Productivity Boost for Accounting and Audit

Accounting and audit is about to get a significant productivity boost.  How? The productivity boost will come from the capability to create a human-machine team using knowledge based systems (a.k.a. what I call a mindful machine for accountancy) effectively. Hybrid artificial intelligence, grounded in rules-based artificial intelligence and supplemented by probability-based artificial intelligence, and operated by a skilled, experienced accountant/auditor will make "the average accountant" above average. A super accountant/auditor.  Entire workflows will have significant reductions in the "friction" that exists today. Quality will skyrocket and attaining that level of quality will cost less.  That is the definition of productivity.

It is tempting to think that achieving this productivity boost will be easy.  It will not. Attaining this productivity gain will be hard work. Personally, I have been working on this task for 25+ years. The article, AI leads a service as software paradigm shift, helps to explain this shift.  But with all due respect to AI; accountants and auditors are going to lead this shift, not AI.  AI is just a tool. Information technology professionals don't understand accounting and auditing or the workflow involved; accountants and auditors understand accounting and auditing the work they do and the workflows they employ. That said, the article does point out, "Humans still design the workflows. Even in the most rote operations, like invoice processing, they perform spot checks and look in on problem cases." Totally agree with those notions.

Here are the pieces of this puzzle as I see them: (1) work, (2) workflow, (3) machine-interpretable metadata, (4) algorithms, (5) agents, (6) standards, (7) machines, and (8) skilled and experienced accountants and auditors. AI will be used by skilled and experienced accountants and auditors; not nobs who think they can pick up a machine and do accounting/auditing.

Allow me to explain.

As described in the article, The Future of Audit: What Will Audit Firms Look Like in 2030?, 

"From the outside, [CPA] firms look modern and digital. From the inside, they feel like an old structure resting on a heavy foundation, entangled in dozens of disconnected tools and tasks that demand constant juggling at every level of the pyramid."

The typical CPA firm which is made up of entry level staff accountants or "associates", seniors who supervise the staff,  managers who supervise the seniors and run audit engagements, and partners at the top of  the pyramid use, on average, 17 different tools to perform their work. Add to this a "reporting team" from, say, a Fortune 1000, who Gartner says uses an average of 800 electronic spreadsheets to document and generate a financial statement that will ultimately be provided to a regulator.

Today, creating such of a "closing book" by a reporting team or "audit bundle" by an audit team is tedious, repetitive work. The drudgery of copying and pasting of information, rekeying information, entering information manually creates what can best be described as a "bucket brigade". 

Most processes/workflows can best be described as kludges that have evolved over a long period of time that get the job done; but things are held together with what amounts to band-aids, bailing wire, and duct tape". Electronic spreadsheets tend to mask errors than help accountants avoid errors and maintain high quality.

These systems are not as capable and effective as they might be. Current accounting systems tend to focus on generating a balance sheet and an income statement, the debits and the credits.  By starting with the debit/credit problem first, entering financial transactions; when the  IT world computerized accounting they basically started at the "end of the chain" rather than at the beginning of the chain where they really should have started, at the business event. To overcome this issue, electronic spreadsheets are created to introduce necessary information into the process. But that information is introduced later than it really should be introduced.

This "hairball" might seem normal, but it is not normal.  Much of the work of many accountants tends to be that of effectively a "data janitors".  Accountants perform tasks such as "transaction chasing". When things cannot be figured, then accountants resort to techniques such as "the plug". There is mounting evidence that the quality of financial statements and audit working paper evidence is falling. Today's processes are nearing their end of life.

Now, don't think of a public company's reporting team generating a "closing book" and ultimately a financial statement that will be submitted to a regulator and an independent third-party audit team generating an "audit bundle" that uses information from that closing book, information from that financial statement, and additional audit evidence collected as separate processes.  They are not separate processes; they are connected.

The creation of something like a "closing book" or "closing binder" that supports a financial statement or an "audit bundle" which supports an opinion related to an assessment of a financial statement takes deep technical knowledge of a team of talented, skilled, experienced accounting/audit professionals.  It is very rare for one person to understand 100% of every step of every task or process related to the work performed to create either a financial statement or an audit bundle. Therefore the work itself tends to be collaborative and the project management of that work also tends to be collaborative.

And when you try and combine the work of the "reporting team" and the work of the "audit team"; workflow and collaboration gets a whole lot more complicated.  Access control to information, information flow, project management, coordination, moving "work" from a reporting team to an audit team, controlling and coordinating last minute changes.

The article, The Future of Audit: What Will Audit Firms Look Like in 2030?, explains this process that results in the creation of an audit bundle at a very high level. The book, The Future of Accounting, explains this process that relates to the creation of a financial report at a very high level and goes into more details related to the accounting related to business events that drive financial transactions that end up as a very significant part of that financial statement.

One tool currently used by teams creating a "closing book" and the resulting financial statement and/or the "audit bundle" supporting the audit work performed with respect to a financial statement is the calculator.  The calculator is an invaluable tool used by accountants to help them get the numbers right; making sure everything "ticks and ties" and "cross casts and foots". And accountants trust that this tool, the calculator.  Accountants and auditors know that the tool works and they understand how to use the tool effectively.

Another tool currently used to create a "closing book" and "audit bundle" is the electronic spreadsheet. How trustworthy are those spreadsheets?  When spreadsheets fail, do they "fail loudly" and let the accountant or auditor that relies on that spreadsheet know?  Not typically. In fact, most spreadsheets mask mistakes. And truth-be-known; not many accountants are very good at creating good electronic spreadsheets. The average accountant is average.

Process knowledge and procedural knowledge represents the steps, parts, pieces, modules; the algorithms; that serve as a "composite set" of all the "information blocks" or "organisms" and the understanding, the "know-how" of the construction of the "closing book" or "audit bundle". That process knowledge and procedural knowledge is also all the "rules" (a.k.a. assertions, constraints, restrictions) that apply to how all the information in those information blocks is assembled and are used to test that assembly of information block organisms to understand the epistemic risk that exists that you got something wrong.  Knowledge is more than just the "information blocks" that make up the "closing book" or "audit bundle"; knowledge is also the "know how" related to how to put those "information blocks" together.

Epistemology is the blueprint, the engineering, and the inspection process which is used to answer questions like: Is this "closing book" or "audit bundle" complete? Was the financial reporting framework followed when the "closing book" or "audit bundle" was constructed? Can we trust the "closing book" or "audit bundle"?  Should we take the risk and submit the financial statement to the regulator?  Epistemology is less about what you know, and more about how you know what you know, and whether you should trust it.  Epistemology is about assessing your control mechanisms.

Accountants and auditors are experts in assessing this risk.  The problem is that (a) there is far to much human involvement of very expensive humans so costs tend to be higher than desirable, (b) humans are, well, human and they make mistakes, and (c) the only tool accountants have are the electronic spreadsheet, the "Swiss army knife" of the accountant, or work very similar to the electronic spreadsheet because they are document or document-like and non standard.

But what if there were a better way? I have mentioned the notion of the "information block".  Allow me to elaborate.

Imagine the notion of an "information block". Imagine being able to "snap" together information blocks into a process.  Imagine information blocks as being like a Lego.  Imagine that human powered "bucket brigade" assisted by Lego-like "information blocks" that "snap" together using logical glue.

These Lego-like information blocks are really "organisms" of information as defined by Atomic Design Methodology.  When these Lego-like information block organisms are assembled to create a "closing book" (a.k.a. closing binder) or "audit bundle" (a.k.a. internal or independent audit working papers), that set of information blocks becomes knowledge.  

It is the process of creating that "closing book" or that "audit bundle"; the choice of which information block "organisms" to include and how those "organisms" are assembled; it is in that process and in the procedural knowledge that we understand how, the "know-how", the final creation is made. It is in that that we trust the information blocks and trust the process and procedures used to create those information block Lego-like organisms.

Procedural knowledge, this "know how" or "practical knowledge" or "imperative knowledge" or "performance knowledge" is the knowledge exercised in the performance of some task. Procedural knowledge is different than descriptive knowledge which is also known as "declarative knowledge" or "business rules" or "knowing-that" which involves knowledge of specific propositions such as "Assets = Liabilities + Equity" or basically facts expressed using declarative sentences.  Procedural knowledge involves one's ability to "know how" to do something such as create a "proof of cash" or a "lead schedule" or how to put together a balance sheet.

What if you could take these Lego-like information block organisms and the "know-how" or procedural knowledge and the "knowing that" of declarative business rules and give all that to a machine, like the mindful machine for accountancy, because you could articulate that information using a global open industry standard and then get a machine and human to work together collaboratively to perform work?

That is the "know how" this accounting information systems specialist has figured out over the past 25 years.  I know how to represent those information blocks effectively  using my Seattle Method. I know how to represent the business rules in the form of a theory, like my Financial Statement Mechanics and Dynamics. I know how to represent accounting working papers and audit schedules using the same techniques used to create financial statements. I know how to represent financial reporting schemes in a manner that software can interpret effectively. I know how to build software that "fails loudly".  I know how to manage epistemic risk and build a complete system that works effectively.

What I don't know yet, but I am figuring out quite nicely, is how to represent that process knowledge and procedural knowledge in machine interpretable form. But; the following is a description of the Procedural Knowledge Ontology (PKO) provided by its authors:

The Procedural Knowledge Ontology (PKO) is designed to support the governance of procedural knowledge throughout its life cycle, from elicitation to management. It enables the explicit modeling of procedures and their executions, ensuring interoperability and effective knowledge transfer in industrial contexts. PKO is built on requirements collected from various industrial use cases and can be utilized by AI and data-driven tools to enhance procedural knowledge management.

In her article, Process Knowledge Management, Part IV, Jessica Talisman examines how a real process knowledge ontology can solve real industrial problems.

Process knowledge relates to how things get built in an industrial setting. The "closing books" of reporting teams and "audit bundles" of audit teams need to be industrial strength.  That is my specific interest; how financial reports get built and how audit working paper bundles get built. The creation of both a financial report and an audit working paper bundle that supports the audit of a financial statement are industrial processes, they are construction processes.

This is an informatics problem and an industrial engineering problem.

Informatics relates to the intersection of information, people, and technology and the practical application of computational systems; understanding how people will "live" in the digital realm within some specific area of knowledge that makes sense to users of that technology. Informatics is the conscious management of information, knowledge, and accumulated knowledge. Informatics spans the knowledge accumulation process of an

• individual member (learning)
• organization (institutionalized knowledge)
• area of knowledge (professional knowledge; subject matter)

Informatics has theories, principles, frameworks, and strategies that can be applied to solve information management problems. Informatics is about harnessing the power and possibility of digital technology to transform data and information into knowledge that people use every day. Informatics is about understanding how people will “live” in the digital realm with an elegance of design that makes sense to users of a particular technology. Informatics is about delivering the best user experience possible.

Similar to how a chef transforms a recipe using kitchen equipment into an unforgettable meal; informatics transforms the use of information and knowledge into a successful experience. Similar to an architect that transforms a building into a livable space by placing doors, windows, and utilities with functionality and ease; informatics improves “digital livability”.

Industrial engineering is an engineering profession that is concerned with the optimization of complex processes, systems, or organizations by developing, improving and implementing integrated systems of people, money, knowledge, information and equipment. Industrial engineering is central to manufacturing operations. Industrial engineers understand tools such as Lean Six Sigma.  Lean Six Sigma involves systematically removing operational waste, reducing process variation, and managing process quality. Industrial engineers understand Lean Six Sigma techniques such as how to use mistake proofing tools like poka yoke.

Creating the final financial statement product and the supporting detailed information of the "closing book" and "audit bundle" and coordinating all that work has characteristics of a "job shop" and an "assembly line". There are tasks and processes, the work performed, are algorithms that are sometimes performed by humans and sometimes performed by machines (i.e. sometimes algorithm steps can be automated and sometimes they cannot).

Algorithms are a well defined sequence of instructions or steps.  Algorithms are always unambiguous and are used as specifications for performing a task or process.

What if there really was a better way? What if we rethink things like the financial statement. What if we created a new vision. What if there was a universal global open industry standard for accounting and audit working papers. What if those accounting and audit working papers  really did "snap" together like those Lego-like information block organisms; think information Legos.

What if we could create a self correcting virtuous cycle using feedback loops?

Up until now it was impossible to do better. The beloved electronic spreadsheet, the accountant's "Swiss Army knife" is struggling to meet the real needs of accountants.  The truth is that the electronic spreadsheet was a "stepping stone"; not the "be-all, end-all" tool that some thing that it is.

Why can we do better now?  The environment has changed.  Fifty years ago we did not have the internet; but we do now. Twenty five years ago we did not have global open industry standard structured information exchange formats; today we do (XBRL, RDF).  We had rule-based artificial intelligence fifty hears ago, but it was too hard to use.  Probability-based artificial intelligence did not exist until about ten years ago.  The notion of a knowledge graph is about ten years old. How do you put all these pieces together?

What is new is that now machines can help humans more.  We can now effectively "team" a human with a machine.  We can create what I refer to as that "mindful machine".  What if we really did create a "mindful machine for accountancy". Another name for this is a knowledge based system.

What if we did put all these pieces together and created new procedures and documented that procedural knowledge such that it was understandable by both humans and also by the machines. What if we could capture important institutional knowledge such that the knowledge is retained within an organization rather than that knowledge leave the organization whenever an employee left the organization. What if we could communicate more clearly. What if we could reduce the threat of inaccuracy or eliminate inaccuracy altogether?

Not all knowledge is equal.  Explicit knowledge is knowledge that is easy to articulate, write down, and share. Explicit knowledge is objective, documented, and easily shared information and tends to be found in manuals, reports, documents, and books. Explicit knowledge is formalized and codified.  Implicit knowledge is the application of explicit knowledge. Skills that are transferable from one job to another are one example of implicit knowledge. Tacit knowledge is gained from unique personal experience that is more difficult to express and tends to be unwritten. Tacit knowledge tends to be more nuanced, experience-based information like intuition or a learned skill that is hard to articulate and understand. Tacit knowledge tends to be important subtleties and nuances that takes deep understanding to get right. And finally we have "common knowledge" which is knowledge us humans takes for granted but computer know nothing about. Remember, computer are dumb beasts. Everything has to be spelled out for them in detail for them to be helpful and reliable.

In an industrial setting, how to create a "closing book" or "audit bundle" is procedural knowledge, the "know how".  That procedural knowledge has both explicit aspects which are documented in manuals and other such documentation, but much of that knowledge is also implicit and tacit and exists only in the heads of your most talented, skilled, and experienced employees.  These experienced employees carry this tacit information in their heads; the process steps, conditions, judgment calls, and other such things that make these complicated procedures work.  This important procedural knowledge exists in experienced hands and heads of employees, as documentation in margin notes of outdated manuals, and in the institutional memory of workers who may retire (are retiring), or move on to other work at any time.

Without the institutionalized knowledge in the form of formalized knowledge representations of both the explicit knowledge but also the tacit knowledge, implicit knowledge, and common knowledge.  The absence of cohesive, formalized knowledge representations of explicit, implicit, tacit, and common process knowledge has significant consequences. 

When procedural knowledge cannot be accessed or reused by machines and people, organizations face increased compliance risks, higher error rates during procedure execution, and substantial friction in training and onboarding new employees. The challenge of capturing processes as formalized procedures intensifies as organizations deploy artificial intelligence systems that require that rich procedural context to function effectively.

Soon, accountants and auditors will have new superpowers and will realize a significant productivity boost by effectively harnessing the power of artificial intelligence.  Magic and wishful thinking will not make this happen, it will take hard work and know-how.  Artificial intelligence will not do this for you, AI is only a tool.  Accountants and auditors applying these sorts of new tools will get us were we want to be. What if we could create modern accountancy?

This is not about yet another incremental change to existing legacy systems (i.e. that kludge). Humans are sometimes underrated.  Elon Musk admitted that Tesla made a mistake and tried to over automate.  Another common mistake is to automate bad processes.  Automation works best when you follow the following fundamental rule: Add automation incrementally and only automate processes that are already working smoothly.  Rethinking accounting, reporting, audit, and analysis means going further than simply automation. Here are some good ideas from the article The Future of Audit: What Will Audit Firms Look Like in 2030?

Re-thinking the audit means going further than automation. It means evolving checklists into dynamic procedures that flex to client risk. It means using client data to expand coverage intelligently, not just pick random samples faster. It means building workpapers that link directly to schedules so tie-outs highlight themselves. It means pulling project management out of scattered spreadsheets and portals into a single system.

The shift is about changing both the unit of work and the unit of value. The unit of work moves from people following steps to systems executing policies. The unit of value moves from hours to outcomes: coverage, assurance, readiness. People still sit at the center but their focus shifts to judgment, handling exceptions, advising clients, and telling the story behind the results.

That’s what sets up the next step: imagining the audit not as a chain of manual steps, but as a system, a machine that brings together data, policies, and people in a completely different way.

Imagine being able to do things like Track-Trace-Walk.  Imagine all information that is connected to be leverageable in order to "drill down" from a financial analysis model to the supporting financial statement line items to the audit working papers (if you have proper access) and/or financial report closing book (if you have proper access) to the general journal to subsidiary ledger and ultimately to the financial transaction or event the business event that generated that financial transaction.  Or, you can navigate the other way from business event, to financial transaction, to subsidiary ledger entry, to general ledger, to trial balance, to lead schedule, to report writer, to financial statement line item and then maybe even to financial analysis model. You don't really need to imagine, this works today.

The beloved electronic spreadsheet, the "Swiss Army knife" is showing its limits.  The electronic spreadsheet will always be a useful tool, but it is only one useful tool.  New tools are necessary for new challenges.  The same technology that is making information more complex, increasing the volume of information flow, and increasing the pace of information flow can, if configured correctly, also be used to solve those problems.

Additional Information:

• Seattle Method Overview
• Process Knowledge Management, Part I: Accounting for How We Work
• Process Knowledge Management, Part II: Collection Development and Organizing Principles
• Process Knowledge Management, Part III; How We Lost Our Way
• Concept Models and Ontologies
• The Future of Accounting
• The Future of Audit: What Will Audit Firms Look Like in 2030?
• Neural Empathy-Aware Behavior Trees Meets Knowledge Graphs for Affective Human-AI Teaming