–

Or use the below podcast player to listen:

Transcript (Note: There may be typos due to automated transcription errors.)

Randy Johnston  00:10

Welcome to the accounting Technology Lab. I’m Randy Johnston with my co host, Brian Tankersley, now today, we’re going to take you a slightly different path on artificial intelligence with the tax research tool of Blue J. Now they claim to be generative AI for tax experts, and the company has been around for quite a little while, founded in 2015 in Toronto. Now I think this is actually worth hearing. You might find it on their website if you’d like to have it again, but on their website, they say they were founded in 2015 blue J is a legal technology company on a mission to improve people’s lives by bringing absolute clarity to the law, and then our software empowers tax and legal professionals to perform exceptional analysis and deliver more insights faster. And I think those claims are true. Now they have operational offices in Toronto and in New York, and it turns out that this company has done some innovative things. A little later in the podcast, we’ll circle back and talk about Blue Js charter on algorithmic responsibility. But Brian, what would you like our listeners to know about the blue J tax research?

Brian F. Tankersley, CPA.CITP, CGMA  01:28

Well, their product is called Ask blue J, and it is a generative AI tool, and so you can ask it questions about tax law and other things like that, and then it creates these answers that are verifiable. Now, one of the concerns we have with using many of the generative AI tools is the the propensity of some of those tools, especially the early versions of them, to do what’s called hallucinate where, when they know something, they they do fine, but when they don’t know something, or the data on the training data is a little thin, it will try to guess and make things up, and it will make something up that’s kind of questionable. And so since this is focused on legal technology, the option of being wrong is not an option. And so, so So this, again, this, this tool in here actually ingests and will spit back to you multiple authoritative services. Again, it works relatively quickly. It will also do the first draft of a client memo or of a memo to the file on different things and and it is good at keeping the main thing, the main thing. You know, one of the hard things about research memos is that sometimes people get off on tangents about the exceptions to exceptions, and they they don’t keep the main thing, the main thing. And so this, this is really designed to do this, and it’s actually used by a large by a number of large firms. KPMG uses this Larson gross Anderson legal business and tax. And again, the the idea here is that we can take this and you can have it go through and look up and give you the authoritative resources on things, and it will give you kind of a summary of what those things say, so that you can, you can kind of get into something that is maybe on the edge of your your competency, now that you know, again, because, because you can’t know everything in the tax code, and you can’t know every decision and everything else, but the generative AI model can ingest All of that, and then it can provide those resources back to you when you’re looking for some of this information.

Randy Johnston  03:48

Yeah, so you know, it turns out the history here is a little bit good to know, because AI’s been around a long time. Brian, you’ve heard me say multiple times I wrote AI code in Lisp in 1975 the AI then was a lot different than AI now. And if we go back to the 2010 machine and how those evolved and so forth, there’s been a lot of changes in the last 15 years. In particular, the CEO of the company, Benjamin alari, was a Associate Dean of the Faculty of Law at the University of Toronto, and he was invited into to IBM Watson competition as a judge, and he started thinking about the use of AI in tax law and the ability to prove use machine learning to predict outcomes and recommendations. So they had their prototype running on the Watson platform in 2016 and of course, today, the IBM Watson now Watson X is also far more sophisticated, but they began selling the tax product commercially, and then expanded over into employment and HR and us. Law now just a privacy statement here, because we have talked about privacy and other sessions with you. The Canadian privacy laws are far more restrictive than US laws. And even though they’re at this point in time, there’s seven US privacy laws enforced, with 19 of them passed, there are Canadian National privacy laws as well as laws, province by province. And this product is compliant with the Canadian laws, the US laws, the European Union laws, the UK laws. I mean, they’re very cognizant about this compliance. So Brian, it probably goes back to this legal background, you can’t be wrong.

Brian F. Tankersley, CPA.CITP, CGMA  05:42

And I’ve, and I’ve actually read that the Canadian privacy law, and it seems like the Canadians are trying to almost one up the European Union and say we’re more private than you are. So anyway, it’s a it’s an interesting, you know, it’s an interesting statute. But the good news is, it, with it coming out of Canada, maybe we’re gonna, maybe we’ve got a better chance for things staying private, even though they’re using heavy generative AI here. Yeah, that’s, I think that’s a critical thing that I think is helpful. And again, the the compliance with both Canadian and European statutes, I think is a is a big win for this product.

Randy Johnston  06:25

Yeah. Now the thing that they did early on february 14 of 2020, they published a charter on algorithmic responsibility. This might come from their academic background. Brian, and I know you’ve done a lot of research, probably more than I have on some of these compliance issues globally and the the evolution of compliance and regulation around AI, but they made eight specific claims on their AI algorithmic position it is you can read these on Your website. You don’t have to listen to us talk about it so much, but it’s under there about us section, under their company area, because there they show not only the people involved, but this algorithmic positioning. And you know, these eight items are pretty interesting. So Brian, would you like to comment on those items?

Brian F. Tankersley, CPA.CITP, CGMA  07:20

Well, one of the one of the major things with AI is that it is only as good as how it’s trained, and so we’ve had a lot of misrecognition of of underrepresented groups and people of color, simply because, simply because so many of the facial recognition algorithms have been trained almost exclusively with people of Caucasian and and Asian descent, as opposed to being trained with, you know, and have not had as many from Europe, or, you know, Southern Europe or Asia, or southern or Southeast Asia, or again, some of the other, some of the other places in the world, and so as a result, they’re trying to build in things that actively avoid those biases. They also monitor the outcomes and provide feedback loops to try to make sure that, to try to guard against this hallucination that we’ve talked about that exists in chat, GPT and other places. They also want human oversight and control. Want to improve the human condition, so it’s not as much about replacing humans as it is about taking ministerial tasks away so that the humans can can focus on relationships and on on, on again, the the overall 30,000 foot view, as opposed to digging down into each little particular rat hole, they invest in data quality, so they try to, they try to make sure that their things are going forward well. They also insist on high standards of accuracy, and they have safe they safeguard they safeguard privacy and security. And they’ve had, they’ve had SOC twos for some time. So I think they’re, I think that’s an important piece that they bring to the table, that they are very, very serious about privacy. Because there are a lot of concerns when we start talking about the privacy of tax data. You know, I was actually, was actually just surfing around YouTube, and I saw a presentation by by Senator John Kennedy out of Louisiana, where he was talking about the talking about the federal IRS leaks, and he talked about his experience As the revenue commissioner in Louisiana. And I think, I think he’s absolutely right that people, people are very, very particular about maintaining their privacy of their information. And this privacy thing, especially when it comes to tax and personal information, is something that cannot be the. Value. This cannot be underestimated simply because it’s a, you know, it’s, it’s one of those table stakes, things that a lot of people don’t realize. And I’m very concerned that with I’m very concerned that with some of the products in the marketplace, we may have made some Fauci and comfort Faustian bargains where we now. We have now let things out of the, you know, we’ve let things out of the barn, and now we’ve got to figure out how to deal with it,

Randy Johnston  10:31

yeah. So, you know, trying to maybe get to the the some of the other things that I want to make sure our listeners knew today, Brian, you know, the eight claims that blue J makes in their AI statement again 2024 years ago, avoid creating and reinforcing inappropriate bias, continuously monitor outcomes, provide transparent and data outputs allow for humans oversight and control, improve the human condition, invest in data Quality, insist on high standards of accuracy and safeguard privacy and security. You covered all those, I believe, but I wanted to call them out like that, because I believe that those largely came from something you and I have talked about for six years, a little more than that, actually, because on May 16, 2018 the Toronto declaration came out, and it was about bias in machine learning. And you and I have presented on machine learning and AI for years, but the Toronto declaration and consequent rulings in other international meetings were quite good. Now, the reason I’m very sensitive to this bias in machine learning, or bias in AI, our listeners will certainly get some amusement over the fact that there are six software tools that sort applicant data. About 20 million applications of employees are processed per month in the Fortune 500 with these AI based tools. I have run my data through all six of these tools, and I’m unqualified for an IT position. Okay, so I just thought you might find that a little interesting, because the bias in the applicant processing data throws me out, but let’s flip it around for those of you want to be a lot more practical. If you’re looking for leverage from AI and you are a tax specialist and you would like to accelerate your tax guidance and be able to base it in fact, we think blue J may be one of your best options out there. We have done other accounting technology podcasts on the new iterations in Thomson, Reuters checkpoint product, the checkpoint edge and its CO counsel. We’ve done that on Walters clewer and so forth, but this is a nifty, doggone little product that’s got a lot of Head Start, and I think could be quite useful. So Brian, other key things you think our listeners would benefit from knowing?

Brian F. Tankersley, CPA.CITP, CGMA  13:12

Well, first off, I think you need to consider that there are also products from both Wolters Kluwer and from Thomson Reuters that integrate with those suites of applications. So if you’re using checkpoint or tax research network or some or, you know, again, accounting research manager, just know that that those tools are out there and and again you can again, this is another product that that I think fits a similar need that will will help, help solve this problem we have of how do we get the right answer faster and get it communicated faster so that we can do we’d be better at our job and be right all the time. Yeah, so,

Randy Johnston  13:57

Brian, you don’t want to just make it up that the tankers leaves to owe no tax. You think that’s a bad decision?

Brian F. Tankersley, CPA.CITP, CGMA  14:03

Yeah, that’s not going to go very well. That’s gonna that’s gonna end in, that’s gonna in a trip to the crossbar hotel life here,

Randy Johnston  14:12

I’m afraid so well, we appreciate all of you listeners with us again today. We’ll talk to you again soon in and up their technology accounting lab. Good day.

Brian F. Tankersley, CPA.CITP, CGMA  14:26 Thank you for sharing your time with us. We’ll be back next Saturday with a new episode of the technology lab from CPA practice advisor. Have a great week.

= END =