How AI can predict tax outcomes
There may be nothing more complicated for a business than international tax. That’s especially true for U.S. companies that want to operate in Canada, and if they don’t follow the rules, says Fabio Bonanno, they could end up with a hefty tax bill.
Bonanno is the senior manager of corporate tax at Trowbridge Professional Corporation, a Toronto accounting firm. It’s his job to help companies navigate Canadian tax law, and for international operators there’s one central question that creates a lot of uncertainty. “Tax rules apply to companies that are ‘carrying on business in Canada,’ ” he says. “Unfortunately, that’s not defined anywhere in the tax legislation. It’s very vague.” It can depend on any number of facts and circumstances, from whether Canadian employees will be hired to where contracts are finalized.
Since January, though, Bonanno has been using Tax Foresight, an artificial intelligence (AI)-enabled tool, available from Thomson Reuters, that scours decades of case law to determine how courts and tax authorities might view a particular issue.
Saving time with AI
Until he started using the software, Bonanno could spend up to a week reading case law and then determining for himself if his client must pay tax or not. With Tax Foresight, accountants input their client’s problem into the program, which then reads previous cases and determines whether the CRA would decide if, in this case, a company’s activities count as carrying on business in Canada.
With Tax Foresight, Bonanno can take those findings and figure out if he agrees or disagrees with them. But all the heavy lifting has been done for him. “It reduces the time it takes me to do this by at least half,” he says. “I’m not wasting time filtering through unnecessary cases, so I can spend more of my time advising.”
Experts can still interpret the findings, but the program does the heavy lifting. “It reduces the time it takes me by at least half.”
Reducing research time
The program was created by Blue J Legal, a Toronto-based start-up co-founded by Benjamin Alarie, who is also a tax professor at the University of Toronto. Alarie, along with Albert Yoon, Anthony Niblett and Brett Janssen, started the company after he saw that his students learned tax-related case law for his exam, only to promptly forget it afterwards. “There’s no way they can really memorize all that and they can’t go through 10,000 tax court cases anyway,” he says.
So Alarie, his colleagues, and a group of U of T computer scientists went to work developing an AI-enabled program that could not only read copious amounts of case law, but also determine with 90 per cent accuracy how a court would view different situations. It also learns more as new cases are added to the system.
While this is ideal for accountants who need to provide advice to clients, it’s also helpful for government auditors, lawyers and judges who deal with tax-related cases.
Always learning
The software isn’t guaranteeing what a court would say, says Alarie—and it’s still important to get legal advice—but it can give people insights into a variety of possible outcomes. For instance, after Bonanno enters his client’s information and gets a conclusion from the software, he’ll change the parameters he had previously input to see what that client might need to do to get a more positive outcome. “We might change the fact pattern to see what kind of impact that might have,” he says.
Currently, Tax Foresight, which officially launched in 2017, can provide feedback on 25 issues, and more are coming. That’s the promise of AI—it can learn new rules as case law is added to the software, it can adapt based on how its users interpret its data and it can work far faster than any human. “It’s not science fiction,” says Alarie. “This is happening in real time and it’s only getting better as we address more cases and adopt the newest algorithms. It’s becoming better at predicting more of the issues tax folks confront every day.”
While the program has several useful features, Bonanno is a fan of its Navigators and Case Finder functions. The former has helped him determine capital cost allowance classes, which “can become convoluted,” he says. The latter has helped him reduce research time significantly, he says. All he has to do is input the facts of the case he’s dealing with, and the program then provides him with all the relevant court cases. “It will provide a list of cases with this same fact pattern,” he says.
This program, and the AI tools of the future, will create more engaging work, says Bonanno. He’s already pleased at how the program operates and how it’s given him more time to focus on important client work. “It’s making us better professionals,” he says about AI. “It makes things more efficient and we can spend more time talking to clients as opposed to doing administration.”
Watch Video:
*(Updated August, 2018: The headline on this story has been revised.)