Wolters Kluwer is introducing generative AI to its tax and accounting suite. The company has also taken steps to dramatically reduce the possibility of error—the widely talked about hallucinations.The company has added GenAI to CCH AnswerConnect, its tax research software, to provide faster research and to enable customers to employ plain language in framing queries.
However, it is not allowing the system to find information outside of Wolters Kluwer’s own source material, according to Dean Sonderegger, SVP and general manager for Canada and Research & Learning for Wolters Kluwer TAA North America.
“We ground the result so that is only based on our content sets,” says Sonderegger. “It significantly reduces probability of hallucination.”
With AnswerConnect in a beta mode, the plans are for a research summarizer and a client letter generator. WK also introduced a invitation-only AI Labs for select users to work with products under consideration.
A major goal of using GenAI in research is to speed up the process. “It doesn’t obviate your need to do your diligence,” Sonderegger says. “But takes 80 percent of the time away. He notes WK has not performed a time-and-motion study to quantify the savings.
WK research has typically used a keyword relevant-based search “that gets boosted by our editorial folks,” he says. “We also augment that so when certain questions come trough we have an editorially curated answer comes back.”
Gen AI can more quickly return definitions and cite relevant results and generally enable workers without great expertise to conduct basic searches.
It can also be used to write summaries that can lessen the amount of time someone conducting a search must spend writing. The resulting narrative, Sonderegger says “doesn’t have to be 100-percent finished.”
Gen AI can produce a first draft that is 80-percent finished which can then be edited. can so be used refine writing—such as asking the system to remove jargon or simplify the writing.
Sonderegger says Wolters Kluwer envisions GenAi enabling firms to get more out of data—detecting “quiet patterns of fact and circumstance”. For example, instead of providing generic information about tax consequences of divorce, GenAI-equipped AnswerConnect should be able to tailor answers to a specific couple’s circumstances.
