AI Annual Report Intelligence

Credit reference agencies gives lenders and and others a range of information about the pontential borrower, customer or vendor. A signifcant part of this information is based on annual reports and other financial statements belonging to the business.

The same set of financial data is also exposed in business information services provided by a large number of actors in the market.

In countries such as Sweden the statutory audit requirements for companies was abolished in 2010 which since then allows businesses under certain size to reduce administrative burden by not having an auditor. This has created new challenges which has remained unsolved by our competitors. Companies can submit financial statements and annual reports that contains forged financial data and authorities have very little to no possibility to detect these.

The Swedish National Audit Office, NAO, wrote in the report 12 december 2017 - Abolition of audit obligation for small limited companies:

One consequence is that the registered annual reports risk containing incorrect totals, which contravenes the requirement that annual reports must give a true and fair view of the companies’ finances. The deficient annual reports are passed on unchanged to various stakeholders, such as other authorities in criminal investigations and credit rating companies, which risk making assessments on incorrect grounds. For example, a company can record incorrect financial information in the balance or income statement to present the company’s finances as better or worse.

​

Solution

Togheter with AI intelligent document processing, IDP, we detect a large number of discrepancies in annual reports such as these examples:

• Similarities in annual reports based on historic annual reports for all businesses, e.g has the annual report been copied from another business or part of the data been modified
• Summary errors
• Unrestricted equity doesn’t balance between the years
• The annual report doesn’t adhere to requirement per local regulations for example in Sweden Annual Accounts Act according to K2 (BFNAR 2016:10) / K3 (BFNAR 2012:1)
• Visually aligning with known accounting systems that typically produce annual reports

These discrepancies are registered on each document with links to which similarities exists with other documents. The discrepancies are summed up and expressed in words what is wrong with a score discrepancy ranking between 0-255 where 0 means just a note while 255 means something very a significant discrepancy.

You will find examples in our API when we return for example:

{
  "intelligence": [
    {
      "companyIntelligenceType": "annualReportDiscrepancies",
      "notes": "The annual report for 2021-01-01 to 2021-12-31 is
      similar to one or many more annual reports that
      has been submitted for other companies",
      "score": 255,
      "firstSeenAt": 1711843200
    }
  ]
}

​

Example

A business with an auditor submits an annual report for the year ending 2020-12-31. The following year another business without an auditor submits a copy of the same annual report but the multi-year overview is wrong and one of the digits in the organizational number has been removed.

Visually compairing the result sheet for both businesses you can notice that the rows and amounts are the same with the difference for the year.

Visually compairing the balance sheet for both businesses you can notice that the rows and amounts are the same with the difference for the year.

The business that made a copy of the annual report report has also copied the overview with the other business change of equity. The following image shows that the outgoing balance should be 372,707 SEK in unrestricted equity.

The following year the summary with change of equity doesn’t match with the prior year. For the year ending 2022-12-31 the business reports that the ingoing unstricted equity balance is 1,245,154 SEK while the prior year had an outgoing balance of 372,707 SEK. There are no remarks regarding the changes of the unrestricted equity.

In the balance sheet of the business it’s reporting a change in credit facility debt of 1,467,157 SEK.

​

How common are financial statement frauds

We have learned by looking at discrepancies for the last ten years that financial statement frauds are more common in businesses that has no auditor but they do exist in both. It’s easy to create a new business, submit a fradulent financial report with good looking numbers and get a descent credit score ranking by existing credit reference agencies. The companies may then be used to receive contributions and borrow money and then eventually they go bankrupt.

Often we see networks of companies using the same set of financial records but with no links between the company representatives. They exist in various industries and cities.

This anonymized example shows a small portion of a network of such companies where eight businesses uses two set of different financial reports but there are no links between the yellow company representatives. These companies could ultimatley be controlled by other individuals.