Cushman & Wakefield (NYSE: CWK) is a leading global real estate services firm that delivers exceptional value for real estate occupiers and owners. Cushman & Wakefield is among the largest real estate services firms with approximately 51,000 employees in 400 offices and 70 countries. In 2018, the firm had revenue of $8.2 billion across core services of property, facilities and project management, leasing, capital markets, valuation and other services.
In the United Kingdom, C&W’s Business Rates Management Area provides a service to clients with large national real estate portfolios which includes management and payment of business rates payable to local councils. These Rates Demands are sent annually and are predominantly paper-based, resulting in the Rating Team processing over 10,000 Rates Demands from 400 different Councils in a 2 month period every year.
The Rate Demands were opened, sorted, manually reviewed and compared to estimates on the Riverlake Rates Management system to allow these to be amended, or approved and paid. Once processed, the paper demands were filed away, making it difficult to retrieve these quickly when queries arise later in the year.
Today, the Business Rates Management team scans all inbound rates demands and emails them to the Rossum Elis Artificial Intelligence (AI) Computer Vision/OCR Solution.
A human reviewer rapidly checks the scanned details in a validation interface, built within the Rossum platform, and marks validated documents as ready for export.
The Open Box automation engine performs a daily retrieval of all exported documents from Rossum, including downloading the scanned Rates Demands PDFs and uploading a copy to a SharePoint location.
The details of the Rates Demands are automatically compared to the estimates in Riverlake, and if they are a match then the demand is automatically queued for one-click approval in the system.
A link to the SharePoint copy of the file is added to the record in Riverlake. Any demands that do not match are summarized in a report that is sent to the Rates Management team for resolution. The business rates administrators are then able to review issues in bulk, make any necessary corrections, and send the file back for reprocessing.
Tech Focus: Data Capture
Rossum's cognitive data capture solution seamlessly processes the email attachments containing the scanned documents and performs an automated data extraction. This data capture consists of two stages: a localization of each data field on the page and an OCR of the field value. The automated data extraction can happen on both “trained” document layouts and previously unseen documents.
The Rating Team have reported many positive effects of the solution:
While additional effort is required to scan the demands when they arrive, this is offset by no longer needing to sort, date-stamp, label and file the paper invoices. All invoices are uploaded to SharePoint as well as linked to the item in Riverlake for easy retrieval.
The downstream impact of having the files digitised was immediately obvious. A client requested a copy of all the demands for their 50 properties in the middle of what is usually the Rating Team’s busiest period, and the team was able to retrieve these and respond within 20 minutes, a task that would have taken much longer when sifting through paper copies.
The bulk validation of the files against the Riverlake estimates allows the majority of the checking and correction to now be performed by the business rates administrators rather than the managers and senior managers who previously did the demand-by-demand comparison and approval. This allows them to prioritize other activities and focus on providing optimum service to clients.
Managers now are only required to review demands that have genuine issues that need to be queried with the councils (about 10% of demands); simple typos and incorrectly entered information can be identified in bulk and corrected by the administrative team.
The ability to identify causes of validation failures at a high level, and in bulk, has created insights into the team’s upstream processes that were not evident when processing the files one at a time. Process changes have already been identified that can be actioned throughout the year to improve the quality of the data in Riverlake and drastically improve the number of automatically matched demands in 2020.