Reports of how artificial intelligence is now—or will soon—affect all areas of business have inundated board members and management. From supply chain optimization and inventory management to enhanced customer relationship management systems and dynamic pricing to personalized shopping and even optimized retail store layout, AI seems poised to revolutionize the way companies work.
Amid the shockwaves, there is one application that may not be getting the sort of front-page attention ChatGPT is attracting, but it offers companies—right now, and not 10 years from now—a whole lot of bang for the buck: AI for tax. By helping tax departments keep track of various laws, adhere to compliance regulations and prepare error-free returns, AI has already added efficiency and accuracy—and it’s also giving companies a way to cope with a shrinking talent pool in tax and accounting.
Perhaps most germane to boards is the technology’s ability to help manage risk.
As Michael Charette, leader of RSM Canada’s tax technology consulting practice, explained in an interview with Chief Executive magazine, AI can identify potential areas of risk within a company’s tax strategy by analyzing past audits and identifying patterns that led to scrutiny in the past. By flagging these issues before they cause problems, companies can make proactive adjustments and avoid costly fines and penalties.
For example, looking ahead at probable regulations involving real-time analyses of value-added and sales taxes, companies must prepare now while there’s still time. “There won’t be as much lead time as people think,” said Charette. “They’ll make an announcement, and 12 months later, you’ll need to comply, so getting ahead of this regulatory environment from a data point of view will be really important.”
The following conversation has been edited for length and clarity.
With so much on the board agenda, why should directors spend time and resources on automation in tax?
It has become a necessity from a talent perspective. You have a labor shortage in general around professional services. We’re graduating fewer people in accounting, and even fewer people are becoming CPAs (certified public accountants)—and then even fewer people choose tax after becoming a CPA. So at some point, no amount of money is going to bring people in because they just don’t exist.
So then you have two options: You can either outsource your tax department or automate. A lot of that outsourcing is offshoring to other countries, and they’re doing what robots would do if we took the time to build the robots. So the question for any professional services firm and the whole industry is if we get our data organized and straightened out, if we have consistency within our organization or if we do the legwork of knowing what’s what all the time, what could automation do for us?
And, of course, as technology costs go down, there will be a return on investment. That’s the biggest challenge right now, showing a provable ROI for automation in tax. If I have a tax department of five people, and I’m spending, let’s say, $500,000 on a particular level of process, it’s very hard to justify spending $200,000 automating something when I get the whole thing done for $500,000.
But then two of the five people leave, and now they can’t get it done. And if they don’t get it done, the fines are millions of dollars. So that’s when we have to remind them: It’s not just the FTE (full-time equivalent) play; it’s the risk management play. The reason you need to automate is because you can’t scale your existing staff.
What best practices do you see as companies move forward on automation?
One is that there should be a strategy around data organizationally, and tax should be a part of that strategy. Leaders across the organization, including at the staff level, should have some level of digital and data literacy to understand what’s possible.
And I think there needs to be a change in the environment where we apply Six Sigma, Toyota’s continuous improvement culture, to finance. We always have to ask ourselves: Is there a better way? If only for our own sanity, because much more data is available as we move into the digital age. Professional accountants absolutely need to be like budget data scientists to some extent, so that skill set needs to be there. But leadership also needs to give people the runway to experiment—within a controlled box, so they don’t create risk.
So the best practices include having a data strategy, making sure tax is a part of that, improving data literacy and creating a culture where people can try to improve efficiencies internally—because it costs too much to hire someone like me to fix every little thing.
What can board members ask management to ensure this is being addressed?
For starters, what are we doing with data? What is our enterprise-level data strategy? Risk management data should be a part of that.
Data is an important organizational asset, so there should be a plan around how we grow this asset. How are we managing this asset? How are we protecting this asset from risks? The C-suite doesn’t have to understand the nuts and bolts underneath it, but they should have a plan, and someone should be in charge of that plan.
Data is not much of an asset if it’s not being used. That’s kind of the next frontier: How will you use that data, and what for? Companies already know this. They’re already doing it in the front of the house. Customer analytics is not a new thing. But back-of-the-house business decisions—even investment decisions—are not always built on the best data. There are a lot of assumptions and a lot of thumb-in-the-air guessing. The technology is there to do better.
Companies now have employees working remotely from all over the world. Does having better access to this quality data enable them to do better modeling on, say, the tax implications for hiring?
It could—and there are other cost-saving opportunities there too. Global mobility is a constant theme. Whether I’m an engineer or other type of professional or executive, when I have to go work somewhere for a month, there are tax implications. If I’m a professional basketball player and play for New York but have a game in California, California wants this pound of flesh for that game, right? That’s just the reality. Employee self-reporting is really unreliable, and it’s unfair to employees to expect them to remember all the dates.
For one client, for example, we were doing this analysis on employee movement for state and local tax in one area and then global tax for employees moving through Europe, Africa and Asia, and we realized we had data that indicated where everybody was. So we started talking to their logistics team about planning projects, and what they didn’t realize was they had this one guy moving from China to Europe, Europe to China, China to Europe—but they already had the same skill set in Europe. So the cost of moving this guy back and forth was huge—when they had a resource sitting on the bench in Europe that they didn’t know about.
The insight was in the data, but nobody was going to look through a massive spreadsheet and pick that out. They had to see it visually because color, shape and size add context. When they saw it on a map, when they saw the little green dot representing that employee, and they hovered over it, and they saw what his skill set was, they were like, “Oh my, why are we wasting all this money sending this person back and forth?”
In all honesty, we were just trying to highlight where everybody was and their tax obligations. But when tax is part of your data set, you can make other business decisions around it.
We’re all hearing the scary warnings about AI and the potential risks of technology replacing humans. Why should directors not be concerned about using AI in tax?
So there’s generative artificial intelligence, like what ChatGPT is at the forefront. Then you have very purpose-built, specific artificial intelligence and machine learning, which is basically a statistical product. It’s saying, “There’s a high probability that this is the answer based on all the inputs you’ve given me.” The application of that to tax is pretty significant.
We are often asking tax people for an opinion, and the things that tax people are good at are their awareness of the issue, they’re aware of the court cases, they know about the laws, they know how to find the information, they know how to write a memo. But AI has been trained in everything. AI can be trained 10 times, then once trained, it can train another AI 10 times, then train another, and so on—there’s exponential learning it can do.
That doesn’t mean we cut humans out of it. It just means that the role of humans is no longer searching for the information, coordinating the information, amalgamating it and putting the opinion together. It means that the machine will take a first pass to say, “Here’s everything I found on the topic—do you like it? Yes? OK, here’s an amalgamation of all those thoughts, a summary of everything that’s happening there. Do you like it?” And then you, as a human, have to explain what it is. But you’re not talking about a generative intelligence that can usurp you from that role.
On the machine learning side, there’s significant value in simply saying, “Hey, there is a statistical probability that this is happening based on all these other things.” Businesses exist within jurisdictions; they operate within products, and they operate within municipalities. So there are all these little factors that affect tax consequences. If we can build machine learning models incorporating all of these factors, we can get better predictive results based on them. But there is no point when the AI will say, “Hey, wait a second—this is illegal. I’m calling the IRS.”
When finance transformation happens—and all finance transformation these days is about data and data strategy—we’re seeing more and more that tax leaders have a place at the table and have input right from the beginning. “How are we going to configure the enterprise resource planning (ERP) system? How are we going to configure our data warehouses? How are we going to make sure all of the systems that our ERP and data warehouses connect to actually think about tax so that when the tax people do their work, the data actually lives in the one source of the truth?”
That’s still the challenge, to get every organization there. But once you have this rich data source aligned to the tax needs, the tax leaders need to reach a higher level of creativity in saying: “OK, well, we can automate all this stuff, but is there more we can do with it?” That is where the next evolution is.
