How AI learns from your general ledger and historical data

Brett Norton

Brett Norton

SVP, Engineering

Save time and improve accounting efficiency through AI-fueled technology that continuously learns from your general ledger and accounting data.

March 15, 2022

5 min read

How AI learns from your general ledger and historical data

Do you remember the cost on the last 20 invoices you approved? Accounting AI does.

Imagine recalling all of the dimensions on the last 20 invoices. It's difficult even when you are familiar with your accounting process.

All companies differ when it comes to their general ledger accounts. Even if your brand new accountant comes from the same industry, it can still take a considerable amount of time for a new team member to learn the ins and outs of your company's chart of accounts, general ledger, or accounting systems. 

Learning these specifics can be very mentally demanding as well as time consuming during the onboarding process as you expand your finance team, which can cause longer ramp times for new staff to get caught up to speed. It is also important to note that each company has a different take on how to classify their costs into various general ledger accounts when it comes to dimensions like department, location, and more. While these variations are important in terms of classification, it requires new employees to learn from the ground up every time you bring someone new on board. 

So how can companies offer stellar training for their new accountants while also cutting their onboarding time down?

The implementation of new technologies, such as Artificial Intelligence can make a world of a difference when it comes to managing your general ledger accounts. More specifically, the implementation of a high-quality AI platform that takes variations of cost classification into account and understands the different kinds of incoming costs. Your company’s myriad of costs need to be coded differently and accurately to their proper GL accounts. With these capabilities, AI can both improve your company's onboarding time and create a more efficient management system with less human error. 


How does accounting AI learn from your General Ledger?

Much like anything you might buy, AI can vary greatly, from skillset and effectiveness to quality. There are two major factors that impact how well an AI platform designed for accounting classifies costs. The first being the algorithms and the sophistication of its technology, and the second being the amount of data used for training the technology. As we know, an AI platform is entirely driven by data, so the more it has to evaluate, the better. 

Fun Fact: has trained our AI on more than half a billion invoice documents!

Imagine if your accountant was able to accurately remember the last 500 invoices they processed, and your accountant used that information while processing upcoming invoices. Looking at the data another way, by knowing due dates on received invoices, the AI can not only sort and classify costs into their proper GL account, but even predict cash flow more accurately. It is ideal to have historical data for an AI system to work properly, so it can understand future invoices or purchase orders.

With's vast amount of invoices processed, it can read header-lever information and accurately code to the GL on day one for new customers. Without historical invoices, it'll need a little coaching from accounts payable staff to hit the ground running and learn how to understand and code hundreds of line items.


Unique Local General Ledger Accounts

So how does implementing accounting AI work for each individual organization or entities within that organization? 

Extracting and reading data from invoices is very similar, regardless of the company or industry, while the general ledger, accounts, departments, projects, and dimensions are unique.

If your accounting team is responsible for managing multiple entities or properties, your team has to constantly switch between different buckets of knowledge every time they work on a different entity. This type of routine can yield negative consequences such as chart of account mix ups, fatigue, and a lack of focus, which can lead to otherwise avoidable mistakes. This also takes away the ability for others to step in and help perform accounting work if you experience turnover or AP staff needs to take a leave of absence.

A well trained AI system can work on any entity at any time, which is a significant advantage. Your AI platform needs to support each companies’ list of general ledger accounts and learn and operate on a very granular, single-entity basis.

The platform is designed in a way where it draws on both a layer of global knowledge, as well as knowledge built up on every specific client to reach its conclusion. It understands and perceives things on a global basis. But it’s also engineered in a way where it learns the nuances of each company to be able to determine the specific GL dimensions for your accounting system.

AI promises to do something different than we’ve seen before. While it’s very easy for the human brain to understand a rule, machine learning is very different from a rules-based system. Machine learning is smarter because it uses big data and has super-human memory that makes it possible to follow trends in real-time.

For example, if the data comes in labeled as Uber, the system will see that as a transportation cost. Then if it sees something come in labeled with Lyft, it will associate that as a transportation cost as well. A rules-based system can’t do this very well, and it doesn’t scale because you have to copy or recreate the rule for every client.

With historical knowledge of common transportation companies, the AI is able to classify these vendors in the same category, automatically. If there is a lesser known  transportation company from an executive who travels to a city without Lyft or Uber and decides to use Silver Fox Limo company as a last resort, the AI may not detect the vendor category the first time it sees the invoice. Once the accounts payable professional tells the system Silver Fox Limo is a transportation cost, the AI will recognize and classify it the next time someone else decides to take a client out in a fancy stretch Hummer.

Artificial intelligence Learns From Your General Ledger Accounts

Machine learning is much more powerful than rules-based automation because it is able to learn patterns and habits from past behaviors to make more accurate and reliable predictions for the future, much like humans. It also provides you and your team insight on its ingestion status through different levels of confidence. When the confidence level is low, it will let you know when to pull in a person for help, as opposed to when the confidence level is high and human intervention isn’t needed, which is called Autopilot. In the image below, you can see how it shows in the interface.

Autolipot invoice processing

This is how works. Let’s look at how the system works on real clients. 

BDO chose because it gave them the ability to see the confidence level in the automation as a part of the view. So if the confidence level was showing 30%, you’ve likely found an area that needs a little more focus on some of the details. If it’s showing 100%, the review would not need to be as in-depth. This was a huge part of the decision-making process on why Ignite Spot chose confidence levels show invoice processing status

Ignite Spot also liked how was able to be flexible with the client’s industry. There’s nothing restricting the industry type from functioning on the platform, where it’s real estate or a food and beverage company. 

So how does machine learning apply to invoice processing?

Machine learning utilizes large amounts of data and information to give you the best recommendations for processing and cost classification. Through the use of sophisticated algorithms, the AI has step by step instructions on how to read your invoice and uses the information provided to deduce General Ledger account classification. Additionally, the more the AI is used, the more it learns from your interactions and patterns, therefore becoming more versed in the variety that often occurs in invoices. 

With systems like, the more it is used, the smarter the system becomes, allowing it to become more accurate over time and eventually operate independently. 

It read the contents of the accounts payable documents and looking through the entire sheet to determine the best recommendations for your invoice number, the date of purchase, vendor information, or any other key information. This really helps to make the process more efficient and keeps it running smoothly. allowed Ignite Spot to mitigate human error that comes into play when information is typed manually or read incorrectly. Something as little as a smudge or a tear in the page could cause human error. It can even detect duplicated invoices. If an invoice previously processed enters the system, it indicates a duplicate with the red box shown in the image below. detects duplicate invoices and prevents double payments

Enhancing your business applications with AI algorithms can transform your accounts payable workflow entirely, allowing you to perform tasks more efficiently and accurately.

Reimagine your financial operations with Artificial intelligence for accounting and invoice processing

Autonomous invoice processing replaces legacy OCR and rules-based methods. In contrast, invoices can automatically be ingested into the system from various sources, including electronic file formats and EDIs, e-mail, PDFs, direct connections, and more. 

Once extracted, the invoice data is reviewed by the AI, which then matches and processes all relevant invoice information, including vendor, dates, numbers, cost accounts, dimensions, assets, and purchase orders. The software is designed to work across one or multiple ERP systems, and the integration can be set up using the company’s API or flat-file support.

By doing this work for you as a CFO, controllers, accounts payable team, or finance and accounting team, you can be freed up from time-intensive invoice processing to dive into more critical functions such as spend intelligence, benchmarking, and cost optimization.

Are you ready to enter the new era of accounting and upgrade your invoice processing software?

Begin your autonomous accounting journey with a solution that understands and works alongside your AP function, and is not limited to a predetermined set of rules that in many ways further complicate accounting.

Click the image below to download the Financial Revolution E-guide now:

10 ways autonomous AP processing will revolutionize your financial operations


Sign up to receive the latest news, updates and valuable accounting and finance resources conveniently in your inbox.