Technology experts in the accounting industry have plenty to say about ChatGPT, staffing issues and a push toward more automation, and how organizations can decide which projects are done in house or through outsourcing.

This is part one of a two-part conversation with:

• Amanda Wilkie of Boomer Consulting;
• Wesley Hartman of Kirsch Kohn & Bridge LLP; and
• Donny Shimamoto, CPA/CITP, CGMA, of IntrapriseTechKnowlogies.

The conversation also was turned into a JofA article. The second part will be published in podcast form on May 18 on the JofA's podcast page.

Resources

• Previous podcast conversation with Hartman about RPA.
• The 2020 podcast episode that featured tech roundtable participants Wilkie and Shimamoto.

What you'll learn from this episode:

• An overview by host Jeff Drew of the guests' background.
• Wilkie's warning about ChatGPT and why she compared it with Wikipedia.
• Hartman's explanation of a new meeting feature in Microsoft Teams.
• What Shimamoto said about balancing in-house and outsourced work.

Play the episode below or read the edited transcript:

— To comment on this episode or to suggest an idea for another episode, contact Neil Amato at Neil.Amato@aicpa-cima.com.

Transcript

Neil Amato: Welcome to a special joint episode of the Small Firm Philosophy podcast and the Journal of Accountancy podcast. This is Neil Amato of the JofA, and this two-part episode marks the return of the JofA's accounting technology roundtable, a group that first met 12 years ago with the goal to pass on tech-related knowledge for accountants to put to good use. Three tech experts in the accounting space are featured in this interview. Here to introduce them and conduct the interview is Jeff Drew, a manager with the AICPA's Private Companies Practice Section.

Jeff Drew: Our guests today bring expertise and experience in all things accounting technology. Amanda Wilkie of Boomer Consulting is a former chief information officer for a top 30 accounting firm. She is also a prolific writer and speaker on emerging technologies, especially in the accounting space. Wesley Hartman is director of technology for Kirsch Kohn & Bridge LLP and also is founder of robotic process automation developer, Automata Practice Development. Last but far from least, Donny Shimamoto is a CPA/CITP and CGMA, who is the founder and managing director of IntrapriseTechKnowlogies, which is an accounting firm that exclusively provides technology consulting and services to small and midsize clients. He also is the founder of the Center for Accounting Transformation.

Thank you all for joining us today. There's been a lot of noise around artificial intelligence and in the artificial intelligence space in recent weeks. Microsoft announced a multiyear, multibillion-dollar investment in OpenAI, the creator of the chatbot ChatGPT. Amanda, since you're one of our emerging technologies experts, can you tell us what ChatGPT is, what it can do, and is this something that accountants should be using now?

Amanda Wilkie: Well, Jeff, ChatGPT, it's more than just a tongue twister. I have a hard time saying it, so I'm probably going to stumble over it a couple of times through this conversation. But we're talking about it really because it is the tech du jour, if you will. Everyone's talking about ChatGPT. As you mentioned, it's a chatbot that was developed by a company called OpenAI. It was just launched in the fall of 2022 and it quickly became the fastest-growing consumer app to date.

It is spreading like wildfire, so we need to definitely be talking about it, but it makes web searching more conversational, so you can build off of the question of the search that you just performed. But it can do a lot more than that. It can write computer code or debug code. It can play games. The possibilities are almost endless. It's been writing college-level research papers. It can pass professional exams, which is, again, something that's really getting attention out there. As you mentioned, Microsoft has now licensed ChatGPT to be integrated with its Bing browser.

Right now that access is limited. People are playing around with it. But one thing that did was it forced Google to speed up the release of a similar technology that it's been working on called Bard. That's what they're calling the Google version of this. Now, we're already seeing accounting and financial professionals use ChatGPT. There was a great article in the JofA recently and the author actually had ChatGPT write an Excel macro and it appeared to do a pretty good job with it.

We're hearing firms consider using it as a tool to help train or advise some of the newer staff on how to build correspondence with clients because it can generate or create content. I think one thing that is interesting about this technology right now is In overall, we're hearing firms really look for ways that it can use this technology opposed to look for reasons not to use it. Back in the day when people started bringing in the first iPhones and they wanted to connect it to the firm's email system, the firms were really concerned about the risk there. Now firms are more open to using some of this technology that is available. Of course, there are some concerns. There are plenty of drawbacks to ChatGPT. For one, it's very, very new technology.

It is not mature at all. It has a limited training dataset, if you will. You have to consider that training set and that some of the information out there could actually be limited. Also, you have to think about when you're asking it to generate content, you've got to be careful. Say if you ask it to generate an article about a tax topic or the latest in how companies are using internal controls and internal control audits to support ESG initiatives, ChatGPT is getting that content from other sources that are already publicly available out there. It could actually be plagiarizing some of that content. You're going want to be careful with that.

Then sometimes when you're asking questions to ChatGPT, you're actually adding to that training set. You have to, if you're asking about specific clients, you need to be careful because we're not really sure what the privacy implications are quite yet. Then the last thing I'll throw out there as far as the drawbacks or some of the concerns is sometimes it can just be wrong. There's been reports of people asking ChatGPT for things like song lyrics. They just got flat out a wrong answer. I would say this technology is something to really keep an eye on, but I wouldn't leverage it too much right now. I think about when I was doing my MBA about 15 years ago, I couldn't use Wikipedia as a source, but I could go to Wikipedia, scroll to the bottom, and get all of the references there. I could use it as a tool. That's what I would say about ChatGPT right now, is you can use it as a tool, but don't consider anything that you get out of it as really being authoritative.

Drew: Yes. That's great advice. Let's turn to arguably the most pressing issue facing the profession, which is staffing. Accounting firms, finance departments that are suffering from crippling capacity issues. There's just not enough accountants out there right now to get all the work done. AI is loosely related to automation and probably will greatly affect automation at some point. But right now let's talk about what automation will help and what tools are available at this moment. Wesley, you are an expert in this area.

Wesley Hartman: Yes. There's a couple of things. Actually, and to piggyback off the AI thing into automation, so Microsoft, for example, announced along with their pushing it and they actually announced they're adding a function to Teams called Intelligent Recap, which is going to be powered by AI, the OpenAI product with ChatGPT, but it will actually go through your transcript and create a summary of the meeting, and a task list, and things like that. Like Amanda says, it's a tool, so it's probably not going to get it right every time, but there's many times where no one's taking notes during a meeting. Someone has to go through and maybe watch the meeting again and make a task list. But just piggying back off of that, continuing with that is just work. Getting the work done and automation and what tools are available. I do think a lot of automation is available to accounting firms of all different sizes.

I definitely do a lot in the RPA side, but I really want to broaden it to just any automation because there are many times where existing technology stacks have automation tools built-in already. Whether it's just sending an email or integrating one software application with another. Really I would say it's important for accounting firms to look at their tech stack and just see what they have and just run a search or even just ask the vendor, who do you integrate with? Then from there, we want to build more tools into that. We have off-the-shelf automation tools like Zapier. Zapier is a good one. They integrate with a lot of products using API calls, but they really give it a nice building stacking, a good interface, really like a flowchart. It's really easy to adopt.

Then if you want to get into the deep stuff, then we're talking about RPA where you are really building custom automations and where you're building an automation that will do a task for you. All of this together I think really can help alleviate the capacity issues in your firm. I'll just use an example. I was just working on building a bot recently for processing single-member LLCs in California where we populate a spreadsheet and then it takes that information and then populates that into tax returns. While the thought processes are well, I'm still inputting into a spreadsheet as opposed to the software, but all those extra clicks of going into a return, finding the correct location to input that information. All that stuff really adds up. If you can automate those clicks essentially and really just input the data into your data source, or even better, export it from another source and then suck it into your existing software. That is all automation. A lot of times automation, using these tools, It's going to be adding up to a lot of little things. It's not just a single, big automation project that will solve a lot of problems. Both can be valid.

Drew: That's a good point, and I will put in the show notes a link to a podcast discussion you and I had, Wesley, where you go into a little bit more detail on how the RPAs work, what time can be saved. I will put a link to that in the show notes. And we may want to jump back and touch on some specific tool names for the audience like Power Automate with Microsoft. But I want to get Donny into the conversation first. Donny, I will ask you a little higher-level question. Is RPA ready to provide enough help automating these repetitive processes that firms can free up their accountants for higher-level work? Or at least have them not have to work 60-, 80-, 100-hour weeks so often?

Donny Shimamoto: That's a great question. It's interesting to hear you describe this whole crippling capacity constraints. I think if I take a step back and we look at automation as a whole, because I actually consider the AI part or the AI tools part of that automation spectrum. Wes did a really good job of starting off in saying, be careful of being a solution looking for a problem in that don't immediately jump to RP and think, 'Oh well, this is going to solve all my issues.' You need to start first back and look at, 'Can I do something within the software itself.'

Is there macro other type of capabilities there? Then if that's not there, then, now let's look at RPA. I do think it's actually mature in fact, actually the year of the pandemic, my prediction for that year was that we were going to actually see the highest level of adoption of RPA than we ever had seen before. Because I thought at that point, I was like, this is ready for prime time well pandemic came. Of course everything's shifted and we have to look at remote work and all the other things.

But I think we're back at that point this year where we can say, this is the tool that we should be seeing and we're seeing a ton of adoption in RPA in our practice and across different areas, tax audit, and in the finance departments. Wes also, I think set up really well, too, in saying this is really about automating repetitive processes. Because the other things I always tell people to look at with this is one thing. First is, is it actually going to benefit you? Like, how long does it actually take the person? If it only takes them a few minutes or an hour, you have to look at that and contrast that on what is it going to cost to actually implement the RPA or build the RPA, or even potentially maintain the RPA if something that you're automating actually changes quite a bit.

Then the other thing I often encourage people to look at the bigger picture is what risks does this actually incur? Amanda actually brought that in with the ChatGPT and the privacy concerns. Are you actually automating something or are you storing something, for example, the password for a software in the RPA that may actually be at risk for a data breach. I know one of the RPA vendors in their software, they store the passwords up in the Cloud. I saw some of them selling to accounting firms where they were building a solution. One of the consultants was building a solution. They were selling the running of the solution to the firm, but everything was all in one single instance that the vendor controlled. Well, in a case like that where it's not segmented by accounting firm, all the accounting firms that this vendor or consultant was serving was all intermixed. I was like, that's scary from a cybersecurity standpoint. Yes, I think RPA is ready. I think AI is ready. Amanda gave some great examples there. But we have to think about cost and we have to think about risks.

Drew: You need to think like accountants. Basically.

Shimamoto: The holistic picture.

Drew: Wesley, I mentioned Power Automate, but I don't know if there are any other tools you want to mention. Then to the whole group. When you're looking at RPA, Wesley has got a business that he and his firm have launched that does RPA development. You get people who actually are very good at developing RPA to do it for you, which can save time obviously and hopefully get a better result. In those situations, what projects might be better outsourced versus done in-house? If we can start with any tools, Wesley, and then jump into the second part of that.

Hartman: Sure. Power Automate, I think is a great tool for everybody because it is, what I call a stacking RPA. The stacking RPA and then there's like flowchart RPA, which is like UI Path. Now, little disclaimer, I have a background in software development, so the stacking style works for me where the RPA goes line by line as opposed to following a flowchart. Now the cool thing is Power Automate Desktop is free if you have a Windows computer, so anybody can just click "Start," type in Power Automate, and it's pre-installed, especially if you have a Windows 11 computer. It's pre-installed.

That is a great tool for everybody. Me personally, I do a lot of RPA now in Python, which is a programming language using Visual Studio. I'm not going to get too deep because I'll nerd out a little too hard, I think. But if you have a programming background, then the Python is a good tool. Now to go into the outsourced-versus-done in-house. I think really one of the things you've got to do is look at what your current resources are. For example I talked to an accounting firm and they have a couple of internal IT people and one of them has programming experience. Honestly for them, they could do a lot of projects, even the harder projects they could probably do in-house, and just have someone as a programming consultant, where if they hit a wall, they can advise them a little bit.

And for I want to say small to mid-range firms if you don't have IT staff. You don't have those internal resources. Really, it's a cost-benefit analysis. I was talking to another firm and even though he would want to do it in-house but he's a sole practitioner, so he doesn't have the time. He's looking to outsource it, but like Donny was saying, he did the math and the project that we were talking about, while he would have liked to have automated, the dollars didn't make sense. He's holding off and seeing if you can do that in-house. Really, a big part of it really comes down to a cost-benefit analysis, as well as your internal resources.

Drew: That's an excellent point. Donny, did you want to touch on what projects would be outsourced versus done in-house? Amanda, you're welcome to jump in, too, if you have anything to add.

Shimamoto: Sure. I think again, Wes laid a good foundation. The way I often think about it, too, though, is even if you're going to outsource, because we see a lot of firms that don't have the right technical people like Wes in-house is, you still want to build awareness of when is the right time to use RPA.

As we said earlier, it's something that's very repetitive and also happens quite often so that you can get that cost-benefit justification in there. Having that awareness of, hmm, this might be something I want to actually have a robot do is good because at least you have your people actually helping identify opportunities that can then get escalated to someone, whether you have a tech-savvy person in the firm that can per se do a quick proof of concept or try and do some basic automation and then hand it off to someone who's more technical.

The analogy to think about on this is like Excel macros. I know a ton of accountants that are able to do really good Excel macros because you just press record and you do it, and a bunch of the RPA products will work like that, too. You press record, you show it what you're doing, and then what you need is the IT person just to do the last piece, which might be make it loop this certain way so that it works through multiple records, or build this last bridge over to the API or to the other software. You can actually leverage both in-house and external expertise and optimize that.

Wilkie: I would agree with both what Wes and Donny said. It really does need to be collaborative and you really do have to look at your resources. I've talked to firms who've said, we need to hire a programmer or we need to hire an RPA specialist. It's like, what are you going to do with them? They think, well, we need these skills, so we'll get them on board and then we'll figure out how to keep them busy. That's not a great strategy, either. There's a lot of resources out there in the profession that you can leverage, so it does need to be collaborative, but you also need to consider the skill set that you have on staff, and if you don't have the skill set needed to move these things forward, then there are resources out there.

I think that a lot of firms who may be putting some of these things off now because it doesn't make sense from a cost-benefit analysis, they'll get to a point where it's going to become important because it's going to be a competitive disadvantage if they're not leveraging some of this technology, and if they're not automating some of those mundane tasks that nobody is getting excited about doing.

Hartman: I just want to add one other thing. When I look at a project on the cost, because we've talked about cost-benefit analysis, one thing actually I would recommend, especially as we move into busy season or we're in the depths of busy season or sometime for some people, busy season never stopped throughout the year, do a recording of what you're doing. Open up a Teams session or something along those lines and just record the steps that you're going through and narrate it, because I will speak just as a developer for myself, or if you're going to have someone run through the process, that is the best tool to actually determine what the cost would be.

Then in addition, you can also determine your own costs. It's like, oh, well, this is a five-minute recording, it took me five minutes to do this under perfect conditions, and then actually, that can help you figure out your cost-benefit analysis, and then you multiply that by how many times a day or a week or a month that you do that task. Just wanted to throw that idea out there.