Use caseBusiness banking

Business onboarding and KYC

Novalis Logistics wants three accounts and twelve cards. Of its eight documents, two disagree. Finding which two used to take a week.

The documents are checked against each other on arrivalBoth gaps go to the customer in one messageThe compliance officer decides, with the file in front of her

The assistantKYB check

Name, capital and registered office match across the documents. But the holding's articles name a shareholder the declaration leaves out: Julie Hartley.

source · 8 documents on the case file
The verdict on the file

Two points need the customer · the AI names both, with their pages

Actions on this case file
01

The starting point

In a nutshell

Julie Hartley owns 30 percent of Novalis Logistics through a holding, and the declaration on file never mentions her. The board resolution carries one signature where article 17 asks for two. Today an analyst finds both in week two, laying papers side by side, and the CFO gets a second request for something the bank could have asked on day one. The gaps sat in the file from the start. Only the comparison came late.

This time the comparison happens at the door. The AI reads every document as it lands, runs your KYB rules across all eight, and names the two points that disagree, with their pages. Emma confirms the gaps, the CFO fixes both at once, and compliance decides on a complete file.

Onboarding ONB-26-1147 · OwnershipCase management
Who really owns Novalis. The holding splits in two, and only one half reached the declaration. On the right, what the file still owes.
Business onboarding queueCase management
Emma's morning queue. Novalis sits between two clean files, sorted by AI verdict like any other field.

None of these screens would surprise a KYC team. The timing would: the documents are read against each other the day they arrive, both gaps go out in one message, and compliance sees only what needs a judgment. The nine steps below follow Novalis from the portal upload on November 3 to the accounts set up on November 9.

02

The file comes in

The company sends its papers, and the AI has read them before the analyst opens the case.

Step 1Intake

The case opens with one slot per document

Marcus Reed, the relationship manager, has agreed with Novalis Logistics Inc., a Columbus, Ohio, freight company, to open a checking account, a savings account and twelve corporate cards. He opens a business onboarding case. Its attachments are a list of named slots, one per document your KYB policy asks for.

Tom Becker, the CFO, uploads his files from the business portal the same afternoon. The portal calls the repository API, and each file lands in the case.

Onboarding ONB-26-1147 · DocumentsCase management
A document filed in its slot gets its type, so the checks know what to look for.
The file is in one place from day oneNo email chain with the CFO, no attachment saved to a shared drive. The papers land in the case, each in its slot.
Step 2Capture

Each document is typed and read

The capture chain classifies each file and extracts the values your team defined for it: company name, registration number, capital, registered office, directors, shareholders, the date of the document.

Certificate of good standing, issued Oct 21Articles, version 2024Holding articlesBoard resolution, Oct 30
Onboarding ONB-26-1147Capture
Each value carries the page it was read from. Emma approves or corrects it on the review screen.
Nobody keys in the certificate of good standingThe fields are read once, checked by a person on the review screen, and written on the case.
Step 3AI check

The AI runs your KYB rules across the documents

Before Emma Collins, the KYC analyst, opens the case, an AI agent runs it against the rules your compliance team wrote in plain language.

Is the certificate of good standing under three months?Do the declared owners match the share registers?Is the resolution signed as the articles require?

The result is written onto the case in two fields: a status and the reasoning. The status is Attention required, with two points, each naming the two documents that disagree.

AI analysis resultsAI console
Each warning cites the pages behind it.
The gaps are found on day oneThe analyst does not compare eight documents line by line to discover them. She starts from two named points and their pages.
03

Who owns, who signs

Both warnings are about ownership and authority. Emma works them with the documents open beside her.

Step 4Viewer

Check the missing owner in the holding's articles

Emma opens the holding's articles in the viewer, at the page the warning cites. Hartley Holdings holds 60 percent of Novalis, and its shares are split evenly between Marc Hartley and his sister Julie. The declaration names Marc and leaves Julie out.

Emma works out the indirect stake herself: half of 60 percent is 30 percent, above the 25 percent beneficial ownership threshold in your CDD policy. She drops a note on the page so the next reader sees her reasoning.

Articles_Hartley.pdfViewer
The note stays on the page for everyone who opens the document afterward.
The analyst checks the chain on the pageThe AI points to the shareholder the declaration forgot. The arithmetic and the conclusion are Emma's, written where the next reader will find them.
Step 5Assistant

Confirm who may sign for the company

The second warning is about authority. Emma asks the assistant who must sign a decision to open bank accounts. The answer points to article 17 and to the resolution, which carries one signature.

Articles_Novalis.pdfViewer
Each answer cites its page. Emma reads the clause before acting on it.
Count them one quarter: how many business files go back to the customer twice, because the second problem surfaced after the first request had left?
The doubt is settled in the articlesNo call to the legal team, no download. The clause opens beside the case, and the answer comes with its page.
Step 6Assistant

Both fixes go to the customer in one message

Emma asks the assistant for a request to Tom Becker. The draft covers both points, with the reason for each and the company name already in it. She adds a line on timing and sends it. The case moves to Waiting for customer, with an SLA timer visible on the step.

Onboarding ONB-26-1147Case management
The request is filed in the case history, with its date and its author.
The customer is asked onceBoth points leave the bank on day two, each with its reason. The CFO fixes them together and nobody waits for a second round.
04

Compliance decides

The file is complete now, and the last word goes to Keisha Brooks.

Step 7AI check

The new documents arrive and the check runs again

Two days later the updated declaration, Julie Hartley's ID and the countersigned resolution come in through the portal. They are filed in their slots, the analysis runs again, and the status moves to Complete.

AI analysis resultsAI console
The earlier warnings stay in the history: the trail shows how each was cleared.
The file comes back readyEmma does not reread the whole stack. The case returns to her queue with the verdict that changed and why.
Step 8Workflow

Compliance reads the file and sets the rating

The workflow routes the case to Keisha Brooks, the compliance officer, because the company has a holding layer. She asks for a summary, opens the two cleared points, and sets the risk rating herself.

One holding layerCross-border freight flowsAll documents current
Compliance reviewCase management
The summary helps the reading. The rating and the approval are Keisha's fields.
Compliance sees only what needs judgmentThe officer reads a complete file with its cleared points marked. Her time goes to the rating and the yes.
Step 9History

The case closes with its full trail

Approval opens a task for the back office to set up the accounts, the mandates and the cards. The case closes, and the next review date your policy sets for a Medium rating sits on the customer file as a field.

Onboarding ONB-26-1147 · HistoryCase management
The trail is part of the case, readable by whoever opens it at the next review.
The audit reads the file as it isWho found each gap, who fixed it, who rated and approved. It is all dated and attributed in one place.
05

The outcomes

Novalis landed in the middle box. Every company file lands in one of three, set by the KYB rules your compliance team wrote, and each box sends the case down its own path.

Complete

Every KYB rule passes

Documents current, owners declared, signatures as the articles require. The analysis lists each rule it ran and the page it read.

Next: the analyst checks, compliance rates and approves.
Attention required

Points need the customer

An owner missing, a signature short. The analysis names each point and the pages that disagree.

Next: one request to the customer; the check runs again when the documents arrive.
Not eligible

A policy rule fails, and you know which

An activity or a legal form your policy does not accept. The AI lists what passes, then names the rule.

Next: the compliance officer reviews and decides, with the reason on the case.
What stays with people

It approves nobody. It refuses nobody. It sets no risk rating and opens no account. Across the nine steps on this page, the binding decisions are taken by the analyst and the compliance officer, with their own rights and their own trace.

The AI reads, compares, drafts and summarizes. The control is in the rules your compliance team wrote, and in the workflow that routes each file to the right person.

06

The gains

Three counts from the Novalis file, November 3 to 9. A company with no holding layer would show fewer gaps, one with three layers would show more.

11 ruleschecked across eight documents before the analyst opens the case.
1 messageto the customer for both gaps, sent on day two, instead of two requests a week apart.
0 fieldsretyped from the certificate of good standing and the articles. Read once, approved on the review screen.

Where these numbers come from. They are counted on the scenario described on this page, with example data. They are not a customer result.

The three principles behind those numbers

Zero wasted comparisonsProductivityThe documents are read against each other on arrival. The analyst starts from named gaps and their pages.
Zero second requestsCustomer experienceEvery gap is found before the customer is contacted, so it is asked for once.
Decision supportA person decidesSummaries and warnings cite their pages. The rating and the approval stay with compliance.
07

Who it is for

An analyst, a compliance officer and an architect each get a vote on whether this goes live. Each has a different reason to cast it.

Business operations

KYC analysts and relationship managers

The week goes into comparing papers, and the customer hears from you twice.

  • The file arrives read, with the gaps already named.
  • The request to the customer covers every point at once.
  • The articles open beside the case, and answers cite their page.
  • Your notes stay on the documents for the next reader.
Risk and compliance

Compliance officers

You sign files you have no time to open page by page.

  • The AI rates and approves nothing. You do.
  • Each warning is a sentence naming the rule and the page.
  • Every analysis stays in the case history, dated.
  • The KYB rules are yours, written in plain language.
IT and architecture

CIOs and architects

Onboarding already crosses a portal, a document store and three back offices.

  • The case is created through the API from your business portal.
  • The AI check is one more step in the process you already run.
  • The capture chain reads the documents where they land.
  • It runs in your data center or your cloud tenant, wherever your KYC data is allowed to live.

How to begin

Pick the comparison your analysts repeat on every company file, such as declared owners against the share registers, and move it to the door. One rule on one queue gives your team real files to judge it by.

Your business portal, your case files and your documents stay where they are. The checks run on top of them.

Let's talk

See it on your own company files

Bring a few anonymized business files, one with a holding in its ownership, and your KYB checklist. We will walk them through these nine steps with you.

The screens on this page are drawn to show one step at a time, with example data.

Uxopian Software helps organizations put AI to work across the systems they already run, without giving up control of their data, their infrastructure or their critical processes.