Digital Financial Reporting

Brad Smith, CEO Intuit, is warning accountants that they must embrace artificial intelligence (AI) in their professions or risk falling behind.

Blackline refers to itself as a finance controls and automation platform and provides accounting process automation says that accountants should gear up for AI disruption. Blackline says that artificial intelligence is set ot revolutionize accounting and change the role of accountants.

Xero sponsored a survey which says that 80% of accountants understand that they’ll need to undertake training in the next five years to adapt to technology and that AI is the new frontier for accounting technology.

The Journal of Accountancy is talking about rethinking the audit and the new type of auditor.

Steve Guggenheimer's Blog post, Introducing Dynamics 365 AI Solutions, explains how Microsoft is "helping developers build services that amplify human ingenuity."

SAP is employing artificial intelligence, as is ORACLE; basically I could go on and on.

The risk of not getting on board with the changes artificial intelligence and other technologies are enabling is increasing.  Might be ready to get ready for the digital age of accounting, reporting, and auditing.

Here are some other articles:

• Forbes: Robotics: The Next Frontier For Automation In Finance And Risk Management
• Deloitte: Its Digital Crunch Time for Finance
• PWC: Hi, Robot
• The Economist: Why the future of finance is (still) automation
• Kofax Advisor: Is the Modern CFO Headed for Extension?
• Ventana Research: Welcome to the Age of Robotic Finance

I took the metadata and tools that I createdfor checking the fundamental accounting concept relations continuity cross checks of US GAAP based financial reports and I started to update that for IFRS.  If anyone wants to check this out, here you go:

1. Excel extraction prototype updated with some IFRS metadata
2. Several IFRS example financial reports (used to exercise the Excel spreadsheet application)
3. Easy to read version of the IFRS FULL taxonomy
4. Easy to read version of the IFRS for SMEs taxonomy

Ultimately I will update my XBRL-based metadata.  I will probably do that based on 20-F filings to the SEC that contain IFRS based reports that use the IFRS taxonomy.  There really are not very many IFRS reports around in XBRL to test things with yet.

If you like Sudoku puzzles, you will enjoy fiddling around with this.  Someone may even be able to create a useful product.

Today, what portion of the checks or other review steps performed to make sure the quality of a financial report is what it needs to be are performed using automated processes? The answer is ZERO.  Well...I guess you could count using spell check in Word an automated task.

Why is that?  Why can't certain quality control tasks and other review steps be performed using automated machine-based processes.

The reason all review tasks have to be performed using human-based processes is because financial reports of the past have been unstructured documents which are impossible for a machine to read reliably.  Well sure, a word processor can "read" a financial report; but word processors know nothing about financial reports.  Word processors understand documents and they can tell you if you messed up the document structure (i.e. word, paragraph, section, etc.).  If this is unclear to you, consider watching the video How XBRL Works.

XBRL-based financial reports created an opportunity to change the review process used to make sure financial reports are created correctly.  XBRL alone cannot change this.  But, if you put three things together a transformation can occur.  To achieve this change you need three things:

1. Structured financial reports: Individual pieces of information in structured reports can be reliably addressed and worked with by computer-based processes.
2. Machine-readable rules:  The logical, mechanical, structural, mathematical, accounting, and reporting rules of a financial report must be put into a form that a computer can understand and work with.  The machine-readable rules are used to make sure the structured financial reports are created correctly.
3. Software:  Computer software is the machine that reads the rules and otherwise leverages the rules to (a) help professional accountants create the financial reports and (b) help make sure the reports are created correctly.

Because of XBRL-based financial reports, machine-readable rules, and software that can effectively understands the rules; it is no longer the case that ZERO percent of the process of reviewing a financial report to make sure it has been created correctly.

Today, it is possible to automate some percentage of those quality checks and other review tasks.  What percent?  While the actual percentage of what can be reviewed using automated machine-based processes is unclear; that percentage will be significant.  The percentage will be determined by the volume and quality of the machine-readable rules and the problem solving logic available from the software.

When will this software be available?  Well, the software is available today.  Here are two software applications that perform automated review tasks:

• XBRL Cloud Reporting Checklist (commercially available, in BETA)
• Pesseract Reporting Checklist (working proof of concept to test the feasibility of automated reporting checklists)

You can fiddle with the reporting checklist here if you are interested; experience the details for yourself. This ZIP archive contains an Excel based application that lets you extract and see the machine-readable rules used by both of those software applications. There are THREE Excel spreadsheets: one for US GAAP, another for IFRS, and another for what I call the XASB reporting scheme which is an imaginary scheme I created to figure some things out.  This document explains the details. This document explains the techniques used.

Currently, XBRL Cloud and Pesseract are both flawed in my view.  The flaw is that they are both batch processes run AFTER a report is created.  Why is that a flaw?  It is a flaw because the software does not help you actually create the report.

What I am doing is creating software that is used DURING the process of creating a financial report.  Basically, what I refer to as an expert system for creating financial reports.  The idea of expert systems is not new.  Expert systems is one of the most successful areas of artificial intelligence.  When I talk about human and machine collaboration or machines augmenting humans; what I am talking about is an expert system for creating financial reports.

Ten years ago I could not explain the difference between "syntax" and "semantics".  Literally. When I created that video, How XBRL Works, on May 1, 2008; I was still struggling with understanding the difference between those two terms.

A lot has changed since then.  I have accumulated a lot of information and step-by-step figured out how to create the expert system I believed could be created.  I figured it out with the help of a bunch of other people.  There are still a few details but the problem of how to do this has been solved.  That is the purpose of Pesseract, to figure this out.

Financial reporting...get ready to be disrupted.

I have always thought Datomic was interesting.  Datomic Cloudis even more interesting.  Datomic Cloud is even more interesting on Amazon Web Services.

See this Introduction to Datomic video.

Here is a list of reasons why I find Datomic so interesting:

1. Datomic is a fact database, organized similar to EAV (entity, attribute, value) or RDF.
2. Datomic has an integrated rules engine, datalog.  See this video.
3. You can never delete facts from the database.  You can mark them deleted, but you cannot make facts go away.  That means excellent audit trail.

++++++++++++++++++++++++++++++++

SirixDB

LearnDatalogToday

JUXT XTDB

I built a working prototype XBRL-based financial report validation tool in Excel probably four or five years ago.  That tool was used to test the feasibility of some ideas which I had.  Those ideas have turned into commercially available software products/solutions.  One software vendor, for example, offers a suite of validation products.  Another commercially available product is being made available. And I have been working with another software engineer to test some other ideas.

These tools might look like toys.  But if you understand disruptive innovation and you look closely; you will realize that these are far from toys.  This is experimentation that will lead to new products and services.

I have updated my older prototypes and synchronized them very closely to the commercially available validation products for my fundamental accounting concept relations continuity cross checks.  You can get Excel spreadsheet applications that currently cover almost 70% of all public companies that submit XBRL-based financial reports to the Securities and Exchange commission.

Here is the VBA Code if you want to check the code before you download or run the macros.  All the spreadsheets have basically the same code, just different mapping rules, impute rules, and consistency check rules.

These two videos help you understand the tool:

1. Video 1 - Overview
2. Video 2 - Additional Information

All the spreadsheets basically work the same, they just have different metadata for the different reporting styles of public companies.  You can use these spreadsheets to analyze across entities like in most of the examples that I have provided.  But, you can also analyze all the filings of one company across all their filing periods.

Here are some examples:

• NO ERRORS EVER: Apple, Google, Urban Outfitters, and Red Hat have never made an error related to the fundamental accounting concepts continuity cross checks.
• ERRORS FIXED: AT&T is an example of a public company having errors; but the errors were subsequently fixed.
• ERRORS:  Fossil, Nexus Enterprise Solutions, Simpson Manufacturing, Tiger Reef, and United Natural Foods are examples of existing errors that still need to be fixed. (Compare those filings with these 1,642 public companies that use the same reporting style but don't have any errors.)

To make the Excel based tools work, you have to get the URL of the file you want to validate or you can put the full path to any local file you want to validate.  You can get file locations from the SEC EDGAR system, from XBRL Cloud's EDGAR Dashboard, using an API such as XBRL Cloud's (human readable, machine readable); lots of different ways to get the files.

By the way, this works with local files also.  See this example. (Clearly this will not run; you need to put the path in to your files.)

You have to understand the reporting style of the company.  Here is a list of reporting styles for all public companies (not all styles are supported yet). Notice the URL; select from the list or just put in the beginning of the name of the company and you will get a list.  XBRL Cloud has an API for getting this information as well.

From the list of reporting styles you can get to a summary of the reporting style; here is the main entry point of the metadata.

Again, the metadata is near commercial quality.  I am not sure how close I will make this, I don't want to take away business from people who have existing commercial products.

Companies that are going to file with ESMA are going to need similar functionality.

Ignorance is defined as a "lack of knowledge or information".  All this information is an excellent opportunity to vanquish ignorance related to how XBRL-based digital financial reporting actually works.  What I discovered from all this poking and prodding of XBRL-based financial reports submitted to the SEC is that validation metadata is the same metadata that you need to extract information from those reports effectively.

But don't believe me.  Fiddle with the Excel spreadsheets.  They are great learning tools.  The code is terrible and you can clearly undrestand the problems of embedding business rules in software code.  Bad idea.