose a policy. The uploaded filename is kept verbatim for audit and is never parsed for policy data — it is customer-supplied text and is treated as such.
When a part of the name cannot be established — no policy number read, an unregistered vehicle, no usable date — the document keeps the name it arrived with and carries a reason, which the portals show as a badge rather than hiding. Fixing the data and re-confirming names it.
- A name derived from the confirmed policy, not from what the sender called the file
- The original filename kept for audit and never used as data
- A "name needs review" badge instead of a silently wrong name
What it does not do
There is no rename button and no folder structure you control. Documents are filed against policies and customers, which is the only hierarchy there is — you cannot invent your own.
Getting a document back
Six ways a PDF is found again
Finding the document is the actual job. Storing it was never the hard part.
By customer
Open the policyholder and every policy they hold is there, each with the document it was read from.
By year
A customer's policies stack year by year, so last year's cover and the document that proved it are one click apart.
By policy number or insurer
Search across the book by customer, insurer, policy number or status, then filter and sort the result.
By vehicle
A motor policy's registration number is stored as a field and is searchable, with a vehicles view in the back office.
By document kind
Schedules, certificates, bonds, renewal notices, endorsements, premium receipts, cover notes and proposals are classified separately rather than lumped together as "document".
By the customer themselves
A policyholder with a portal login downloads their own documents without asking you for a copy, which removes the request rather than speeding it up.
Who can reach a document
No policy PDF has a public URL
The mechanism, stated plainly, because "secure document storage" means nothing on its own.
Stored objects are private and nothing in the store is publicly readable. Object keys are generated by the server from the document's own identifier and never from a filename a user supplied, so there is no guessable path and no way for a crafted filename to decide where bytes land.
Every download goes through a request that re-derives the caller's scope on the server before any bytes are served — an agent gets their agency's documents, a policyholder gets only the policies assigned to them, and a sub-agent gets what the agency granted. Opening a document in the in-app viewer additionally needs a signed link that is bound to that user and expires in five minutes.
The practical consequence is the one worth stating: a link to a policy PDF is not a link you can forward. It stops working, and it was never valid for anybody but the person who requested it.
- Private storage, server-generated keys, no public read access
- Caller scope re-derived on the server for every download
- Viewer links user-bound and valid for five minutes
- All traffic over HTTPS
What it does not do
There is no document-level sharing with people outside the platform, no public link you can send, and no watermarking. Sharing a policy with someone who has no account means downloading it and sending it yourself.
The same policy, uploaded twice
Why duplicates matter more in this product than in most
In most document systems a duplicate is untidy. Here it is also expensive: policy processing is metered, so a second copy of the same file would mean a second charge for reading a document that had already been read.
So an upload is matched against the account's existing documents on the content itself — a SHA-256 hash of the bytes together with the byte length — rather than on the filename. A match returns the document already stored instead of writing a second copy or spending a second policy credit.
The matching is scoped to the owning agency. Two different agencies holding the same insurer-issued file are not duplicates of each other, and neither one's upload tells the other anything.
What this does not catch
It catches the same file twice. It does not catch the same policy arriving as two different files — a schedule from the insurer and a photograph of the printed copy are different bytes and will both be stored, correctly, as two documents on one policy.
Nor does it merge records. If the same policy was uploaded twice before it was confirmed and ended up as two policy records, joining them is a manual correction.
Where the documents actually sit
Stated because it is the thing most often assumed wrongly. The application and its database are hosted in Singapore, and uploaded PDFs are held in Cloudflare's object storage — not in India, even though the product is built in Gujarat for Indian insurers.
The Privacy Policy sets out which providers are involved and what each one does. There is no published retention schedule yet, and nothing is deleted automatically, because the platform is built to keep policy records findable for years.
Not a general document system
What this is not built to be
A document management system in the general sense does several things this one deliberately does not.
Insurance documents only
A classifier rejects anything that is not an insurance document before extraction runs — an RC, an invoice, a bank statement, a payslip, an agreement. You cannot use Insoora as a filing cabinet for the rest of your paperwork.
No version control or e-signature
There is no check-in and check-out, no revision history on a file, no annotation layer, no approval workflow and no signing. A revised policy is a new document on the policy, not a new version of the old file.
No full-text search inside documents
Search covers the fields that were extracted and the customer record. It does not search the body text of every stored PDF, so a phrase that appears in a policy wording is not findable.
No external sharing and no retention policy
Documents cannot be shared by link with anyone outside the platform. Nothing is deleted automatically and there is no published retention schedule — a deletion request is handled by hand and some records may still be kept where law, tax or a backup cycle requires it.
The screens
Documents, where you actually meet them
Three views, drawn with invented data: arriving, filed by year, and in a customer's own hands.
Arriving
A batch in progress, with the state of every file — read, needs review, failed — so a backlog is a finite piece of work.
Filed by year
A customer's policies stacked year on year, each with the original document kept against it.
In the customer's hand
The policyholder's own portal on a phone, where they download their documents without asking you for a copy.
Frequently asked questions
Questions this page should answer
Keep reading
Where to go next
The pages that cover the next part of this, with what each one adds.
Upload the folder nobody has opened since last March.
Each document is read, named for its policy and filed under its customer.