Adam Larson: 00:04

Welcome to Count Me In! In this episode, we welcome back Tala Khalifeh, Chief of Staff at Stacks and a leader for IMA's shared interest groups. With a career shaped by her deep commitment to management accounting and ethics, Tala brings a unique perspective on the evolving crossroads of artificial intelligence and ethical decision making in the accounting world. She shares candid stories from her own journey, including pivotal moments that tested her personal values, and offers practical insights into how AI is transforming, but not replacing, the critical need for human judgment in finance. So whether you're curious about AI's growing capabilities, concerned about ethical risks, or seeking strategies to keep your organization's financial data trustworthy in a tech driven era, this conversation is full of essential guidance for navigating the future of accounting.

Adam Larson: 00:56

Well, Tala, thank you so much for coming back on the Count Me In podcast. We're excited to have you here as we're going to focus on ethics and accounting and that you, you know, you've kind of built your career around that at that crossroads. And I think the combination of makes your perspective on this topic very interesting. So, you know, we're going to talk about AI and ethics in there, but can you maybe paint a picture for us on how you ended up in this space? What first drew you to management accounting?

Adam Larson: 01:18

And when did the ethics dimension really start to take shape for you?

Tala Khalifeh: 01:23

Hi, Adam, and thank you for having me today. So if I think about how I ended up really interested in accounting and ethics, it actually goes back to my early career. There were two chapters that truly shaped who I am as a professional and they each left a different mark on me. The first one was internal audit. That is where my career began and it gave me something that has stayed with me ever since, which is a risk mindset.

Tala Khalifeh: 01:52

So as an internal auditor, you look at any process and ask the same questions. Where's the risk? How can we mitigate the risk? How can we enhance internal control? So when I later moved into accounting roles, I carried that lens with me.

Tala Khalifeh: 02:06

So I was always that person in the room making sure dual signatures were in place, that maker checker processes were being followed, cash was properly handled. Yes, I was this annoying person, but it wasn't because I was told to, but because as an internal auditor, I had seen what happens when those controls are missing. So that first chapter made me more disciplined and detailed and also like precise. But the second chapter is what made ethics deeply personal for me. I remember at a certain point in my career, I found myself working in an environment where unethical practices were happening around me.

Tala Khalifeh: 02:46

And the more aware I became, the clearer it was that I could not stay in this environment. So I chose not to be part of it and I resigned. I walked away from that role to protect my integrity. So that was not an easy decision, but it was the only one I could live with. That experience fundamentally changed how I see my profession.

Tala Khalifeh: 03:07

It showed me that ethics is not a topic we discuss in a webinar or we write about in an article. It is a daily choice. It is something you either stand by when it matters or you don't. So after that, I joined also the IMACE Committee on Ethics because I could see that awareness was genuinely missing. People were not seeing these situations clearly or they were seeing them and staying silent.

Tala Khalifeh: 03:31

So I wanted to change that and I wanted to know more about this topic too. So spreading awareness is something I am deeply passionate about today. That's why I like to write a lot about it and I enjoy presenting webinars and organizing events about this topic.

Adam Larson: 03:47

Like you said, it's a daily choice. I think for a lot of people, don't realize how one simple choice can change the trajectory of, you know, everything that you're working for or working towards. That's it's not like it's suddenly like, oh, oh, I see this person, they're a criminal and they're going to make the unethical choice. It's more like, hey, it's just a regular person like you and I who just might make a choice and they start walking the other direction.

Tala Khalifeh: 04:10

Exactly. It's like those daily small decisions that add up to change, to have like major change in the reports and the credibility of the report. So yeah, it's a daily, it's a daily thing.

Adam Larson: 04:25

Yeah. So the topic for today, we wanted to kind of cover like ethics and AI, the ethics around AI. And back in 2023, you published an article in strategic finance that kind of had the little question, you know, is AI going to replace accountants? Know, we're we're three years later, maybe walking back to that moment and some things you've learned since then.

Tala Khalifeh: 04:46

Yeah, well, 2023, the year 2023 was really a turning point for finance profession. AI was suddenly everywhere, not some like a future idea or a concept, but as something people were actually using day to day. And it was also positioned as a tool that could categorize transactions, prepare reports and even build dashboards faster than an entire team. So that created real anxiety. People started questioning their role or their relevance or where they fit in a world where so much could be automated.

Tala Khalifeh: 05:23

But a personal experience made it even more real for me. A close friend reached out to me because she needed help with her accounting exam. She'd been using CHAD GPT to study, but it kept giving her incorrect answers. She got frustrated and she was also starting to lose confidence in herself when in reality the issue was the tool. So that immediately caught my attention.

Tala Khalifeh: 05:47

So when the opportunity came up to write about it, I didn't hesitate. I was already curious, already asking questions. And honestly, like everybody else, I was anxious too. I wanted to understand what these systems could actually do that maybe I couldn't. So I started testing the tool myself using the real accounting scenarios.

Tala Khalifeh: 06:07

I questioned the outputs carefully and I also paid attention to where it worked, where it failed and why. So that hands on process is, is what made this article also practical, not just a theoretical.

Adam Larson: 06:23

You did, you kind of, you asked chat you like, it's like you interviewed chat GBT for the article, which which is kind of fun. You know, you're like, hey, AI, I'm going to interview you for this. Did

Tala Khalifeh: 06:32

you like,

Adam Larson: 06:33

as you were doing that, you were watching it fail. Did it improve as you went through it or did it get worse as you went through it?

Tala Khalifeh: 06:39

It did. It actually did. The first time we tested it, we were giving the instructions in a table and a structured financial data. And then we translated this information into plain text. So this is where ChatGPT back then gave us the right answers.

Tala Khalifeh: 06:57

It's when we gave it the instructions or gave it the given in a format that was easily understood and interpreted by the system. Back then, there was a limitation to the ability of CHADGPT to understand and interpret properly the given and the instructions. It was more about the way the data is presented rather than the content itself. And that's why it was failing. But then I know that the CPA exam failed at the beginning and then a few months later it passed and it scored 85 out of 100.

Tala Khalifeh: 07:37

So I think that's the technology was moving fast and it's still moving fast. Like what ChatGPT or any AI platform today can do is way more advanced than what, it could actually do three years ago.

Adam Larson: 07:54

So in your opinion, how does that make you feel? Do you think that those passing scores and that we can trust it more or does it raise other questions for you as you're like, okay, well just because it passes, does that mean it's like, oh, we're good to go?

Tala Khalifeh: 08:07

No, definitely not. So AI can now directly work with, structured financial data in a much more meaningful way. It can build reports, forecasting model, dashboard and interactive scenario planning models directly. Whereas in 2023, was nearly impossible. Back then AI couldn't even interpret an Excel sheet.

Tala Khalifeh: 08:31

Now the platform can build this model from scratch and not just build an Excel template, but also include formulas and an interactive scenario planning model. So this is really impressive. And also remember Adam, that back in 2023, AI wasn't integrated into financial systems. It was simply ChartGPT, Excel input, manual input. Now they are integrated into ERP systems, BI tools, treasury system, AR and AP platforms.

Tala Khalifeh: 09:12

So, but we also need to be, honest about what a passing score actually means. So passing an exam proves that the system can apply the knowledge and recognize the patterns in a structured setting. It does not prove professional judgment in unclear situations. An exam has clear questions and defined answers, whereas real finance is nothing like that. It involved incomplete information, competing priorities, judgment, and real world consequences.

Tala Khalifeh: 09:44

So AI has advanced and now I'm sure that even if we test it again, I'm sure it'll pass any CPA exam or any other accounting exam, but it still does not mean a system can be trusted with professional judgment.

Adam Larson: 10:03

Yeah, it's like any anything can be trained to take a test, where does it come where those real role choices come into play? Where, where, like where will it shine up against that?

Tala Khalifeh: 10:14

Exactly. So, the technology, even if it's moving fast and even if it was trained and even if we gave it the right information in a very, in a format that it can be easily interpreted, that should push us to step up and not step back because finance professional is the most important skill. And this is where AI now is still falling short in context and explainability. So it's not the AI system does not understand your business. It does not understand your vision or your strategy.

Tala Khalifeh: 10:51

It does not understand what you're planning to do next year or why, for example, you spent so much on marketing last year. So if AI needs to give you an analysis or wants to support in decision making, it might not be as accurate as a finance professional because the context of your business still lies with the finance professional.

Adam Larson: 11:14

Yeah. Yeah, I think that there's that there is that tension that's been kind of built between AI and and protect practitioners, whether it's in accounting and finance or pretty much any any industry at this point, because you have these AI tools that can run through millions of lines of data at a speed that, you know, no human can match. Like none of us can keep up

Tala Khalifeh: 11:31

with Exactly.

Adam Larson: 11:32

And so your intuition is like, oh my gosh, like how do I compete with that? You know, and you kind of mentioned a little bit there's the human judgment side of it, but it's still is a little scary because you have, you know, more advanced technology coming up each and every day and so you're like, well, at some point or something like say, oh, I can live with the judgment of the robot.

Tala Khalifeh: 11:53

Yes. Well, no, we can't. We can't. We haven't reached this, level yet and I'm not sure we will ever. Human oversight will always be needed, even though systems will get more and more advanced and speed is where AI has truly delivered.

Tala Khalifeh: 12:10

Tasks such as reconciliation, transaction categorization, dashboard creation, forecasting work that once took hours can now take minutes. PwC's 2025 Treasury survey found that around 74% of finance teams are already using or expanding AI capabilities. So we are no longer testing it, we are also using it. And I see this in practice every day through my full time role with an accounting firm where I get to see how accountants and CFOs are working with these tools and through also my freelance project with small business owners. So AI has significantly changed how quickly I can prepare my work, but that frees up my time to focus more on the most valuable part of the role, which is helping clients understand what the numbers actually mean for their business.

Tala Khalifeh: 13:07

And this is where AI still falls short, context and explainability. So as I told you, it doesn't know the client, the strategy, the vision, the story behind the numbers. So even if AI is advancing, the rule remains the same, that output still need professional review. Also, if we look at the adoption numbers, we can see that we are still far from full transformation. Bain and Company's 2026 CFO survey found that only 15% to 25% of CFOs have fully scaled AI across their finance functions.

Tala Khalifeh: 13:47

And satisfaction levels rise significantly once organizations scale properly. Okay, so once companies implement AI across their finance functions, they get to be more satisfied than with companies that are still testing it in a small project.

Adam Larson: 14:09

Yeah. So it seems like more people, the more people use it, the more they're accepting and understanding of it.

Tala Khalifeh: 14:14

Exactly. Exactly. And the more they link them together, like the more you link software's together, the efficiency of course, improves. This is mainly AI's, best, feature is speed.

Adam Larson: 14:30

Well, yeah. And if you and if you have an established system that's working well and you insert some sort of AI technology into it, you're not having to worry about, oh, is my is my is my information getting sent out there? It's housed in this enclosed space and then it can actually utilize your your numbers to get the, you know, to give the right answers. When you're doing that in practice, you know, what if, you know, an accountant says, oh, this doesn't look right. I don't I don't feel like and leadership's like, well, but no, but we have the process in place.

Adam Larson: 14:58

It's fine. You know, like, does that look like when you're when you kind of because we might come with the other side where people were fudging the numbers one way and now it's like, well, the AI did it right because they set it up that way, but what if there's something wrong and nobody wants to look at it?

Tala Khalifeh: 15:11

Exactly. So, this is the problem. The problem is that when outputs are not questioned. So I remember I mentioned the black box in my article, like you can see the input and the output, but you can never see how the results or how the conclusions were reached, how they were reached. So the thing is that senior accountants or finance professionals in general can really look at this at any output and sense that something is off.

Tala Khalifeh: 15:45

And the problem here is that we need to question these outputs and we need to challenge these assumptions to verify that everything we need to verify that they are accurate and they are real because remember that accountability still lies under the human or the accountant is still the person who is in charge and he is the one that should make sure that this information is credible, that this information is reliable. So accountability did not transfer to the machine. So we cannot just say, 'Hey, we trust the process. We know we've tested this. It's okay to accept it.' No, whenever we are in doubt, we need to investigate further.

Tala Khalifeh: 16:30

We need to check and we should also redo the work if we are in doubt because we need to make sure that the information is credible. Like imagine AI is like having an extremely capable junior accountant, fast, efficient, and always available, but you still would not send that work directly to a board or senior management without review. So that same principle applies here. So human judgment is not about processing more data or just relying on the machine. It's also about asking the questions the machine still cannot answer such as, does this number actually make sense for this business in this specific industry at this specific moment?

Tala Khalifeh: 17:14

So these questions will always belong to the accountant.

Adam Larson: 17:20

Yeah, you have that, you need that professional judgment and there's the learning expert Elliot Mazie. He always says, you know, you have the artificial intelligence, but you also need the HI, the human intelligence, and they have to work together in order to be successful, especially when you're building something that, you know, you can always say, okay, hey, I, you know, build the beginning of this presentation for me, but you can't that can't be what you put out into the world. It has there has to be some sort of human interaction. So we still have to have our critical thinking skills, even though humans seem to be saying, well, I'll just let the AI make the decision for me, which probably isn't Definitely. The best ethical decision to make

Tala Khalifeh: 17:53

Because, the machine can always be wrong. You always need to challenge the outputs and question them whenever you are in doubt. That human layer, hasn't disappeared yet.

Adam Larson: 18:09

No, it hasn't. And and especially as professionals, you know, you have to if you suspect that the AI output might be wrong and there there might be organizational pressure to push forward, which is, you know, the your typical kind of ethical decision for accountants like, hey, well, the numbers are good enough. Just just push it forward. You know, do you think that will start to happen, especially in, you know, professional organizations when they're just pushing the AI output now?

Tala Khalifeh: 18:36

Definitely. Definitely. Think this will now create a new ethical dilemma, which is accepting AI's output without questioning or without really assessing the credibility of the information. But let's go back to the IMA statement of ethical professional practice. So we have principles that we need to abide by, which are like honesty, fairness, objectivity, responsibility standards such as competence, confidentiality, integrity, credibility.

Tala Khalifeh: 19:10

These do not disappear simply because AI is involved. So organizational pressure does not transfer accountability away from the individual. So even if the system prepares or processes the data, as I told you, the ownership still sits with the accountant. But what we can do is we can have a process for escalation. So let's say an accountant doubts that this information is credible or that net income is correct, we need to have a process when to escalate, when to accept, when to flag.

Tala Khalifeh: 19:46

So that's why the most important thing is to have a So the first step is governance. Companies need a clear AI policy that defines which tools are approved, how they should be used, what data can be processed, and how outputs must be reviewed. This is the most important step in order to ensure at least to establish a foundation for credible information. Without this, you won't be doing a control over risk. So employees may use unapproved AI platforms, expose sensitive data, or rely on AI output without proper review, which this will also increase the likelihood of inaccurate or unreliable financial reporting.

Tala Khalifeh: 20:32

The second step, we need to determine where AI can operate with minimal oversight and where human review is required because not all financial tasks carry the same level of risk. So the objective is to focus human review on higher risk and more material areas rather than reviewing every AI generated output for sure, which would reduce the efficiency benefits of AI. But if we doubt that a number is wrong, we need to still interfere, review and fix it. And also the third step, we need to establish review threshold and escalation rules to identify what needs to be, like, let's say any payment above $1,000 must be reviewed by like the senior accountant, not accepted it automatically depending on, the materiality and the risk associated with this payment. These kinds of rules that we can apply to mitigate the risk of AI.

Adam Larson: 21:38

Do you think that, organizations are going to start defining like an acceptable margin of error for your AI output, you know, kind of like how they do for other things? What would that look like in a real finance team? And then is that also like giving permission for letting errors slide? Like how, where's the balance there?

Tala Khalifeh: 21:54

Yeah, definitely. Well, you know that there's no currently no universal standard for how much error is acceptable in AI generated financial work. And that creates a challenge because many organizations are already using AI without clearly defining what level of risk they're willing to tolerate or how they will catch errors when they occur. But I think that the goal is not to accept a certain margin. The goal is to reduce it.

Tala Khalifeh: 22:23

And in finance, we already know that and we already do this, and this is why we place strong internal controls. So finance professionals already work with materiality estimation risk and control frameworks almost every day. And I think that AI should just sit inside that same structure. So as I was saying before that the first step is governance, have a clear AI policy and also assess materiality and risk for each task to know the level or when human intervention is needed more than other transactions, other tasks. So I think this is how we can control it and mitigate the error.

Tala Khalifeh: 23:06

And this is not a one time step. This should be an ongoing process. Controls need to be monitored and adjusted as processes and AI tools evolve. The AI tools are going to evolve, they're going to become better, they're going to be more advanced. So that's why we need to stay also on top of it.

Tala Khalifeh: 23:27

So we also want to test it every now and then to make sure that we're still applying the proper control tasks, at least to match the new skills of AI.

Adam Larson: 23:42

Yeah. Well, and for everything that you're saying, it sounds like nothing really needs to change within the finance organization. If you have good internal controls, if you have good processes in place, you can fit the AI into that system

Tala Khalifeh: 23:55

and

Adam Larson: 23:55

if loans are still following good, good proper accounting practices, it should fit in the system pretty well.

Tala Khalifeh: 24:02

Right. That's correct. Like any other software that you're using any anything, other like we as finance professionals, we apply this in everything we do. So, this is exactly similar. This is not different.

Tala Khalifeh: 24:19

So, we need to treat it with the same concept.

Adam Larson: 24:23

Yeah. So, you know, somebody's listening to this conversation. They're inside an organization that has already gone all in with all their AI tools. What is the mindset you want them to take away from this conversation?

Tala Khalifeh: 24:34

I want them to know that the biggest mindset shift is recognizing that trust and financial outputs is becoming one of the most valuable skills in accounting. AI is already automating large parts of our traditional finance work, but organizations still need professionals who can review the outputs, interpret results, challenge assumptions and confirm that the conclusions actually make sense. So that human layer is what makes financial information credible. So the role of the accountant is shifting today from processor to interpreter and increasingly also to guardian of outputs. So the professionals who will remain essentials are the ones trusted to step in when the system reaches the limits.

Tala Khalifeh: 25:25

And the good news is that the profession is responding. Credible structured learning on AI and finance is finally available. And that matters for a long time that simply was not there. The IMA's AI and Finance credentials is one of the strong examples of that. And when I heard about it, I was genuinely so happy to see it because I'm happy to see that the profession is catching up and this is exactly what we need.

Tala Khalifeh: 25:52

So after also watching this evolve over the last three years, I can say this clearly. AI has been a genuine win for finance, has upgraded the work, but it has not replaced the need for human judgment. So the mindset shift is this: stop asking whether the machine will take your place. Start asking what you can do that it cannot, but requires your judgment, your ethics, your accountability. So no model can replicate that.

Tala Khalifeh: 26:22

So own it completely.

Adam Larson: 26:24

Yeah. No, no, thank you so much. That was great, great, advice and some great insights. I really appreciate everything you put into this conversation and sharing your insights with our audience today.

Tala Khalifeh: 26:35

Thanks, Adam.

Announcer: 26:38

This has been Count Me In, IMA's podcast providing you with the latest perspectives of thought leaders from the accounting and finance profession. If you like what you heard and you'd like to be counted in for more relevant accounting and finance education, visit IMA's website at www.imanet.org.