The client who sends everything at once: bulk upload and smart matching

You send a client ten clearly separated requests. What comes back is a single email with the subject “Here you go” and a zip of forty files: bank statements, a handful of invoices, two versions of the same lease, a payroll run, and a scan of something you did not ask for. No item numbers. No note about which file answers which ask. Just the whole drawer, tipped out at once.

Some clients will always work this way. They are busy, they gather everything in one sitting, and sending it all together feels like being helpful. In an email-and-folders world, their generosity becomes your sorting problem. Someone on your team now spends an afternoon opening files, guessing what each one is, and matching it back to the request it belongs to.

Why the dump is a sorting nightmare in email

The trouble is not the volume. It is that a pile of files has no idea what it was sent to answer. Email carries no memory of your ten requests, so nothing connects the third attachment to the item that asked for it. Your team reconstructs that link by eye, one file at a time, and every mistake, such as filing the FY2024 statement against the wrong period, quietly enters the record.

Do that across a portfolio of clients who all bulk-dump near a deadline, and the sorting alone becomes a real cost. The work is not skilled. It is just unavoidable when the tool has no concept of what was asked.

An unassigned pool, then suggested matches

Clivanta handles the dump by giving it somewhere structured to land. A client can bulk upload a whole set of files at once, and anything that does not obviously belong to a single item drops into an unassigned pool for that Engagement rather than scattering across inboxes.

Then the AI pre-checks go to work. For each uploaded Document, they read the file and suggest which Request Item it most likely answers, using the document type, the document date, and the entity name they extract. The forty-file pile arrives pre-sorted into proposals: this looks like the Q1 bank statement, this looks like the lease you requested, these two look like duplicates of each other.

The pile is still forty files. But instead of a blank sorting job, your team sees a first pass that is usually close, with the evidence for each guess sitting right next to it.

You confirm the match — the AI never finalises it

This is the part that matters, and it is deliberate. The AI suggests. A person decides. Every proposed match is confirmed or overridden by a human before it becomes real. The Firm can work through the pool and accept the assignments, correct the ones that are wrong, and split the odd file that answers something unexpected. The Client can do the same from their side of the portal, tidying their own upload against the named requests.

What the AI will never do is finalise an assignment on its own. It does not quietly move a Document onto a Request Item and call it settled. The pre-checks are advisory only, across the whole product: they classify, extract, flag and suggest, but they never accept, reject, move or conclude. A wrong guess is a suggestion you decline in a second, not a mistake buried in the record that you have to unpick later.

That restraint is what makes the assistance trustworthy. You get the speed of a machine reading forty files, without handing it authority over what the audit file says.

Duplicates caught before they clutter the record

Bulk uploads are where duplicates breed. The same bank statement gets included twice, or a client re-sends the whole pack “just to be safe.” Clivanta detects duplicates by file hash, so an identical file is flagged as a copy rather than filed as a fresh Document. You are not accepting the same evidence twice, and the item’s history stays clean.

When a file genuinely is a corrected version rather than a duplicate, it lands as a new attempt against the item, with the earlier one retained as a previous attempt. Nothing gets silently overwritten, and nothing gets silently deleted.

Attachments by email, routed to the right place

Some clients will not go near a portal at all. They reply to your request notification and attach the files, because that is how they have always worked. Clivanta meets them there too: attachments sent by reply are ingested and routed to the right Pack, then treated exactly like a portal upload. They enter the same unassigned pool, get the same suggested matches, and pass through the same human confirmation before anything is settled.

So the client who insists on email is not a hole in your process. Their attachments arrive as structured Documents against the right Engagement, not as another thread for someone to reconcile by hand.

The pattern underneath

Bulk upload and smart matching solve a specific, familiar problem, but they follow the same principle as everything else in the product. Structure the work as explicit Request Items. Let the machine do the reading and the sorting it is good at. Keep every decision that touches the record in human hands. The product overview shows how the pieces connect.

The client can still send you the whole drawer at once. Now it sorts itself, and you just say yes.