Digital Financial Reporting

A lot of people are talking about "continuous accounting" and "continuous reporting" and even "continuous auditing".  Some provide grand visions and buzz words to try and describe continuous accounting and reporting.  Some people even give you very broad brush strokes that help you understand what it means. Some provide high-level blueprints for implementing continuous accounting.

But think for a moment.  If you actually wanted to achieve continuous accounting, reporting, and auditing; exactly how would you do that? How exactly are you going to automate the tasks and processes involved?

First, what gets in the way of continuous accounting and reporting?  Here are some specific things that I see:

• Inappropriately set up chart of accounts.
• Inappropriate mapping between chart of accounts and financial report line items. (i.e. lead schedules)
• Information (metadata) used to correctly categorize report information missing from the accounting system therefore automation is impossible.
• Information that is unavailable to automated processes. (i.e. policies not in accounting system, qualitative disclosures not in accounting system)
• Errors.
• Complexity. (i.e. multi-currency, multi-gaap, multi-subsidiaries, multi-special ledgers, multiple accounting systems, etc.)

So all those things add up to the hurdles that need to be overcome to actually get continuous accounting and reporting to actually work.

Second, you actually have to physically implement something.  This video walks you through a basic implementation of continuous reporting.  So this fundamentally works.  Here are all the details of two implementations that have 100% of what is necessary for a basic working system where accounting tasks and reporting tasks are automated: Accounting process automation/record to report; Pixel perfect rendering.  (If you want the complete scope, you need all of this which is simply more volume of what is in the first two examples; but the two examples I point to really are complete.)

To make all of this work effectively, you must be able to control the system effectively.  And so, if you can overcome all of those hurdles and control the system effectively such that the system is reliable and predictable then you can use that system to perform useful work. Let machines do the grueling, gruesome, repetitive, mindless tasks.  Maybe not all tasks and processes will be automated, but certainly some will be automated.

All of this is about augmented intelligence.  Another term used is intelligence amplification.  In his book, Principles, Ray Dalio advises,

“By developing a partnership with your computer alter ego in which you teach each other and each do what you do best, you will be much more powerful then if you went about your decision making alone.”

Augmented intelligence is about computers and humans working together rather than machines replacing humans.  Augmented intelligence applications combine human and machine intelligence. This is important in systems there is low tolerance for error or where artificial intelligence is not evolved enough to take humans completely out of the loop.

That is essentially how continuous accounting, reporting, and auditing will be achieved.  Everything else is really simply a detail.

The Proof+Common+Rendering example representation is a new addition to my set of examples which can be used to master XBRL-based financial reporting.  What this example does is show you the essence of an XBRL-based digital financial report and then reconcile that essence to a 10-K.

Key information for this example representation includes: 

• Files: Every file is available for your further analysis.
• Documentation: This walks you through the example and helps you see what I am trying to get across.
• Database: You can download the Microsoft Access database that I used to create the example.  The only thing not included is the code that generates the disclosure mechanics and reporting checklist rules.
• Pixel Perfect Rendering (Inline XBRL): This is an Inline XBRL instance document that was generated from the database above.  This version references a web-based taxonomy that supports the XBRL instance.
• Pixel Perfect Rendering (PDF): This is a PDF generated from the information in the XBRL instance using XSL-FO and a FOP to generate the PDF.

There is a wealth of knowledge in this information.  This builds on the accounting process automation or record to report example.  In essence, if you read through all the examples you can see that you can put transactions into a general journal and if that is done correctly you can generate a pixel perfect XBRL-based financial report and/or a human-readable PDF, HTML, Word, or other such representation.

Essentially, this points out that there is a completely new workflow possibility that can be employed for creating financial reports.  Whether one of the outputs is XBRL or not makes no difference.  In fact, this is bigger than just reporting; it impacts accounting, reporting, auditing, and analysis.

We live in a world where the questions remain the same; but the answers have changed. This TED Talk video, Why the majority is always wrong, explains.  While 97% continue to do more or less of the same things over, and over, and over; 3% try new things.

Consider thinking outside the box.  Why settle for normal when you can be exceptional! The digital transformation is about talent, not technology.  Learn to leverage your tools effectively.

This excerpt from Ray Dalio's book Principles is one of the best discussions about artificial intelligence that I have run across: SYSTEMIZED AND COMPUTERIZED DECISION MAKING.

This YouTube.com video, AI and Algorithmic Decision Making, provides an excellent overview that is very approachable to business professionals.

There is one notion that I disagree with that comes across in this article.  The article gives the impression that you literally "program" rules (i.e. put rules in code).  For a number of reasons, that is not a good idea. If logic and rules is expressed in "code"; then only programmers can add new logic/rules, maintain the logic/rules, and business professionals have to educate the coders on the business logic that they are trying to get programmed.

There is a better way.  If the "computer code" and the "business logic/rules" are separated, then programmers can write code and business professionals can work with and maintain the rules/logic.  Using this approach, the rules/logic are represented in declarative form in some machine-readable format.  A rules engine or logic engine then uses those rules/logic.

So, rather than writing your "principles" code, you write them as declarative rules.  Closed systems are necessary, you cannot just add "stuff" haphazardly.  Open ended systems might not work reliably, there is a chance of error. If the "future can be different than the past", you have to be very careful.  Open versus closed systems seems to be described by control theory.

Plenty of rules engines or logic engines already exist.  The missing piece is the computer readable business rules. Rules for creating a financial report look like this. (Here are many more examples.)

And so, while I would not dispute that understanding how to code is helpful.  But, you really don't need to "learn to code".  If you want to understand the details better, I would invite you to read Artificial Intelligence and Knowledge Engineering Basics in a Nutshell.

I updated an example that now does something really interesting.  If you go to this PROOF+COMMON+RENDERING example you will find this Inline XBRL Instance.  That was generated using the Microsoft Access database in this ZIP archive.  Now, keep in mind that I am not really a very good programmer, but it will give you an idea of how to generate renderings.  I don't have an interface for editing the "cells" that drive the rendering, I just edit the database table.

That might not look like much, but you could create a complete XBRL-based 10-K for Microsoft, Apple, Google, Amazon, or anyone else using that Microsoft Access database.  Clearly you would want an XBRL processor and XBRL Formula processor to verify that everything is working correctly (that is what I do).

A Harvard Business Review article by Becky Frankiewicz and Tomas Chamorro-Premuzic, Digital Transformation Is About Talent, Not Technology, points out:

“...the best way to make your organization more data-centric and digital is to selectively invest in those who are most adaptable, curious, and flexible in the first place.”

“Technical competence is temporary, but intellectual curiosity must be permanent.”

“A much bigger competitive advantage is to harness valuable data, having the necessary skills to translate that data into meaningful insights, and above all being able to act on those insights.”

Excellent read.  It seems as thought the risk of not adapting is increasing significantly.

This article points out that digital transformation is a leadership problem.

Digital transformation in developing countries.