Digital Financial Reporting

The terms "accounting process automation" and "robotic process automation" and "machine intelligence" and "artificial intelligence" seem to have very different meanings to different people.

Here is a list of vendors that seem to provide process automation:

• Blackline - Accounting Process Automation
• Blackline - Finance Controls Automation
• Docuphase - Accounting Automation, Robotic Accounting Information
• Kofax - Robotic Process Automation, Finance and Accounting
• Winpro- Robotics and Congitive Automation
• IBM - Robotic Process Automation
• IBM - Automation of Month End Closing
• Blueprism - Robotic Process Automation
• Oracle - XBRL stack
• SAP - XBRL stack (uses Fujitsu)
• IBM - XBRL stack
• Altova - XBRL stack
• Arelle - XBRL stack (open source)
• XBRLQuery - XBRL stack (open source)

This video showsa really good example of what most people seem to be calling "RPA" or robotic process automation.  What RPA seems to be is a "bot" performing specific detailed tasks by "scraping" or "reading" information from a web page and then "copying and pasting" fields from a web site or application to another.  It is like "recording" a macro of a routine task which is then repeated over and over by the robot/macro rather than performed by a human.

So, what RPA seems to be is similar to what a business rules management system does plus some sort of user interface to "scrap" or "read" information from one system and put it into another.  Neither business rules management systems nor RPA seem to have an over-arching conceptual model which can be leveaged.

These seem software vendors seem to provide similar capabilities, but they take a different approach.  Each of these uses the semantic web approach and does provide the ability to create a conceptual model which can be leveraged by business rules:

• TopQuadrant- TopBraid, Data Governance
• Coherent Knowledge - Egro Suite
• Cognitum - Fluent Editor

Basically, RPA is a computer performing a specific set of routine tasks in order, replacing a human that performed these mundane routine tasks.

So, I can see the benefit of RPA.  I have a different vision for automation.  I am taking the expert systems approach.  I would not consider RPA an expert system, it is just stringing tasks together.  And I am not using "deep learning" or "machine learning" commonly associated with artificial intelligence.

These different approaches each have different capabilities.  The pros and cons of each approach are not currently clear to me.  What is clear is that different companies are trying to figue out the best way to make automation work.

##########

Oracle XBRL Stack

This TedX video, The Real Reason to be Afraid of Artificial Intelligence, by Peter Haas helps you understand some things you want to keep in mind when thinking about artificial intelligence.  The problem is not the artificial intelligence software.  The problem is people's ignorance of how these tools actually work.

If you want to understand, I would invite you to read the document Computer Empathy that I put together.  Tools such as artificial intelligence are great, but they deserve to be evaluated with a healthy skepticism.  Be a leader, not a follower.

########################

• Elon Musk on the Risks of AI (July 2017)
• But then, the problem with over-automation (April 2018)

Below is a link to a web page that provides a list of 14 pretty high quality XBRL-based financial reports of public companies. (Or you can click here.)

(Click image for larger view)

Each of the 14 XBRL-based financial reports were evaluated using techniques summarized in the document Blueprint for Creating Zero-Defect XBRL-based Digital Financial Reports.

I am comparing the validation results obtained from different software validation tools, this is a work in progress.

For each of these 14 XBRL-based financial filings of public companies, the following is provided: (the links are all for the FIRST filing)

• Link to the XBRL instance
• Link to the SEC filing page (where you can get the HTML, XBRL, or whatever else was submitted to the SEC)
• Link to the XBRL Cloud Evidence Package (human readable information about the report)
• Link to the XBRL Cloud Viewer
• Link to the XBRL Cloud Edgar Dashboard (which shows inconsistencies for submitted filings)
• Link to the Reporting Checklist/Disclosure Mechanics results (shows validation results)
• Link to the Fundamental Accounting Concept Relations results (shows validation results)
• Link to information about the Reporting Style of the company

Also, links to all the machine-readable rules that were used to validate the reporting checklist, disclosure mechanics, and fundamental accounting concept relations rules are provided on the summary document.  These rules have been implemented in two different software applications, XBRL Cloud which is commercially available software and Pesseract which is a working proof of concept used to test the feasibility of some ideas.

If I can find other publicly available information about these filings, then I will link to that information to this web page.

More information coming soon.

More and more evidence is showing that there is going to be an artificial intelligence revolution over the next 20 or so years.  My observation is that techno-optimists tend to over promise what will actually be delivered and business professionals tend to have a rather pessimistic view of what artificial intelligence can deliver in terms of functionality.  What will likely transpire is something in between the optimistic view and pessimistic view.  But undoubtedly, the changes that are coming will be very significant.

As a result of documenting what I am calling Computer Empathy, I am recognizing something that was somewhat unexpected related to this coming revolution.

My realization is the major role that philosophy, in particular the discipline of logic, will play in the artificial intelligence resolution.  Here is why I say that: Logic is the common language shared by business professionals, information technology professionals, and knowledge engineers. Both computer science and philosophy are concerned with the representation of information and rational inference. Philosophy enhances the kind of thinking required to be clear, analytical, and concise.  This is a graphic that explains what I see:

(Click image for larger view)

Logic is a natural, common language that is intuitive, can be agreed upon, and which therefore enables precise communication between different parties.

Consider the graphic below. Think of any information technology project.  Really any project for that matter where you have people across many domains that need to communicate precisely.  Communications problems cause many projects to fail.  Consider the different parties in the graphic below.  What one thing do these parties have in common? The answer is logic. 

(Click image for larger view)

Comparing the above graphic to the graphic below makes the point.  What tool do different parties, particularity parties in different domains, use to agree on something?  What does every domain have in common? If you enabled this ability to communicate, then this is what you would get:

(Click image for larger view)

Reasonable people tend to be naturally logical.  That is a very good start.  But what if you FORMALIZED this understanding of logic just a little bit, giving people more discipline and rigor, making their use of logic more systematic?

The domain of philosophy, which invented (or discovered) logic, offers that solution.  But, the solution needs to be PRACTICAL, understandable by the typical business person; this logic is NOT used to create endless philosophical debates; the point is to understand how you can use the logic to agree if it is the case that you are trying to agree and you WANT to agree in order to solve a problem.  That is what business people like: solutions to problems; making things better, faster, and cheaper.

And then, what if you took this intuitively understandable language of logic and made it machine-readable? What machine-readable syntax would you use?  Consider the example syntaxes below; which do you find preferable?

(Click image for larger view)

What if a controlled natural language that is understandable to humans could be created to represent information in a machine-readable syntax?  Maybe a controlled natural language is too much to ask for, but what if something understandable to business professionals could be created?

Further, there are multiple technology stacks technical people can use to solve problems.  The decision as to which technology stack to use can be arbitrary or driven by past choices which cannot be reversed.

As I see it the discipline of logic from the domain of philosophy is a common denominator that could enable precise communications and enable the artificial intelligence revolution to solve more business problems than without this common language.

What do you think? Others seem to share this view.

#####################################################

• Philosophy, Computing, and Information Science
• Why Computer Science Needs Philosophy, Barry Smith
• Computer Science and Philosophy at Oxford
• Computer Science and Philosophy at Stanford

Ran across the Future of Accounting Podcasts today.  This seems worth checking out.