LEVERTON AI Automates Lease Review Ahead of IFRS16
How artificial intelligence is set to revolutionise lease management in the telecoms sector.
LEVERTON AI have developed an artificial intelligence solution which can automate and streamline the extraction of important information from lease contracts, saving hours of time in reviewing legal documents. TowerXchange speak to LEVERTON AI’s Global Sales Director, Richard Belgrave, to find out more.
TowerXchange: Please can you introduce LEVERTON AI and when the company came into existence?
Richard Belgrave, Global Sales Director, LEVERTON: LEVERTON AI was founded five years ago, originally as a commercial spin off from the German Centre of Artificial Intelligence. The company has created a platform which applies artificial intelligence (AI) and deep learning to automatically extract, structure and manage data from complex corporate and legal documents. Originally our focus was on the real estate market in Germany, extracting key pieces of information from lease contracts, but we have since expanded into further sectors, countries and languages. We now have offices in Berlin, London and New York and our platform is capable of reading various different types of corporate and legal documents in over 30 languages.
TowerXchange: Please can you explain the platform in more detail?
Richard Belgrave, Global Sales Director, LEVERTON: LEVERTON’s platform is a cloud and web based system which is used to turn unstructured documents into insightful, structured data.
The first step is to turn the document into machine readable data. A PDF, which is the format that original documents are typically found in, is actually an image and so this must first be turned into text. Our system is capable of recognising different document formats automatically. The second step in obtaining the data requires the overlaying of a neural network, whereby deep learning algorithms automatically extract the required information from the document. This information could be contract length, renewal date or any number of other metrics and details. This data is then transferred to a structured repository linked to the original document. We use an open source format and so the data can be exported into excel or uploaded to third party systems.
TowerXchange: What level of accuracy is the system capable of achieving?
Richard Belgrave, Global Sales Director, LEVERTON: The speed and accuracy of the process is dependent on the document type and the language it is in. In the German real estate sector where we have the most experience, the system gets 92% of data correct and reduces a four hour manual review of a real estate lease to just 15 minutes.
When the data is extracted from the document it is displayed in a panel adjacent to the original document, with the extracted data linked to the original text (figure one). It is possible to review and validate the extracted data with a simple click and in doing this you are effectively training the machine to become more accurate.
TowerXchange: Why did you see the potential to apply AI to the management of lease contracts in the real estate sector and what benefits do you see this bringing to the telecom industry?
Richard Belgrave, Global Sales Director, LEVERTON: Generally, when it comes to real estate, companies often have a large number of high value assets which typically have long, detailed and somewhat verbose lease contracts. Reading through such contracts to extract key information such as billing amounts, lease renewal dates et cetera is an extremely time-intensive task and a suboptimal way of managing things. Lawyers can spend weeks reading through boxes of contracts, during which time the bill continues to mount up for the client, making the process not only laborious but also expensive. LEVERTON’s system can be trained to extract and structure such information automatically.
The telecom industry also has a large number of lease contracts which need to be reviewed and the introduction of the new IFRS standards has made the review of such lease contracts increasingly important. With the standards coming into effect as of January 2019, there is a pressing time issue when it comes to studying and extracting key data from such contracts, determining whether the leases should belong on or off a company’s balance sheet.
TowerXchange: Please can you tell us some of the clients that you have worked with?
Richard Belgrave, Global Sales Director, LEVERTON: LEVERTON currently works with over 100 corporate and investor clients across the globe. We have a strong footprint in the real estate sector having worked for major companies such as JLL, Sports Direct and Siemens and our software has also been used by a number of investors including Blackstone and Invesco.
In the telecom industry, our first client was a German property manager with a mandate to review 800 leases for Deutsche Telekom every month. Because of the homogeneity of the data, our system reached 98% accuracy in reading those leases in a very short period of time saving a huge amount of time and energy in the process.
TowerXchange: Finally, what differentiates LEVERTON AI from competitors in the field?
Richard Belgrave, Global Sales Director, LEVERTON: LEVERTON uses a form of AI called deep learning. Deep learning uses neural networks and therefore functions like a human brain, whereby the more information we feed our platform the more intelligent it becomes over time. Often, other AI based providers use a more rule-based approach, in that their platforms need to be constantly taught how to interpret information. If you were to consider these ideas in-light of document abstraction, LEVERTON needs to see a few hundred instances of a document type to learn how to interpret that document, agnostic of the documents structure. With more rule-based approaches you constantly need to teach the platform where to find information in a document, as they can’t automatically interpret unknown structures. Due to the nature of the AI, LEVERTON’s clients are buying into a shared ecosystem of everything that has been processed before, rather than having to teach a tool from scratch. Another way in which LEVERTON differs is our focus area.
LEVERTON has traditionally focused on extracting information from leases and so we have a very good grip on this document type. When you consider our AI, by training our platform on a single document type, we have ensured the intelligence of our platform is not diluted.
The article was originally published on the TowerXchange website.
Author: Laura Graves, Managing Director – EMEA, TowerXchange