A Framework for Autonomous Accounting

Today we are excited to share a framework for how (we believe) artificial intelligence will impact the accounting industry.

Similar to autonomous driving, self driving cars didn’t happen overnight. They required intentionally-built, robust systems combined with advancements in artificial intelligence to unleash their potential. In San Francisco, it is now common to see cars from companies like Cruise driving passengers with no one in the driver’s seat.

We believe Autonomous Accounting will be similarly ubiquitous and transformative.

💡 Definition: Autonomous Accounting
‍Autonomous Accounting is a comprehensive system that can generate complete and accurate financial statements, with accompanying financial metrics, insights, variance analyses, and reconciliations in real-time, without human involvement.

‍

Why autonomy matters when choosing an accounting solution

Speed and accuracy of information provide a competitive advantage — and accounting should play a more significant role in achieving this advantage. But much of an accountant’s month is consumed by manual, repetitive work for future compliance, rather than dynamic business building today.

To make matters worse, accounting is becoming an increasingly unattractive profession, creating risks to accurate and validated financial information across organizations. The WSJ reported that 300,000, or 17% of all CPAs and auditors, quit in the last two years, and that accounting is a declining college major.

Autonomous Accounting has the potential to provide real-time insights for a company's most important financial decisions, maintain accurate and auditable records for compliance and taxes, and make the accounting profession more dynamic and attractive as a career.

‍

Level 0: No Automation

With level 0, there is no driving automation. Most vehicles on the road today are level 0 automation. Human drivers manage the route, speed, and direction.

For centuries, accountants prepared financial statements by manually stitching together financial information using paper, hand-written ledgers, and eventually calculators. Spreadsheets provided many benefits, but the entire financial statement creation process remained manual.

Level 1: Task Automation

With level 1 driving automation, specific tasks are automated to augment the driver (driver assistance). Cruise control, automatic parking, and lane assistance each provide support for one element of driving without taking control.

With level 1 accounting automation, some tasks are automated to assist the human bookkeeper. However, a significant amount of manual labeling and calculation is required. Humans review and approve transactions (usually in a data quarantine) before they become part of the financials, and most tasks are not automated.

Market examples:

• QuickBooks and Xero automate many tasks such as bank feeds and application of user-created rules.  Today, these are table stakes in any core accounting solution.

Level 2: Function Automation

Level 2 driving automation uses advanced driver assistance systems (ADAS) to create partial driving assistance by automating a single function of driving, from steering to acceleration to braking, but a human still needs to be in the driver's seat and remain aware of the surroundings.

Level 2 accounting automation can perform a complete set of tasks, such as creating and reconciling a subledger or executing a data transformation workflow, but human oversight and adjustments are still needed to generate complete and accurate financial statements.

Market examples (core accounting):

• Quickbooks and Xero include some function automation if you use their bundled products, such as payroll and accounts receivables. These modules deliver a complete function and the associated journal entries to the general ledger.

Market examples (other accounting workflows):

• Ramp uses built-in learning to categorize payments and push journal entries to the general ledger. They use AI for document parsing, receipt matching, and reconciliation.
• Digits uses vector embeddings in their categorization engine to vastly improve categorization of all bank feeds, as well as interactive dashboards powered by AI.
• HighRadius uses AI-based workflows for financial close, accounts receivable, and treasury management.

Level 3: Problem-Detection Automation

Level 3 driving automation involves advanced “environment detection” capabilities that significantly — even if subtly — expand the comprehensiveness of Level 2 automation. They decrease the risk of applying Level 2 automation because they can identify when to alert the driver.

Level 3 accounting automation can generate financial statements, but also perform validation checks and identification of anomalies — this is the “map” to define the path to more accurate outputs and a “vision system” to identify issues and exceptions from perfect conditions. Accountants can make informed decisions on high-risk areas while relying on Autonomous Accounting to identify standard tasks and execute them consistently.

Market example:

• Puzzle is the first core accounting system to achieve level 3 automation with a system that generates financial statements with minimal human intervention, while proactively monitoring the outputs for accuracy. Learn how Puzzle’s Autonomous Accounting system works.

Level 4: Problem-Solving Automation

Level 4 driving automation involves the vehicle’s ability to actively respond to complex and ambiguous situations, known as conditional driving automation. While a human can take control at any time, the autonomous vehicle can operate in full-self-driving mode in almost all circumstances.

Level 4 accounting automation generates entire financial statements while proposing solutions (such as adjusting entries) to detected anomalies such as accounting policy violations, missing transactions, or reconciliation differences.

Market example:

• We have not seen any core accounting solutions with full level 4 capabilities. However, some elements of proactive response to anomalies have proven possible and will become more prominent.

Level 5: Problem-Resolution Automation

Level 5 driving automation do not require human attention to get from point A to point B. The “dynamic driving task” is no longer performed by a driver and the car can adapt to changes in weather and road conditions without intervention.

At level 5, Autonomous Accounting do not require human attention to generate a full accounting package. Once configured, it’s a self-sustaining system that executes against policies and controls, adapts to solve inconsistencies and anomalies, and delivers insights to facilitate management, board, or resource allocation discussions.

Market example:

• None. The most significant challenges to reaching this level include technical challenges such as data that is not integrated into the core accounting system, multivariate calculations that require both data transformation and judgment, and judgmental decisions that differ from industry to industry.

What does Level 5 mean for accountants?

We believe that accountants offer far more value than processing routine transactions, performing reconciliations, and producing financial reports, across all industries and company sizes. According to a recent study conducted by FloQast and the University of Georgia, 65% of accountants aspire to take on a more strategic role within their organization. McKinsey research cites none of general accounting operations are “difficult to automate.”

Autonomous Accounting will help accountants become more valuable strategic partners. In-house accounting teams will be more efficient, and professional service firms can take on more high-value business. With less manual preparation, accounting will become a more analytical profession, using financial knowledge to separate signal from noise amidst abundant financial information.

The Autonomous Accounting Framework: a final word

This framework is exactly what its name suggests: a framework, and not a perfect one.

Its focus is on the core aspect of accounting: generating financial statements for taxes, compliance, and fundraising. Other accounting automation solutions have been included to highlight the innovative advancements happening in the field. With technology evolving rapidly, we anticipate the emergence of new players and the re-architecting of core infrastructure by legacy players to suit the AI age.

We welcome feedback — the framework will evolve along with the industry, technological advancements, and your input to help us improve.

Want to experience level 3 accounting automation? Sign up for Puzzle.

‍

Thank you to Andrew Robinson (CPA) and Alice Ko (CPA) from the Puzzle team for co-authoring, and thank you to our advisory board Andy Toung, Casey Woo, Emily Westerhold, Kathy Ryan, Lauren White, Mike Etheridge, Steve Love and Wesley Gutknecht for reading drafts.

Autonomous Accounting is a registered trademark of Puzzle Financial Inc.