In this episode of The Accounting Technology Lab, Randy Johnston and Brian Tankersley are joined by Hrishikesh “Rishi” Pippadipally, partner and CIO at Wiss, to explore how accounting firms should evaluate and select AI partners. The Accounting Tech Lab is an ongoing series that explores the intersection of public accounting and technology.
View the video below:
Or listen to the podcast using the below player:
Transcript (Note: There may be typos due to automated transcription errors.)
SPEAKERS
Speaker 1, Brian F. Tankersley, CPA.CITP, CGMA, Randy Johnston, Hrishikesh Pippadipally (speaker 1).
Brian F. Tankersley, CPA.CITP, CGMA 00:00
Welcome to the accounting Technology Lab. Brought to you by CPA practice advisor, with your host, Randy Johnston and Brian Tankersley,
Randy Johnston 00:09
good day. Welcome to the accounting Technology Lab. I’m Randy Johnston with my co host, Brian Tankersley, we’re pleased to have a guest today. Rishi from wisson company is with us, and we’re going to talk about lots of different AI applications that the firm has used, and some products that are in play, and so forth. So I think you’ll find today’s conversation to be pretty insightful. So Rishi, if I could ask you to introduce yourself and give us a little background that would be appreciated,
Speaker 1 00:37
Sure, absolutely. And thanks for having me little bit about this before getting into my background again. This is tech enabled Accounting and Advisory form. We strongly believe the future of profession is being rewritten by technology, and we’re not waiting for the disruption. We are helping build it. So that’s about vis and me. I’m the partner and CIO like, yeah, my focus is technology, which means, increasingly, AI, my role blends between strategy and execution, and I do actually spend tremendous amount of time evaluating AI vendors, running the pilots, making sure what fits for the firm. Like, how can we leverage technology all the time to be more proactive.
Randy Johnston 01:21
Yeah, and you know, I know that many of the CIOs I know around the country have been pursuing these technologies, and I think that my understanding of the way you look at this, you are quite cognizant of accounting regulatory and client data constraints. And I’m going to set the question up just a little bit further, because Brian and I may be the most client data protective of any public presenters or podcasters out there, because we recognize it’s actually part of CPA licensure, and we also recognize that a number of the vendors and suppliers are not protecting the data appropriately, and we’ve talked about that in other labs. So what would you want us to know about industry context? Because, again, when I think about it from a CPA firm perspective, or I think about it from an industry accounting perspective, you know, we tend to live in that world and have for decades.
Speaker 1 02:22
Yeah, and definitely, I think data privacy, data confidentiality, is pretty much top of our mind too. Like, when we are serving the clients, we want to be very cognizant about, like, what actually being used by this vendors. There’s a processors, like, really careful about that, again, like it’s while evaluating a vendor, right? It’s always important to know, like, what they’re doing with data. Like, what, what kind of like retention rules do they have? Like, how are they using it? Anonymizing stuff like benchmarking, like really need to know, and ability to export out the data or like, their ability to like, especially in the AI world, right? Learning. So what exactly they are learning? How can we push when we no longer want to continue the relationship, right? So these are all really important and industry specific rules, like, again, in the AI world, like we see now, hundreds of companies trying to pitch and saying that they’re going to solve certain problems, right? So, like, how do we even believe that? Like, when coming to llms and this, foundational models are good with general knowledge and stuff like that, but there’s lot of I cannot think it’s full of nuances. This compliance controls industry specific rules like, and not the least the client level customizations, right? There’s so much nuances that you need to evaluate and make sure, like, how these companies gonna handle these or like the startup companies that they’re saying that they can solve the problem. And it’s flashy demos, but when it comes to real world, it’s really a messy one, and really need to make sure data privacy and data security is top of the mind, and the quality and accuracy of the results that they are producing is consistent with the, again, AICPA guidelines or the controls that’s needed, right? So that’s why the context is really important. Like, I’ll just give a simple example, right? Like, if you just look at, like, plain foundational models, like, anything, I don’t want to name specific, but like, just ask, like, what would my depreciation? Accelerator depreciation gonna look like? It’s a 5050, chance that it’s going to consider all the new tax code, like without considering the tax code, then let’s say, if it misses, I say, Hey, can you consider oppb law and give me my accelerator description? Then it revise that is not thinking like an accountant. That’s basic for an accountant, right? So like that industry. Context, like, very specific to the tax code, or, like, if it’s an audit guidelines, whatever it is, right? So there’s a layer of training that you need, and I call it like, again, like, the industry also calls it more of a vertical layer, right? So there’s few players who can actually kind of help in this, like, beyond the big foundational models, like, again, co counsel is one we’re working very closely with. Basis for many years, they’ve been positioning themselves as, like, accounting AI. So that’s really important, right? Like, when the AI quality is so much not comparable for an obvious accountant, that option rate not going to be good. So like, again, like to sum it up, I think industry context is really critical for an option, right? Yeah.
Randy Johnston 05:50
And you know, when we’ve talked about the frontier models, and I actually very specifically rank their security, I have open AI and chat GPT far down the food chain for those reasons, and copilot further up the food chain because of the similar reasons. And the other context here, though, is the accountant centric. So if you think about tax research, you know, in effect, the Blue Jays and the tax gpts and the accordances of the world, where they actually are thinking much more like an accountant, or maybe even more true to tax law. You know where the BNA AI is living, or Thompson Reuters co counsel, checkpoint co counsel, and cch is living. So I am tracking with you on that as well. Now, when it comes to data governance, I know that we’re going to see more data governance content from our own k2 group, because we’ve taught data governance in the past, and really wanted to step up the data governance in AI again. And part of the reason for that is what vendors do with data in their own models is one thing. What they do when they pass that data to sub processors is another thing. And you know, with the open AI Intuit relationship announced in the last in the last week or so, it is pretty clear to me that we have to be cognizant about that. So just help us understand how you are currently evaluating vendors when it comes to those things.
Speaker 1 07:29
Yeah, I think, like, again this, some of the basics still stay the same in terms of like, making sure SOC two, making sure they have all the penetration testing, like, in terms of how robust their systems are, right? But again, like, now coming to more specific to AI Right? Like, how are you supporting the multi client model? Like, how your data is stored? Like, how do you separate them? Like, what are the walls between that, right, that generic data now, even the AI learnings, every time you do something on your client data, and you’re correcting the patent to saying that this is how you kind of do your client preferences, it’s learning something. It’s just like an accountant would have learned in your office, right? Like, how do you get trained? So that learnings are going to another database, maybe it’s a vector database, whatever they’re trying to use the right model for, right? So those are the learnings. Like, again, like, how are you segregating that? Like, what are the rules of the road for that, right? So really understanding the moral again, like, it’s so premature at this point in time. Like, unfortunately, we don’t have any kind of a federal guidelines and stuff like that. It’s like innovation versus like compliance, right? So, but still, we need to make sure, like, we understand what they are doing, and make sure the vendors are transparent and letting us know, like, how they are using all the data that you have, right? And like, define the rules like, how are you purging it? Like, what? How long are you going to retain it? Because, not necessarily the vendors that you might try now might exist in six months, one year, or you might be happy or not happy, you need to always make sure your client data is protected and your IP is protected.
Randy Johnston 09:16
So, yeah, understood. And today, have you built any agentic AI across WIS, yeah,
Speaker 1 09:23
we, we’ve been partnering very closely with basis on the agentic AI side. Like, I mean, we do realize that, like past couple years has been conversational. Ai, now it’s agentic AI. We do little experimentation in house with different Microsoft world, where we live in, but majorly bank on the partners to build all these agent tech workflows. We have multiple vendors that we are working on. Some of them are more like generic purpose accounting use cases that gonna solve, which is more, I would say, basis. But there are certain the. Vendors that we work is very specific and needs to like a particular piece of the puzzle, right? So that’s how we kind of solving it. But out to say, yes, it’s gonna work. But the lot of edge cases, the vendor that acknowledges and understands that, yes, there are nuances what works, what doesn’t work is where, like, really, you can bank on saying that, yeah. I mean, these people are not just a technology professionals. They understand the industry. They understand the nuances. So, like, we kind of weigh in and try to see, like, who is the long term partner that we need to go with. But yes, I would say, like, 2026, especially coming up, going to be a cycle of maturing the agentic workflows. I see like two step tasks, three step tasks being executed. But when it comes to orchestrating a bigger end to end task, it’s still not there. That’s why we are partnering and CO building with all these some of Yeah,
Randy Johnston 10:58
and for our listeners purposes, you know, one of the things that I thought was interesting is your firm has been running with the basis product for about three years. And, of course, multiple other basis partners are ones that I’ve referred so when I think about the way that things work, and you’ve probably had a good working relationship with Matthew and Mitch just because of your position there, or the rillet product that Brian and I covered in a separate accounting Technology Lab podcast here recently. Or perhaps, like probably a newer vendor to you from the sounds of it, the tabs with Molly shields and that team, you know, you’re used to dealing with vendors who see you, I think is an attractive target. And so how do you vet what they say they do in their privacy statements and in their user license agreements, in their EULAs, as well as what happens in the sub processors. Or, I assume you’re digging that deep with most of these vendors to affirm that the client data for Wisp is protected tightly.
Speaker 1 12:13
Yeah, I’m glad that you mentioned those names again, like just taking a step back, right? Like the way, these are again complementary vendors, not necessarily acting on the same side. So we have developed an internal process that we call this lab. Literally, it’s a funnel that kind of takes in all these vendors. Like, I would say last year, at least 100 to 200 vendors we kind of evaluated and like, we kind of bucket them in the sense of, like, is this product, like roulette, for example, it’s more they play in the ERP and native ERP space versus basis is more accounting, automating accounting workflows and generate purpose, right? And there are some very specific niche, I would say, like k1 kind of automation technologies, right? Like, you need much more deeper expertise to build it, and we can’t count on this. So there are different layers of how we see these vendors playing right, and how we develop the relationships, and how do we see them growing strategically, right? So that’s how kind of, like, we even tier the partners. Like, we have a different tiers on, like, what’s our involvement with these vendors, because sometimes it’s heavy co building, where we have weekly calls trying to do internal calls, feedback calls. Then we work with the vendors on like a filtered version of that. There are like monthly calls. We participate in like development cohorts, which are run by them. So there’s different levels of partnerships, and we tear that so, like we have a mature framework that kind of naturally evolved, like we didn’t plan for that, but the way the things were going on, like we put them in the lab, we’ll say, put them in the lab, and we have a kind of a strong internal group too, because we cannot rely everything On the vendors, right? So we have the tiers internally, like, especially the business transformation team, who help us bridge and evaluate this at multiple levels, like at the domain level, trying to see the fit there, the technology level and then the compliance, like the part that you have been asking so, like, again, like, it’s the probing is what we do and try to see, like, everything that we have has been checked off or not, right? Again, like, we believe in the that, like, whatever transparency that they have, like, again, there’s a layer of how much we could go like, but we do ask them to list all the sub processor like, what kind of models they are using. Some cases they use proprietary models. Want them to list it out, at least we are aware of it right? So then we can make a call like, how what’s our risk appetite on trying to go like, does this fall in the acceptable. It or not, because there’s no proper guidelines over there, but we try to make sure we are as conservative as especially when it comes to data and compliance.
Randy Johnston 15:08
Yeah, makes good sense to me, so I’m going to restate what I think I heard you make sure that I’ve said it right. But you know, over the recent past, for your firm, you’ve evaluated hundreds of products, AI products through your funnels. And, you know, I always say I learned something new every time I speak with somebody, but you are tiering them, you know, primary, secondary, tertiary, possibly tiering the solutions, and then, you know, vetting them that way. And that’s actually a pretty good framework for people to consider. Now, Brian and I and other accounting technology labs have talked about AI governance frameworks, and we’ve seen several vendors publishing these frameworks this year. Field Guide, I think, has one of the best published frameworks that I’ve seen for a while. But, you know, I want to ask some questions about products, but I think Brian, you’ve maybe got a comment or question.
Brian F. Tankersley, CPA.CITP, CGMA 16:00
I want to get back to brass tacks in practice. Okay? Because you led with WIS being an advisory firm, okay? And so one of the things that you know, one of the problems historically, with advisory is you only have so many partners in the firm, and all the juniors, or as I refer to them, the kids, want to you know that are in there early in their career, they want to jump immediately into advisory. And the challenge is, you have to have an industry competency and experience and understanding of the context in an industry to be well, you know, we’ve learned that over the years, we’ve been the podcast, that context is one of the things that AI does bring to to to different to to technology. So the question I have for you is, give me three or four concrete examples of how AI has helped you and WIS make think these things more leverageable, so there are more tasks that can be downloaded, that can be delegated to juniors, and so that the process is and the practice and the advisory is more scalable,
Speaker 1 17:04
so you’re talking very specific to advice. So I’ll give you some examples here. I think I have you on top of my mind, definitely. So, like, account coding, right? Like, yes, QBO has it. Like, few other ERPs do have it. Like, really, like, especially in the cash world and, like, outsource accounting world, like, you don’t need a junior accountant sitting trying to code those transactions, right? So you feed in a two years ledger, you try to give again. Like, there is issue over here. Sometimes I’ll be more practical in this, because I’ve seen those cases like you feel that two year learning of your journal will make AI understand it, but are cases where you might have done differently for the same vendor and classifying it differently? There could be reasons, right? So that’s where the human in the loop comes. And we believe a lot on that. But like all that is like more than 94% is kind of automated, like nobody spends time more than reviewing it and making sure they approve the transaction. So that, that’s a classic example. And AI looking at, kind of reconciling to, let’s say, payroll data and journal entries and payroll actually proposed the AI actually proposing the journal entries for payroll to make sure, right? So all these things like that, you expect, traditionally, staff accountants or the seniors go through that like now, they don’t have to do it. They just review it, more from a preparer point of view to a reviewer, which gives them a lot of time for on the advisory side and higher level times.
Brian F. Tankersley, CPA.CITP, CGMA 18:44
So, but how are you is there? Do you have a AI doing anything to capture specific knowledge at an industry level of operational type issues that you’re using to advise clients
Speaker 1 18:57
on the advice and on the clients? Is like, again, like we pretty much have gone deeper into the basis AI, and it’s been like two years plus in the maturity cycle, I would say, like all of our kind of cast practice goes through that like it has a full view. They train on the accounting AI, which is where I was like, going from foundational model frontier models to like more specific to accounting. So we rely on them to train all the accounting specific and industry specific. What we give is a client knowledge, right? That could be from the contextual knowledge of your prior transactions industry, or that could be a like a corrections from our staff knowing the patterns of very specific what your client customizations are. So it’s a mix of team effort. Yeah.
Randy Johnston 19:47
So Rishi, on that point, just enumerate for us. I’m going to ask you to do this with basis. I’m going to ask you to do with real it. But you just name 567, things that you say basis is doing this well for Wis.
Speaker 1 19:59
You. A basis is doing well for the again, like all these players, kind of like competing people have an overlap, like, now it’s kind of more converging to lot more of overlap. But I would see say that, like, the consistency in the quality is high. And again, I would say, like, there are certain areas of research, which is, let’s say, co counsel, like, they have a lot of proprietary material, so the guidance that we give internally to we use a lot of AI, like, we’re not zeroing on one or two. Like, nothing can match co counsel when it comes to research, because it’s Thompson proprietary, right? Like, it’s very clear where to use AI where not to use what type of AI, right? So what we do internally is, like, clear guidance, right? And going back to your basis question, right? So it’s good in, like, these kind of quick learnings and codings, and they’re probably the best partner that I’ve seen in terms of taking the feedback, accommodating and coming back in the fixed cycle. Like some of them are, like, we give the feedback today, they fix it tomorrow, like it’s super fast, where our users do feel that okay, they’re listening and they’re fixing and I can trust them. So that’s the key, important part of like, an option that we try to see. And going back to more on the AI side, I do feel they also better at like dealing with excels, like when these different players interacting with Excel, and the quality of output that you get in Excel is not so good. Like they had a fixed data extraction, especially the bank statements and all like, a year ago, or even more, with high consistency, right? So that’s another thing. Like data extraction has been like, almost taken care. They’ve been forefront in that.
Randy Johnston 21:54
And by the way, I want to stay on the path that we’re on, because what you just claimed about rapid modification cycle. That’s absolutely been true with basis and clearly the extraction and classification. That’s another key area. You know, I usually do call out things like rapid reconciliation, the interfaces with QBO and sage intact, the reporting capabilities in the platform. You know, there’s just so many things that it does from a productivity perspective, and I think it’s because of the AI. So are there other things that your team have been reporting with basis
Speaker 1 22:31
in terms of the product development, or is it more like you mean, like the AI efficiency
Randy Johnston 22:37
science, the AI efficiencies, I think, is the nature there of the question, yeah.
Hrishikesh Pippadipally 22:41
Like, one other good thing again, the basis, or any partner that I would say, is, like, you need to partner with them to even come up with the most impactful use cases, right? Like, often we think something gonna work, but the quality of and output playing through an agentic workflow may not be the same. So like, the and sometimes it’s reverse, they could explain us, like, where actually their AI works with it, it doesn’t, and that’s an open feedback that always appreciate from them, and probably a quality that you need to look for in any partner or vendor partner that you’re looking for, knowing what you’re not good at, because you’re not chasing the wrong thing.
Randy Johnston 23:24
Yeah, that makes sense. And you know, I’m going to do the same type of issue with real it, because you’ve got your I’ll call it professional AI ledgers out there, the digits and the puzzles and so forth. But obviously you picked one of the ERP ledgers real it campfire being two examples in in that area. So what has RIT done from a productivity AI for your firm, and what are some of the key features that you’re leveraging there?
Speaker 1 23:55
Yeah, like we didn’t talk about AI, but, like most of it is also in how you’re integrating with lot of other systems in the like kind of ecosystem, making sure you have the data available, and like, not chasing for the data, but you already have talking to these multiple and so they’re probably one of the top people who are integrating with as many people as they can, so that you don’t have to put those payroll journal entries or something else reconcile like, you don’t have to do that like, because they can talk and they can match and surface you like, what are the breaks? Right? So I think that’s where I believe what the foundation is, and layer in AI on top of it, to give those kind of like analysis and judgment for like someone, for a human to review, is where they’re Excel they’re excelling at.
Randy Johnston 24:46
That makes great sense to me. And really this was probably getting to the third directional point that I wanted to make sure we discussed. Because, in effect, with your use of basis, your use of tabs and so forth. Worth, in effect, you’re really working with what you might consider to be true, end to end partners there. And so you’ve talked a little bit about how you’ve discovered an adoption, but talk to us a bit about change management and team reskilling.
Speaker 1 25:14
I mean, that’s a like, biggest problem, I would say, like, it’s consistently across any technology or platform that you bring in, right? So again, we do put a lot of emphasis on training. Like, forget about like, tool specific training, like generic, like, upskilling people in terms of, like, what does aI mean? Like, what is conversational AI what? Why do we have different models, like, starting from the basics? Like, I think that’s really important for them to understand, and what we try to incorporate, and which we did, like, last year, early last year, and we are doing a retraining of that is like, like, if you give generic examples, they will never understand it. Like, we can try to curate it like tax specific, like, give it the tax examples, like we even looked at, like, multiple outside third parties, if they can do it, but none existed a couple of years ago. So we build our own content and which is more geared towards, like, what they can relate to, and see how that played out. So it’s really important for them to understand. And like, what is a prompt like? What is that context like? Why is it important to it? Like, with the trainings and all like, always feel that, like, I have a third and third rule, like, 1/3 of the people are, like, motivated. You don’t have to give them much information. They they’re on board and they’re off the ground. The other thought I see is like, you need training to them, like this kind of training, like, explain everything, then they understand this. Other thought is like, it goes in multiple iteration, and you need to keep educating them again and again. And it takes time for that. It’s just the nature of different segments of the people that we see at the phone, right? So we make sure we train as much as possible and a clear guidance, because I often see that it’s overwhelmingly technology now, right? Like, why do we have so many tools? Like, what is it? Too? Like lot of people are with confusion, so we give a clear guidance on when to use what and what are the strengths. So like, training is, like, a big emphasis on at first, like, especially upskilling people, because, like, without that, like, there’s going to be not much of adoption, right? So that’s on the people side, upskilling, coming to the partners, right? Like, again, like you already said that, right? That, like, the true partner is like someone who’s there with your internet like, you should feel that they’re an extension of your team. Like, especially with this new age AI technology, I don’t see any mature player out there, like someone who could help diagnose a problem, design a solution, because there’s a lot of feedback. Good feedback comes from accountants on the ground, right? Like, running the pilots. Like, that’s very important. Like, we don’t want to go any big bang, right? Train your staff, like, in the product. Like, and when you say, train your staff in that product. It’s process re engineering, like, that’s where we put a lot of emphasis, saying that, hey, this is the old way of doing this is the new way of doing this, too, with the examples, right? So, I mean, all that cycle, we partner with that firm, and we have a strong internal team. I mean, that’s something I would recommend anyone, everyone, to have it like we cannot put everything on the vendor. You need that team of process architect who can re engineer the process with the new technology, right, and kind of be the change management change champions.
Randy Johnston 28:55
Yeah, great, makes great sense. Well, you’ve got a ton of experience in this area and managing these technologies is there as a kind of a final parting thought, any particular piece of advice that you’d give to firms that are trying to look for AI partners and how to avoid common pitfalls? What would you suggest is your key learning. I mean, like,
Speaker 1 29:25
on the SIR under the surface, right? Make sure the transparency is there. They have the audit trails, like how they’re coming to the conclusions and explanation references, like, even if AI is recommending something, like, where did it draw the conclusions from? Right? Like, those are, like, more, like, detailed under the hood kind of thing. Like, again, I don’t want to go deeper into that, but, like, on the surface, right? Like, don’t just buy the tools, like, buy the partners. Like, look at the depth of their knowledge, not just technical, the domain expertise. Just don’t buy the hype and the demos. Like, I. And make sure you go through the pilots. Lot of people offer free pilots, sometimes they are paid pilots, but yeah, whatever, it makes sense for you, based on the situation, based on the conviction, like, definitely go with the pilot.
Randy Johnston 30:13
Okay, makes great sense. So Brian, final questions or thoughts.
Brian F. Tankersley, CPA.CITP, CGMA 30:19
You know, it’s just very nice to hear somebody else talking about worrying about privacy and other things in accounting firms. And I’m very glad to hear that WIS is again trying to take a leadership position in this area, and so it’s very nice to hear this no other questions. Just very impressed and looking forward to seeing what you and the rest of the team come up with it with that we can learn more from. You know, with the smaller firms that we talk to a lot.
Randy Johnston 30:44
Well, super well. Christian, thank you very much for your time today and your expertise. It’s greatly appreciated, and we hope to have you back in the future. And for our listeners and regular followers, we’ll talk to you again very soon on another accounting Technology Lab.
31:03
Good day. Thanks, Jenny. Thanks Brian for having me. Thank you.
Brian F. Tankersley, CPA.CITP, CGMA 31:08
Welcome to the accounting Technology Lab sponsored by CPA practice advisor, with your hosts, Randy Johnston and Brian Tankersley.
= END =
Thanks for reading CPA Practice Advisor!
Subscribe Already registered? Log In
Need more information? Read the FAQs