Differentiating agentic and generative AI — and more with a Tech Q&A author

Hosted by Neil Amato

Sponsored by Thomson Reuters

In this episode, Wesley Hartman, co‑author of the Journal of Accountancy’s Technology Q&A column, discusses how AI is reshaping work for accounting firms.

He explains the difference between generative and agentic AI and why both matter for firm workflows. Hartman also outlines the most pressing AI risks for CPAs, including hallucinations and emerging deepfake‑driven scams, which he wrote about in the February Tech Q&A. He closes the conversation with practical guidance for adopting AI tools methodically while avoiding common pitfalls.

Also, here are a few Technology Q&A columns related to the discussion:

What you’ll learn from this episode:

  • The ways Hartman uses AI in his own work.
  • The difference between agentic and generative AI.
  • Why “confidently wrong” AI responses can present risks for firms.
  • How inaction or “wait‑and‑see” thinking can create its own form of AI risk.

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 back to the Journal of Accountancy podcast. This is your host, Neil Amato.

AI risk is everywhere, and that risk is the topic in one of the Journal of Accountancy’s Tech Q&A columns out this month. I talked to the author of that article about AI risk in mid-December at Digital CPA. You’ll hear the discussion right after this brief sponsor message.

[Sponsor message]

Amato: My guest today is JofA Technology Q&A co-author Wesley Hartman. He’s the founder of Automata Practice Development. Wesley, welcome to the podcast. We’re glad to have you on and glad to see you in person.

Wesley Hartman: Yeah, fantastic. I think this is the first time we’ve actually met in person, you and I.

Amato: It is. I’m glad we’re able to make it happen with a podcast that is tied to some of your JofA articles but also just generally the topic of AI.

Hartman: Absolutely.

Amato: Can you first, I guess, briefly describe how you have been using AI in your business?

Hartman: As an automation developer for accounting firms, we use AI to help us code. It integrates directly into our tools, and it actually sees the code that we’re writing and it functions as an AI assistant. We are the orchestrator of the knowledge and logic of the accounting firm in the business use case, but it helps us with some of that nuanced code to make sure that we’re doing things very efficiently. Also in some ways, from a coding perspective, it does the equivalent of spell checker to make sure we’re not missing our punctuation with our code.

Amato: I like that. I like the spell-checker comparison. That’s a good one. This may be a simple question, but I think it’s a good one and it’s from Jeff Drew, JofA editor-in-chief. How much has AI changed what you do?

Hartman: It has changed massively how our business works and runs. When I started my business, ChatGPT, the big name, and some of the other tools had just started coming out, but they were not really integrated as code tools, or it was just a novelty at the time. But over the course of my business time, as we figured out how to use it and use it well, we’re able to develop automations much more quickly from a code perspective, which really lets us focus on the relationship and helping our client accounting firms get the impacts of automation. In some ways, while our clients, maybe they’re not using AI directly, but they are definitely getting the benefits of us as a consultant using AI.

Amato: That’s great. Now, you’ve written a lot about ways to use generative AI and agentic AI in the Tech Q&A column. First, actually, I want to ask you, can you briefly define the difference between generative and agentic AI?

Hartman: That’s a great question, actually. Generative AI is what a lot of people initially are seeing with generation of text. I input some text. It goes into the magic AI black box and then it creates new text, generating it, essentially. Now the difference between generative generating text versus agentic — agentic takes action. I was here doing a workshop for how to use AI tools. I showed attendees an example of agentic workflow where you have an email address that we’ll call AP@mybusiness.com.

Invoices come in, and those invoices automatically go into an AI agent, and that AI agent will input the invoices into your accounting platform, and then it pushes the results of that to another agent that will then review the accounting work that was done by the previous agent. Then after that, that second agent creates a report, which then goes to you the user because it’s important to have a human in the loop with this sort of work. But the big difference is generative is creating text while agentic is actually doing work.

Amato: I think that’s really well said, and we like to toss around those terms a lot, but sometimes it helps to just go back and actually say, “Wait, what are the differences between these two?” So, thanks. In the February Tech Q&A, you’ve written about AI risks. Because as great as AI is, it comes with plenty of risks. First, what inspired you to address that topic?

Hartman: I think there’s an excitement around AI, and there’s a lot of discussion about how it’s going to change everything. But I also know that in the accounting space — my father-in-law is an accountant and I worked at a firm for 15 years — accountants are very risk-averse, and so they want to know, “What are the issues, what do I need to worry about?” and not always just dive in headfirst.

That really inspired me, especially over time, as AI researchers are looking into edge cases of how to use AI, it’s starting to come out to show, OK, here are problems, here are things that we’ve been able to experiment with, and some of them are very concerning and we really need to keep them in mind as we implement them in firms.

Amato: You’ve written about a good number of the risks in that article. Do you want to highlight, I don’t know, two, three, four — something like that?

Hartman: Sure. I’ll tag a few of them that I think are some most obvious, some most interesting, some most impactful in the accounting space. We’ll start with the obvious one, which is probably talked about a lot, which is hallucinations. That’s where AI gets things wrong. Part of the problem is not necessarily the AI getting it wrong, it’s that it is confidently wrong. It will sound like it knows exactly what it’s talking about wrong.

Because of that confidence, we can get a false sense of what it is. From an impact on a firm perspective, the problem you’re going to run into is clients are going to say, “Well, the AI told me this. I’m a software company. Why can’t I deduct my dog’s food? The AI said I could.” That’s the hallucination side of things, both from an internal firm and external.

I think one other important one is, that’s more of an inadvertent risk, we just don’t think about. But then on the flip side, we have malicious actors out there, whether it’s hackers or scammers that are using things like deepfakes.

There is a situation where scammers deep-faked a video call and a voice — I think it was a CFO of a large company — and were able to initiate a wire transfer of, I think it was a seven-digit sum of money from the company to offshore bank accounts, and it just kind of disappeared, I believe. That’s, I think, the two biggest risks. You have the inadvertent risks of hallucinations, but then you have external risks of people actually using these tools maliciously, and we got to worry about mitigating those.

Amato: Those are some of the risks you’re highlighting in the February JofA. Overall, what do you think are the three biggest AI risks for CPAs?

Hartman: We got the hallucinations, like I mentioned, clients bring up information, scamming. I think probably one of the other bigger risks involved is choosing not to do anything. Like, analysis paralysis, are we going to wait and see? I’m not saying implement every single AI tool that’s out there, but the important thing is to get some of these AI tools in the hands of yourself, your staff, and set up guidelines on how to use it. Because people will use tools that will help make their lives easier.

The biggest concern is that if there’s no guidance from firms and accountants, then it’s the Wild West. And they’re going to put in information that maybe they really shouldn’t be putting into thinking, well, it’s helping themselves, but it’s not secure or it’s not used properly. That’s definitely one of the other big ones.

Amato: And then, what do you think are the three most important things for CPAs to know as far as the trends in AI looking ahead in the new year?

Hartman: I think one of the most interesting things is that there already has been a flood, and the flood is only getting bigger of vendors that are bringing AI tools to market and, for lack of a better word, trying to help the space. But they see it as there’s a good opportunity.

You’re going to get overwhelmed. Accounting firms are going to (have to take it slowly). Just remember, take a breath and then just go methodically, make sure that if you’re implementing a tool, you’re solving a problem, not just implementing the tool so you can say that you’ve implemented AI.

That’s definitely I’d say one of the biggest things to start with. I think one of the second things is that, yes, it feels like AI is steamrolling everything as well, and if you’re not on it, get on the train. I do think that you need to be aware of it and that sort of thing. But even from the CEOs and the heads of the AI companies, it turned into, oh, this is going to be the decade of AI agents. There’s that concern like, OK, if some of the larger [companies] are saying, “It’s a decade of AI,” you don’t have to necessarily worry that, if I don’t do something at this immediate moment, I’m going to be left behind and suddenly I’m fossilized or something like that.

I think it’s important to just remember to be methodical, solve problems, don’t do things just because. Then I would say the other thing is remember that in your usage of AI, AI is also there to keep you engaged, which can be good, especially for inspiration and creativity. But also, going back to it, make sure that when you’re using AI, you’re using it with purpose, to solve a problem, to reach a goal so that you can evolve your firm to the next level.

Amato: At this conference, Digital CPA, Erik Asgeirsson of CPA.com said, to make a baseball reference, It’s kind of like the third inning of a nine-inning game. It’s by no means over. You’re not out of the game, but you need to be ready, but also don’t feel like it’s too late.

Hartman: Absolutely. Just remember also that, like in the AICPA, you have your third-base coach that’s there to help guide and let you know when is it time to sprint versus when is it time to hold and wait.

Amato: Oh, it’s great. I like that analogy. Wesley Hartman, thank you for being on the JofA podcast.

Hartman: Absolutely. It’s great meeting in person.