Digital Financial Reporting

The future of accounting, reporting, auditing, and analysis will depend heavily on curated machine-readable knowledge that is also readable by humans. Think digital ecosystem.

While the US GAAP XBRL Taxonomy and IFRS XBRL Taxonomy are necessary for the future; what they provide is not sufficient for this inevitable future.  As such and as I pointed out, these two critically important XBRL taxonomies will be supplemented with additional metadata and will become high-quality, highly expressive ontologies, on the right side of the ontology spectrum.

As I have been able to show, it is very possible to represent the US GAAP Financial Reporting Ontology and the IFRS Financial Reporting Ontology using the XBRL technical syntax. I used this method to create highly expressive ontologies for four different reporting schemes.

As I have tried to show with this cross reference, with this prototype, and with this prototype; information can be both machine-readable and generated into human-readable information that the average professional accountant can relate to.

So, who will create and maintain these highly expressive machine-readable ontologies?  Just as important as knowing who it will be is understanding who it will not be.

It will not be the FASB or the IFRS Foundation.

The FASB and the IFRS Foundation has for about 15 years created little more than what I would characterize as human-readable "picklists".  The IFRS Foundation has religiously aligned the XBRL taxonomy that they create with "the bound volume".  "This appears to be the IFRS Foundation's comfort zone," as one professional colleague put it.  The IFRS Foundation could change their mindset, but history has shown that is probably doubtful.  The FASB provides more common practice information, but the US GAAP XBRL Taxonomy is still a low-function picklist.  The FASB seems comfortable with this.

If you think about this and ask yourself if the FASB and IFRS Foundation can provide the following you will likely reach the conclusion similar to what I have reached:

1. Do the US GAAP XBRL Taxonomy and the IFRS XBRL Taxonomy provide what is necessary to implement XBRL-based internal reporting within the enterprise? Nope, not even close.
2. Do the US GAAP XBRL Taxonomy and the IFRS XBRL Taxonomy provide what is necessary to leverage those XBRL taxonomies for AI assisted audits?  Nope, not even close.
3. Do the US GAAP XBRL Taxonomy and the IFRS XBRL Taxonomy provide what is necessary to leverage AI in other ways for automation of accounting, reporting, auditing, and analysis?
4. Do the US GAAP XBRL Taxonomy and the IFRS XBRL Taxonomy enable the creation of high-quality XBRL based financial reports.  Well, the quality problems that persist in XBRL-based reports submitted to the SEC yield the answer of a resounding no.
5. If an audit requirement for XBRL-based reports were mandated today; would the US GAAP XBRL Taxonomy and IFRS XBRL Taxonomy be up to the task?  Definitely not.  There is zero probability that an auditor could withstand the scrutiny enabled by automated processes that exist today.
6. Do the US GAAP XBRL Taxonomy and IFRS XBRL Taxonomy provide what is necessary to enable Deloitte's vision of "the finance factory".  Nope. What about the expanded vision that I see?  Again, nope.
7. Are the US GAAP XBRL Taxonomy and IFRS XBRL Taxonomy up for the vision of "Audit 4.0" as outlined by Rutgers University? Not even close.
8. Are the US GAAP XBRL Taxonomy and the IFRS XBRL Taxonomy sufficient to enable the replacement of paper-based reporting with XBRL-based reports, could quality be maintained?  Very doubtful.
9. Are the US GAAP XBRL Taxonomy and IFRS XBRL Taxonomy up for Blackline's vision of the modern finance platform or DFIN's vision or PWC's vision? Unfortunately not.
10. Can you use only the US GAAP XBRL Taxonomy or IFRS XBRL Taxonomy to create an expert system for constructing a financial report.  Nope...and that is why I created what I created.

All this begs the question as to whether XBRL itself is even up for the tasks above.  Based on my testing and based on the Method of Implementing a Standard Financial Report using the XBRL Syntax which I have developed; I can prove that XBRL is up to the task if it is implemented correctly.  If you look at the evidence, you can see that six filing agents are capable of achieving high-quality.

Curated machine-readable metadata that are high in the ontology spectrum will improve quality even more.  In fact, curated machine-readable metadata will supercharge what has already been achieved. Folks, all this is based on science. If people stop making four common mistakes and if they took the time to understand what XBRL actually is; then people will figure this XBRL-based digital financial reporting out.

But again, I still have not answered the question: curated by whom? Will the Big 4 finally figure out that there is, in fact, a business model here and each create their own curated ontology?  Will the AICPA?  Maybe CCH/Wolters Kluwer?  A second tier CPA firm?  Perhas a university.  Maybe an institute.  Will I successfully complete the versions of the metadata that I already have for US GAAP and IFRS?  Maybe a software vendor like Intuit maybe?  Will there be a multitude of curated versions of the US GAAP and IFRS Financial Reporting Ontologies?  Will it be an individual effort, a collaborative effort, or a mixture of both?

Will the US GAAP Financial Reporting Ontology and the IFRS Financial Reporting Ontology be free or will you have to pay for it's use?  Probably a mixture.

I am quite confident that one or more curated versions of these important financial reporting ontologies will  be created.  As I pointed out before, the existing US GAAP and IFRS XBRL taxonomies are like crude oil. Some will refine them into gasoline, but others still will refine them more and create high-octane racing fuel.

If I can figure this out, others can also.  The answer will be revealed over the coming years.  Heck, the answer might not even be XBRL it could be RDF + OWL + SHACL or some other rules mechanism.  Or, it could be both XBRL and RDF/OWL/SHACL, some at OMG prefer that syntax.  What technology stack will prevail?

Regardless of who or what syntax, the future of accounting, reporting, auditing, and analysis will be machine-readable curated knowledge.

Anyone think otherwise?

As I point out in the document Demystifying the Role of Ontologies in XBRL-based Digital Financial Reporting:

In simple terms, ontology is about naming parts and processes and grouping those parts and processes together into categories.  An ontology is a description of what exists within some field or domain; the parts and the relationship and hierarchy of the parts relative to one another.  Why is ontology important?  Ontologies help you think about a field or domain.  Ontologies help you have precise discussions about challenging questions, to build theories, to construct models, to help you better understand the field or domain represented by the ontology.

This is a prototype of both human-readable and machine-readable information about a US GAAP classified balance sheet:

• US GAAP Classified Balance Sheet

The information set is not complete.  For example, I don't have formal class relations, I don't have formal roll up relations, I don't have formally defined properties that can be used to separate the "current" from the "noncurrent" items.  There is a lot that I don't yet have.  But, I have a starting point.  I am going to figure out how to add all that other information.

It would be very hard for me to believe that this information is not helpful to accounting students, professional accountants creating private company financial reports, etc.

More to come...

The concepts, documentation, labels, and references in the US GAAP and IFRS XBRL Taxonomies are pretty decent.  There are some needed improvements, but the dictionary of concepts and associated documentation and references is very good.

But the relations in both US GAAP and IFRS XBRL taxonomies leave a lot to be desired.

Both the US GAAP and IFRS XBRL Taxonomies tend to be monolithic representations of "stuff" that is not particularly well organized, relationships are informally described (as contrast to being formally described), very important information is left out, both tend to be more internally inconsistent with how they represent things than they should be; and therefore the taxonomies are harder to use than they really need to be.

So here is a list of a significant set of the networks of information provided by the US GAAP XBRL Taxonomy.

Take the first network, the "Statement of Financial Position, Classified" or commonly referred to as the balance sheet.  Here are four methods for looking at that network of information:

• Flat list, HTML page: This is a flat list of the information output as an HTML page.
• Tree view representation: This is a representation of the information in the form of a tree or hierarchy that I generated.
• CoreFiling Yeti tree view: This is an off-the-shelf professional quality commercial tool for working with the taxonomy information.
• Workiva tree view: This is another off-the-shelf professional quality commercially available tool for working with the taxonomy information.

All the tools basically provide similar information.  Some tools have more information, some tools have less.  Some tools are good for some things and bad for other things.

But all tools fundamentally provide the same organization of that network from the US GAAP XBRL Taxonomy.  Take a look at that network, study it, and consider the following:

Is there a better way to orgainze this important taxonomy information so that it is easier for both humans and for machines to work with?

Consider the following.  What if rather than representing that information as one monolithic network, the information was represented as several smaller sets of information.  For example:

• Assets [Roll Up] and Liabilities and [Equity Roll Up]: This is two pure roll ups without other stuff intermingled within the roll up.  Clearly the XBRL calculation relations would also be provided.
• Preferred stock parenthetical information: This is the parenthetical information that is not part of the roll up; it provides additional details about each class of preferred stock.
• Common stock parenthetical information: This is the parenthetical information that is not part of the roll up; it provides additional details about each class of common stock.
• Treasury stock parenthetical information: This is the parenthetical information for treasury stock.
• Partnership alternative for equity section: This is the partnership alternative concepts that would be used instead of the stockholders equity concepts in the roll up should the reporting economic entity be a partnership.
• Limited Liability Corporation alternative for equity section and parenthetical: Similarly, this is the LLC alternative that would be used for the equity section and parenthetical disclosures should the reporting entity be an LLC.
• Partnership alternative for parenthetical information: This is the partnership alternative for the parenthetical information that a partnership would use to report.
• Dimensions information for a restatement: This is the dimensions information should you need to report a restatement.
• Dimensions for class of stock information: This is the dimensions information should you need to report class of stock information.  This has a problem though; you should never report preferred stock and common stock together.  It really should look like this: (i.e. they should be separate)
  • Preferred stock classes
  • Common stock classes

Notice how much easier that the information is to understand as contrast to having to work with the single monolithic network that is not particularly well organzed.  Also, consider what software developers need to do.  For example, suppose that your entity is a corporation.  Would you EVER use the partnership or LLC concepts?  Of course not.  Wouldn't it be nice if non-relevant concepts were simply never shown to software users?  That reduces the number of concepts that you have to sort through.

So go back and take another look at this LLC related set of concepts and the same information in this network provided with the US GAAP taxonomy.  Really look close and try and understand the information.  Notice a difference between the two?

Fiddle around with some of the other networks.

There is one additional advanced topic that I will mention.  While both the US GAAP and IFRS XBRL taxonomies define simple terms neither defines functional terms that are made up of sets of simple terms.  For example, there is no real notion of a "disclosure" in the US GAAP or IFRS XBRL taxonomy.  You have sets of concepts for form disclosures, but there is no real way to directly refer to a specific disclosure.  The closest thing to a disclosure is the Level 3 Disclosure Text Blocks.  But those are simple terms, each defines one disclosure.  Can you match that Level 3 Disclosure Text Block to the associated Level 4 Disclosure Detail?  No, you cannot.  There is no way (i.e. a functional term) that is provided to explicitly and reliably identify each Level 4 Disclosure Detail.  But what if there were?  I provide a means to do this. (See the column Disclosure Name here.)  How useful is that?  Here are examples of specific functional terms:

• Roll Forward: Property, Plant, and Equipment [Roll Forward]
• Roll Up: Assets [Roll Up]; Liabilities and Equity [Roll Up]
• Set: Preferred Stock by Class [Set]

Mindbridge.ai has a product called AI Auditor.  You can watch a five minute video that describes their product here.  This article describes how AI Auditor works.  This short video provides a concise summary of their approach.

How much more do you think an AI auditor can do if the financial report itself is machine-readable structured information as contrast to unstructured information?  Think about that.

Here is everything you need to know to have an intelligent conversation about XBRL. As more and more metadata becomes available, then AI auditors will be able to contribute more and more.

Here is US GAAP metadata that makes AI auditor more effective.  I have the same sort of metadata for IFRS but not quite as built out.

Here is a framework for creating even more metadata.  I have carefully tested US GAAP, IFRS, IPSAS, and a testing reporting scheme that I created called XASB to make sure the framework will work for any reporting scheme.

We are one step closer to the Deloitte vision of the finance factory!

What are others saying?  EY explains how AI will transform audit. (This is a very good video, worth watching.)   This article by the AICPA explains how AI can improve the audit process. This article by FEI explains the impact AI will have on audit.  Other tools already exist. Here is OnPoint PCRthat the AICPA is behind.

AI is real. While the AI assisted audits probably provide a bit of assitance now, that level of assistance will go up, and up, and up.  The tasks performed by human accountants will move around.

Ontologies will power our future.  The document, Demystifying the Role of Ontologies in XBRL-based Digital Financial Reporting, will help you understand what ontologies are, get a sense of how they work, and see the sorts of things you can accomplish with ontologies.

In simple terms, an ontology is logic represented as machine-readable metadata. The more metadata you have, the more you can make software applications do.  It really is that straight forward.

Here are several examples that give you a sense of how ontologies can be useful for XBRL-based digital financial reporting:

• US GAAP Cross Reference: Imagine a "professional accountant knowledge portal" for each financial reporting scheme.  Imagine machine-readable disclosure checklists (rather than human-readable memory joggers).
• US GAAP Financial Report Ontology (Prototype): Imagine a machine-readable resource that is also human-readable that organizes the knowledge used for accounting, reporting, auditing, and analysis of financial information.
• XBRL-based Financial Report: Walk through this self guided tour of an XBRL-based digital financial report and think of such reports as reconfigurable pivot tables (rather than inflexible documents).
• Logical Model of a Report: Contrast this logical model with how XBRL-based reports were explained to you. This set of details explains each of those pieces. Imagine a software application that can interact with all this machine-readable metadata.
• Expert System for Creating a Financial Report: Watch these videos that are an initial descrption of an expert system for creating financial reports. More is coming soon, so stay tuned and watch Pesseract come to life.

Using a war metaphor, strategy is about picking the right battles and tactics is about execution of each battle effectively.  It might be time to reevaluate your strategy and tactics when it comes to XBRL-based digital financial reporting.

GAAP Codification: An Ontology Perspective