Wolters Kluwer headquarters in the Netherlands
It’s widely known that to succeed with AI, a company needs to have some distinctive information. Wolters Kluwer, the Netherlands-based professional information, software solutions and services company that does business in over 180 countries today, has never lacked for that resource. It was founded in 1836 as a schoolbook publishing company, and over the years merged with other publishers and eventually began developing and acquiring digital information capabilities.
Nancy McKinstry, the CEO and Chair of Wolters Kluwer, became its leader in 2003. She began to transform Wolters Kluwer into an expert solutions company, hiring and developing experts with deep expertise in areas like healthcare, tax, risk and compliance, and legal. The company also created a global Digital eXperience Group (DXG) to help speed time-to-market and innovation in digital products, as well as a Global Business Services (GBS) group to provide strategic execution services. Today, the company’s revenues from publishing are less than 5% of the total—down from over 80% when McKinstry became CEO.
By the time AI became more prevalent in the late 2010s, Wolters Kluwer was well into the business of providing “expert solutions” to its customers. At this point there were hundreds of experts in various fields providing expertise to customers, and fortunately Wolters Kluwer captured the data on the advice it provided and the outcomes for the customer. One might notice that this is a perfect situation to begin modeling and predicting those outcomes with machine learning. By 2016, the company had created its first AI-enabled product: CCH IQ used machine learning to help tax service providers identify which clients are affected by changes in tax legislation, assess the impact of the changes on client tax returns, and understand opportunities for additional tax services to clients.
Building AI-Based Product Capabilities
In addition to the data, Wolters Kluwer began to build the necessary capabilities to create AI-based products. Its Financial and Corporate Compliance (FCC) division created an organization—called Customer Information Management/Operational Excellence (CIOx) —to better understand customer needs and build new products and solutions. Members of the group are expected to spend time with customers and identify their needs. The goal is to create solutions that fit customers’ workflows and develop solutions that address their problems.
The group developed a product development process with the steps of “ideate, test, incubate, scale.” A team of product developers employs the process. The company also began to hire a number of data scientists and embedded them with domain experts. Its formula for product development success became “Data + Domain Knowledge + AI = Solution.”
The CIOx team, working closely with the business lines within the FCC division, also developed a set of principles for successful product development projects. For example, the team recognized that it needed an equal seat at the table with other team members, and should be viewed as data-driven, trusted advisors. They wanted to be seen not as a research group, but as having a solution mindset. The metrics for their success—and resulting reward mechanisms—should be the same as the businesses with which they were working. In short, they viewed themselves as “blue-collar AI workers” focused on achieving outcomes rather than simply developing great models.
In addition to high-quality people and operating principles, CIOx has also established a set of technology capabilities to accelerate AI-based product development. They have created a series of platforms that shield developers from legacy technology and enable faster data access and modeling. A set of data stores and feature engineering repositories are part of this capability. Wolters Kluwer has also adopted a machine learning operations (MLOps) workbench for sophisticated neural network model development, deployment, and ongoing monitoring.
Analyzing Legal Bills with AI
One of Wolters Kluwer’s most prominent AI products is called LegalVIEW® BillAnalyzer, offered by Wolters Kluwer ELM Solutions, part of the Legal & Regulatory division. It supports large companies’ chief legal officers in reviewing law firm invoices for compliance with outside counsel billing guidelines. It compares these companies’ legal services agreements with what law firms actually bill them for, and frequently finds errors in the bills. Beginning in 2017, LegalVIEW BillAnalyzer was based on human review by legal experts. It was a successful but labor-intensive product.
Now, however, AI systems extract key provisions from legal services agreements—and automatically analyze bills. LegalVIEW BillAnalyzer has over $160 billion in legal invoices to use in training machine learning models. For each line item on the bill, a model calculates a risk score—i.e., the likelihood of a billing anomaly based on provisions in the legal services agreement and adjustments in past data. If, for example, the outside law firm is billing 16 hours for a deposition that usually requires only four hours for that particular type of case, it will assign that line item a higher risk score. If the customer verifies that the line item needs adjustment, that data is used to refine the model and can increase the risk score.
LegalVIEW BillAnalyzer can save companies up to 10% on their outside legal service costs and increases compliance with billing guidelines by up to 20%. The system also saves in-house counsel considerable time on investigating instances of overbilling.
Generative AI Products in the Making
Wolters Kluwer, like many companies, is also experimenting with generative AI. But consistent with the CIOx focus, it has specific products in mind. One of the forthcoming generative AI-enabled products, for example, involves an existing product called OneSumX® ProViso. It employs AI technology, validated by regulatory experts, to monitor, capture, review, and summarize vast amounts of changing laws and regulations for banks, insurers, and other financial services companies. This has been a successful, expert-based product for several years. Now, however, the CIOx group is working to determine how GPT-4 can add value to the offering. It is having the language model read a regulation and summarize it with thus-far promising results. The technology appears to do an excellent job of creating a first draft and produces outputs similar to human financial services legal experts. Ultimately, the business is working with product owners to identify potential use cases and understand how ChatGPT and other generative AI solutions might help its customers.
Wolters Kluwer also has maintained legacy computer systems throughout the company that have been installed for many years and are specific to a particular content domain. The company has explored the idea of training its employees in old and new programming languages in order to translate the old code into new code, but this is very time-consuming, and it is difficult to attract and retain talent for the job. But since Wolters Kluwer is often working in compliance-oriented domains, it has always been important to preserve the quality of program code.
Now the company is exploring the idea of using GPT-4 to translate obsolete program code into new languages, as well as to produce documentation of the new code in English. Notably, product operations specialists in several business areas are working with data scientists to analyze and perhaps rejuvenate over one million lines of code. Wolters Kluwer is adding artificial intelligence capabilities into its products, allowing it to supplement deep domain expertise with the latest technology.
The Ongoing Role of Humans
Wolters Kluwer has no intention, however, of using AI to replace its human employees. The company has invested heavily in employee education and “upskilling” programs, with the assumption that jobs of the future will rely heavily on AI and other new technologies, working side by side with its experts. Grassroots communities within the company, including one called “Addicted to Learning” with 800 members, have arisen to encourage further learning about new technologies, including AI and generative AI. The assumption is that such focused education will drive innovation and future demand for Wolters Kluwer products. Thus far, at least at Wolters Kluwer, this maxim has proven to be an accurate indicator of innovation for 187 years.
