With deadlines looming, organizations are racing to get the lease data necessary to comply with updated FASB ASC 842 and IASB 16 rules — especially the public companies that have a deadline of early 2019.
Artificial Intelligence (AI) technology for lease abstraction can reduce the time is takes to extract relevant data from piles of hard-copy lease documents. Many organizations are evaluating AI over time-consuming manual lease abstraction, which can take as much as five hours per lease to distill all the qualitative and quantitative data points. When you have hundreds or even thousands of leases in need of abstraction, that can add up to a significant amount of time and resources.
AI tools can help speed up the lease abstraction process to procure the information you need to meet your end goal: gathering essential data for FASB/IASB lease accounting as well as information for ongoing lease management. But can a technology takeover provide the quality you’re looking for? Can your organization rely on AI tools?
Keep these six points in mind when considering the use of AI.
1. AI is only as intelligent as you make it.
AI software must first train itself to understand and decipher the information you need for lease management and compliance purposes. AI tools can be reliable for extracting important data points that are necessary to carry out your calculations — but they are also very literal and will only abstract exactly what they are told. To go beyond this, AI tools require precise training, sometimes involving tens of thousands of documents to get up to speed.
When relying on AI, your organization may need to enlist experienced human lease professionals to review what the tool has assembled, as the intelligence is far from error-proof. Some tools, for example, may copy any language that matches a key term or overlook synonyms for a key term, which can entirely derail your data collection.
2. AI isn’t always AI.
Because Artificial Intelligence is an emerging technology, there are no real standards for using the term when describing a commercial software product. For example, AI and Optical Character Recognition (OCR) are often used interchangeably, though the terms are not synonymous.
An OCR system can scan hard-copy text — such as typewritten or handwritten PDF — and interpret it into machine readable text. The process results in the rudimentary extraction of key words and phrases. True AI, in contrast, applies machine learning to the lease abstraction process, “training” itself as noted above in order to improve its success rate over time.
3. Costs and volume may be a barrier to entry.
There is a volume requirement to conduct proper AI (machine learning). Because it takes time and a significant amount of work to train an AI tool, that can translate into significant upfront costs to develop a lease abstraction process. It may cost thousands of dollars just to begin putting your records on an AI platform, then hundreds per record to perform the lease abstraction. For this reason, AI is only practical for large, complex asset portfolios rather than small- to mid-sized portfolios.
4. Critical data points: Think beyond immediate needs.
To perform the calculations you need to be compliant with the new lease accounting standards from FASB and IASB (IFRS), you need to capture a specific set of data points.
For certain areas — equipment leases or simple leases for example — AI may be a sensible option for extracting very basic data points like dates or payment amounts. But you will likely need manual intervention to capture the more complex FASB data, or important data that you’ll want to extract for legal and administrative purposes.
To position your organization for better lease visibility for the long term, you should consider a thorough abstract between 150-200 data points, depending on who is going to be reading the abstract and how the information may be used.
5. Time savings can be deceiving.
When comparing manual versus AI-driven lease abstraction, be sure to take into consideration the “human intervention” time that will still be necessary for the latter.
Manual lease abstraction — It could take up to 4 hours for a full abstract of data from a standard commercial lease, while FASB compliance-relevant data may take 1 hour to extract. An equipment lease could take 30 minutes for a full abstraction, while FASB-only data will likely take less than 15 minutes.
AI-supported lease abstraction — You’ll gain a head start by automating 5-50% of the lease abstraction process. However, keep in mind that a thorough quality assurance (QA) review conducted by trained professionals will still be necessary. And it is quite difficult to put a time stamp on the QA process.
6. The human element: Still a requirement.
Despite its automated properties, AI still requires administration by lease abstraction professionals as mentioned above. These are the professionals who review each and dissect each document to identify the multiple documents within, then organize and drill down each individual component.
AI can take over the organizational task, but the rest is up to a set of eyes that are keenly focused on the details. If there is an error or oversight, AI does not make corrections. That also requires a professional to go into the documents to manually implement changes.
The most effective approach to lease abstraction: Technology & people working together
Consider a hybrid strategy, using automated tools to meet the easier data requirements and depending on human experts for more complex lease abstracting and data validation. If you are going to take advantage of automated lease abstraction technology, you still want to have professionals administrating it, as well as overseeing the QA and any complications that may arise.
As you develop your implementation timeline and choose lease accounting technology, you must have a realistic plan for extracting the data from hundreds or even thousands of lease documents. For many, lease abstraction is the most time-consuming part of the FASB/IASB readiness process, but when used in conjunction with a professional eye, AI tools can help speed along the process for many large organizations.