Achille Ettorre:
From a revenue perspective, I truly believe that using these tools or these models can increase the revenue for any organization that understands the processes involved and the people that are actually working these models.
Narrator:
Welcome to Accounting for the Future, a BDO Canada podcast for financial leaders to navigate change, and achieve business growth. We'll uncover the challenges financial leaders may not have dealt with yesterday, but we'll definitely have to manage for the future.
Anne-Marie Henson:
Hello and welcome to BDO Canada's Accounting for the Future. I'm Anne-Marie Henson, and I have the pleasure of welcoming Achille Ettorre to the podcast. Achille sits on the advisory board for the Master of Management Analytics and Master of Management of Artificial Intelligence at the Smith School of Business at Queen's University. He has over 15 years of experience in building and leading analytics programs, creating infrastructure and leveraging data insights for global organizations. Currently, he serves as a faculty member and executive consultant at the International Institute for Analytics, the authority on analytics, maturity and best practices. He also presented at TEDx, his presentation titled How AI can build the life you want has over a hundred thousand views on YouTube. And his mission is to elevate the industry and transform organizations with data-driven solutions. Achille, welcome to the podcast.
Achille Ettorre:
Thank you for having me, Anne-Marie. I'm excited to be here.
Anne-Marie Henson:
Fantastic. So did you ever think that your experience would lead you to be the guest on an accounting podcast?
Achille Ettorre:
Truth be told, no. Although I do have a background in finance, earlier on in my career, I did work in a finance role. I did not think I'd be in a podcast like this.
Anne-Marie Henson:
Fantastic. Well, look, we're really happy to have you. Our listeners are probably used to this, but what we like to do is to talk about trends and things that are happening in the world and then talk about how they could impact the accounting and the finance function. So your background in finance probably suits us perfectly and we're really excited to talk to you today.
Achille Ettorre:
All right.
Anne-Marie Henson:
So I guess we'll start with a couple of more basic questions to set the tone. So could you just talk to us a little bit about how you define generative AI and how that's connected to what's known as big data?
Achille Ettorre:
Well, we'll get right into it. I mean, for me, generative AI refers to a type of artificial intelligence that can generate content, whether it's text, images, audio, quite frankly, even code, and it does it in a much quicker way than an actual human can. The models, the importance of it is that the models are trained and elevated in such a way that there's minimized bias within the actual output of the model. Maybe a little deeper in terms of a simple definition per se, but in the space today, the arms race is working with the models and training them in such a way that they're ethical and provide outputs that are meaningful and non-biased, ethical to whoever's using them.
Anne-Marie Henson:
Interesting. And then what would you say is I guess the connection to big data and how would you define big data for those of us who are a bit less accustomed to those kinds of words?
Achille Ettorre:
Big data. So big data as the definition holds itself, its data structured in such a way that can be any data point that an organization or anyone creates manifests or processes. It can be done at the POS, it can be done at a PO, it can be a GL account. It's the point at which you can actually use and create multiple data points or touch points so that you can put it into ingest it into a model. Today we're talking about generative AI or any type of model. By the way, you can do an Excel spreadsheet or anything of that nature to produce the data or information or insights that anyone is looking for.
Anne-Marie Henson:
That's fantastic. No, it's really, really helpful and for you to put it in that way makes it really easy for us accountants to understand, we definitely know what GLs and those types of data points are. So you spent many years now working with businesses on optimizing their processes through analytics. How has AI been able to impact the speed or the ability at which businesses can improve their processes?
Achille Ettorre:
That's an interesting question because the quick answer is it's a really powerful tool that's supposed to increase the speed of anything that you're trying to find out. So today is LLMs or ChatGPT, but in the same token or in the same breath, some organizations, larger ones may have some trouble actually productionalizing the actual output of these models in such a way that it's insightful just because of the size of the actual enterprise. So I'll use my previous employer someone like a Loblaw or a Walmart or a DHL. In another token, I'll use my family business as an example. My family business rents construction equipment and we are able to build models and almost bridge the gap from what some of those larger organizations can do given the resources that they have.
Some examples include go-to-market strategies, dynamic pricing, understanding what a loyalty program should or shouldn't be for particular customer segments. All things that are traditional in a larger organization, it's productionalizing that model in such a way that it's meaningful and doesn't have biases - that's the hard part. In a smaller organization that I mentioned, which is myself or any other small business, it provides the opportunity for that entrepreneur to close the gap and being more competitive or competitive within its current marketplace.
Anne-Marie Henson:
Oh, that's really interesting actually. Because you would think that the large enterprise type organizations that have access to such rich data would be the ones that would be able to benefit the most or the quickest from AI. But I guess it sounds like what you're saying is because sometimes there's I guess, such thing as too much data or not knowing what to do with that data and how it can help you make decisions. But for smaller SMB type organizations, it allows them to scale and to actually be much more competitive with those organizations because the speed at which can adopt these kinds of processes.
Achille Ettorre:
Absolutely, and if I could build upon your note or your comment, I would say a larger organization has so much more inputs, data inputs that they really have to have a governance program to ensure that all the throughputs are done in such a way that it's systemic, methodical, and correct every time. Because if you do not set something up in an article correctly in the beginning, it'll only create a huge amount of opportunity as it goes through the process up right until you hit the cash really where you sell it to the individual. Your sales could be recorded incorrectly, your cost of sales could be recorded incorrectly, and that's literally what the larger organizations see or what I've seen from what I work with smaller organizations, not big data or as much data, so easier to control, but they're more susceptible to ethical decisions whether... What type of profit margin is reasonable to endure for particular service that you're offering.
Anne-Marie Henson:
Oh, that's super interesting. I know we've already started talking about this a little bit, but I'd like to expand a little on what the impacts could be. There are a number of different facets I'm sure, but we could talk about how it could impact a company's revenue, the revenue model, as well as the risks and opportunities. So I'd love to hear your thoughts on those.
Achille Ettorre:
So it's interesting from a revenue perspective, I truly believe that using these tools or these models can increase the revenue for any organization that understands the processes involved and the people that are actually working these models. So you can do this through a dynamic pricing type strategy within ChatGPT, and I saved that ChatGPT because that's the most popular one I find. You can input what particular funnels you would want to implement for inventory that's aging, and you want to just deplete within your organization to release the cash that's being held within that inventory. You can actually look at minimizing your actual cost structure by structuring your buys within your supply chain. So adjusting your mins and maxes to say, when I reach a certain amount of inventory widgets within my distribution facility or my supply chain, I'm going to put in a minimum order of a thousand widgets versus doing four orders of 250 each. That's on the revenue perspective.
You talked about the risks.
There is risk to that, and what I would suggest is that for instances where you can go to speed to market in such a way that's very quick, not only do the individuals need to understand the impact of what it is that they're doing, but the individual needs to understand that the operations can actually deliver what the model suggests. If you're running 20,000 SKUs and you want to increase prices by 30% or 32%, the work required to actually get that to the storefront can be pretty laborious unless you have digital signs that you just adjust at the shelf. So it's something to think about if you're hitting a weekly flyer or if you're hitting any type of deadlines. So risk to the company go twofold. One is at the person at the individual level, they have the ability to really increase or decrease the pricing of a particular product.
They need to understand the scope of the competitive landscape so that what they're doing actually is able to sell through the product for what it is that they're doing and that the operations is able to fulfill what the changes are, whether they're week to week or month to month. Those risks are generally handled well when companies have a strong governance structure and processes are fully defined within the entire organization. And what I mean by that is typically your typical silos that exist everywhere are connected through the different data points of the data flow, which is otherwise known as data linkage throughout your organization. The more that you can understand the processes that a company has imposed or written out for their organization, the better position they are to minimize their risks for anything to go incorrect.
Anne-Marie Henson:
That's really interesting actually, and I want to expand on one thing that you said. So we did actually a podcast episode a couple of months ago on AI and the finance function. And we did talk about the importance of governance, which you're talking about quite a bit actually. And one of the things you just mentioned was the importance of having the proper governance process throughout to understand the inputs, to understand the value chain and what needs to happen at each sort of step of the way to really be able to maximize and to reduce your risks when you're implementing AI.
But I guess what I hear a lot of is that AI and the speed at which technology and innovation is moving is so fast that it's difficult for a company to stop and say, "We're going to put in all of these the right processes and controls to be able to have the right governance process to then be able to maximize our use of AI." A lot of companies are just going ahead and saying, "If we don't go ahead now and figure it out as we go, then we're going to lose out on increased margins, new revenue streams, and things like that." So what do you have to say about the balance between speed and governance?
Achille Ettorre:
Well, for the companies that choose to go down the path that you just suggested, it gives people like me, independent consultants, the opportunity to come in and really focus on helping them get to where they need to. So it's a whole industry or there's a whole body of work that can be done afterwards. And I can tell you as a consultant, it's easier for me to come in and help you set up that structure or anyone to set up that structure and help you build on that maturity on doing it correctly the first time. So that when you're truly ready to scale, you don't implode two, three months down the road. And what I mean by implode is you can implode operationally. You can implode by not following any type of tax regulatory requirements, whether it's locally, I'll say locally within Canada or any type of international standards and reporting requirements that are out there.
Being someone that's regimented to set things up beginning and spending that extra time, money, and resources in the beginning will save you a significant amount of heartache down the road if things were to implode. And your risk level for me anyway, wouldn't be worth... It wouldn't be worth it for me. Now, if there is a sandbox environment or a play environment or call it an innovation center where you kind of want to run these type of models and see what happens in a fake environment, that's great. I'm all for that. But to put it in production I think would definitely not be the best decisions for organizations. And I think that if they're led to believe that they're going faster, I would suggest that they're going to slow down, they're going to hit a wall pretty quick.
Anne-Marie Henson:
Right. No, and it's good to know that... Like you said, there's a way to go relatively quickly, but with the right expertise and the right people in place to help you put things together so that you're not missing out on those opportunities. I guess I'd like to hear about your work or your thoughts on more the governance and the director level. So audit committee members or directors, how can they get up to speed on AI and what can they do to improve their AI readiness so that they can help companies make the right decisions in terms of what direction to take?
Achille Ettorre:
That's a great question, and I think for board members specifically, spending some time on whether it's informal education or some formal training on understanding what AI can actually do within their organization, or quite frankly entire industry would definitely bode well. And I think that education doesn't stop at the board. I think that education is really throughout the entire organization or society. This is not going away. It's not perfect right now, it's really not, but it's going to get better a generation from now. It'll be a lot different. I think they need to embed it right at the beginning. I do think bringing in consultants or people that are comfortable and have done things from a AI perspective or an analytics perspective, standing up a governance team or running a governance team or even failing at governance can hold a significant amount of positive education or insights as to how they can proceed within their own organization or their projects of whatever they're doing.
I also think, I do think people in finance and audit, they need to be at the front of the table when things begin. And I've seen board members actually sit at the table and being part of the projects and providing obviously what their insights are. But getting a true understanding of what the potential hiccups could be right out of the gate in prior pre-AI and/or pre-big data, you would have time to respond, you would be able to set up for a board meeting, and it's a quarter report. Now, you're not afforded that time. You can literally stand up a model and an AGPT for your personal organization in less than a day. And so you would need to spend the time to be there. I think those are three big nuggets that would definitely help and the advice that I would provide.
Anne-Marie Henson:
Yeah, and I like the one on having accounting and finance people at the table at that same time. I think oftentimes the finance function tends to be looked at or at least in the past, more in terms of, okay, well here's the contract we entered into six months ago and here's the revenue that came in, so now you need to account for it, or you need to set this up. And I think like you said, with the decisions that need to be made and the changes that could happen to organizations, having those functions at the table to talk through what could go wrong or what would the impact be of option A versus option B, and that probably goes not just for finance function, but procurement and HR and IT, and actually all the different functions within an organization that have to be at the table to make decisions almost on a real-time basis.
Achille Ettorre:
Absolutely.
Anne-Marie Henson:
And maybe talk to us a little bit on the impact of governance on AI and just expanding on maybe some key questions that an audit committee or directors, board members, should be asking of their leadership team.
Achille Ettorre:
That's a tough one. I think if I look at the principles of the questions that should be asked, you're looking at how the actual data management is handled, you need to understand the ethical considerations of that. More than that, I think you need to understand the regulatory compliance, and I don't think there's enough of that in a lot of experiences that I've had. And then the risk management component. And I think those four pillars need to work together and almost formulate, I don't want to say a new group. Because I think the groups exist or the functions exist in certain parts, but they have to work together. And working together, I think they'll be able to noodle out such things of what misuse of the actual model would be or the actual function would be, what biases there would be, or quite frankly what errors there are because the process wasn't thoroughly thought through.
And maybe you're not in a situation where you have an enterprise ERP system such as SAP, you have something that doesn't stand up as strong because you've just grown too quickly and haven't been able to implement something as strong as robust as that. And so in saying that with those four pillars, you need to understand where you sit in the maturity model within your organization on how you use data and how you use analytics and how you implement AI. And as you understand where you stand, you obviously work on moving yourself up to being more mature, but not everyone's going to be an Amazon.
Not everyone's going to be a Google. Not everyone's going to work like Loblaw or Walmart. There are smaller companies out there, and I'll say like my family business, there's a few of us out there that work together and we can sit down around the table pretty quickly. And I think given my experience, we're probably not the normal company, but I can tell you quite quickly, I mean not understanding regulatory compliance and what the impact would be on the reporting structure for your HST filing is not something that will go... You can't just let that go for months or a year or two. You have to really stay on top of it or it'll handcuff you and quite frankly, even destroy good businesses.
Anne-Marie Henson:
Right. Thanks. And I actually really appreciate the fact that you're giving examples of very, very large companies, but also your family business. I think it helps put these things in perspective to understand where companies stand and the responsibilities they have. At the end of the day, compliance doesn't care how large you are, so you do have to make sure that you meet those regulations. Otherwise, you could be out of business really quickly. With the speed at which AI has been growing, I know you talked about it being imperfect today, but it will change a lot over this next generation for sure. But what do you foresee happening with regards to AI in the next say five years?
Achille Ettorre:
I think hands down will be a part of all of our lives. If I were to say within the audit industry, here's what I would see. I would suggest that a lot of the routine tasks would be automated so that there's oversight and a review versus actually compiling the data that will be automated. And that function, I think would resonate whether it's in an ERP system or your own, you can build your own. I think about fraud. Whenever I think about audit, I think about fraud. And I think that the tool chest that the audit and advisory folks, industry folks would definitely have a lot more in their toolkit to detect fraud and parse it out and maybe look at it from every different angle just because of the sheer speed and amount of data that's out there and other data that you can layer in just to make your predictive analysis or what you think may be happening in your current situation.
I also think... And I don't know, this is a stretch for me, Anne-Marie, I'm going to say this. I don't necessarily think auditing once a year or once a quarter is... I think it's right. I think it's good, but I think AI can probably automate that yearly audit to kind of say, Hey, these things should be audited monthly per se, or weekly per se, and you review them for that month or that week. And you can kind of continuously do that. So when you're in a situation where you're doing the audit for the year, I'm not sure what it would look like. I have an idea on what it looks like today, but in the future, I think it would be a lot easier for the folks in the industry given the tool sets that they would be provided.
Anne-Marie Henson:
Yeah. No, and it's actually... I don't think it's a stretch at all. I've heard that before, this concept of continuous auditing, which when we had a lot of changes with regards to Sarbanes-Oxley and controls and things like that, there was an expectation that that was going to bring in a lot more continuous auditing. I'm not sure if that's exactly the case, but it certainly seems like with the tools that we have today with AI, that we'd be able to do a lot more auditing throughout the year and then come in at year-end and do top up procedures rather than a whole audit for the past 12 months. So no, it'd be interesting to see where that goes. We've talked a lot about where AI could be beneficial or very useful to organizations. Is there anywhere that you've seen or that you think that AI would actually... Shouldn't be utilized?
Achille Ettorre:
It’s a controversial topic here, because I'm definitely pro AI, but I do think for instances where there's an emotional impact or sensitive type of data or sensitive type of decision that needs to be made. You definitely want to use AI to help you make that decision. But I wouldn't solely cut my workforce by 10% using AI. And I'm sure there are companies out there that have done this. I would suggest to everyone that I think you can definitely use AI to kind of help with that decision if needed. But I do think that it requires human oversight and maybe not just one human, probably a few. I think when you get into the kind of strategic planning or thinking about the culture and the direction of the company, I'm sure that you can have conversations with the new ChatGPT and have some interesting conversation. But I do think that that still resides or should reside with the human touch.
I definitely think AI can help out in augmenting or setting up the framework, but I do think that should resonate within the leaders of the organization. And then in areas where you think there's high bias, where maybe it's the model that's not created correctly or maybe you know that the model will say something, but given the experience of the individuals that are working within the current situation, just no differently, and this could be a health discussion where you're going down a path where you may need to work on or need to get some help with a particular disease. You definitely want AI to help you if it can, but you want the human expertise there to kind of look at it and make sure that it's making the right decision for the particular individual.
Anne-Marie Henson:
Right. It's good to know that humans aren't yet completely obsolete, and there are some areas where we're still really critical to that decision making or communication process. You mentioned bias, and this is actually quite a passion of mine as it relates to AI. I am very curious about how we can ensure to remove or at least significantly reduce bias in AI and in this predictive analytics. So can you talk a little bit about how you think someone or an organization could develop an AI product that is free or at least almost free of bias?
Achille Ettorre:
That's a very challenging task, but I definitely think you need diverse data. And what I mean by diverse data, you don't just want your internal data, you'd want external data to kind of validate whatever it is that you're trying to prove out. I think you need to have... Quite obviously, I guess you need to have some bias detection measures in place just to ensure that there's no bias that's creeping into that particular model in any such way. When I say bringing the audit people to the table, I think you need to bring everyone in DevOps to the table so that you're transparent in what your algorithm actually does. And this doesn't mean you actually going through all the math, but you're basically charting through in a decision tree, if you need to, when this input happens, these are the different criteria’s and then this spits out and this is what happens in this situation.
And I really think you need to really spend the time in understanding what all the ins and outs are so the process that's undertaken is definitely not susceptible to bias in the given situation of the model. I also think that we talk about... And I talk about this quite a bit, people, process, technology. On the people front, I think you need diverse people or an inclusive group of people. And so you don't necessarily need the entire team to be able to write the code and algorithms. I definitely think you need a few of them and it'll definitely help. But you do need people from all different facets.
Whether it's actual customers, potential customers, different employees from different teams, some outside counsel or panel, some consultants to review and vet that whatever it is that you're doing does pass a sniff test and wouldn't have a bias. And then once you're done, then you definitely have to monitor it. And monitoring is continuous, right? So if I use my friends at Disney, Mark Shafer's team does a couple million forecasts a day on capacity. They want to check to see that what it is that they're spitting out is actually right, so that when people go into book their Disney cruise or their time at Disney, it resonates with the offering that it is for that particular individual or that family wherever they are in the world.
Anne-Marie Henson:
Right. No, thanks for sharing that. And I think the continuous monitoring is something that's so important because you could make all the efforts at the beginning as you're setting up a process and building a strategy around an AI tool that you're going to be using. But not forgetting the fact that you have to check the output every once in a while, once you've implemented it, is something that's really key. Well, I guess we've talked a lot about AI and how it can impact organizations and a little bit about how it impacts auditing as well, but I'd like to know a little bit more from you. Where do you see your future in the next couple of years?
Achille Ettorre:
That's interesting. I mean, I'm fortunate that I'm able to do a lot of different things and things that I want to work on. Since COVID, I've put a lot of thought or something resonated with me with doing some work with, I'll say some of my friends at Harvard. And Karim who runs the analytics practice at Harvard or D^3 at Harvard, the D^3 Institute, his famous quote is, "AI won't replace humans, but humans with AI will replace humans without AI." And that kind of really resonated with me. Because I would tell you that I think about not only myself, but I think about - I have three kids. I think about how it is that I can best set them up for success down the road. And I've taken a lot of my experiences, both successful ones and not so successful ones, and I'm pretty close to finishing up and publishing a book, which is kind of one of those, I'll say bucket list things for me.
Once I write that book and publish the book, I'm going to utilize that to kind of do different things and help out different organizations set up or stand up their analytics practice or their AI practice. Or quite frankly, just come in on a particular project to see if I can help a lending hand. And it will be at a corporate level, but I'm also looking to help out people that are in school and graduating from high school or an undergraduate capacity because I think it's fun in both aspects. The corporate one is something that I find very intriguing because I get to learn a lot, but I learned just as much as dealing with some of the younger folks in this world, and I find both fulfilling. So I don't know where that's going to lend me. Hopefully we can meet again, Anne-Marie and we can do a touch base, but that's what's in stock for Achille in the years to come.
Anne-Marie Henson:
Fantastic. Well, congratulations. I'm going to be really looking forward to hearing about the release of your book, and we'd absolutely love to have you back anytime. So it was really great exchanging with you and I learned a lot, so I hope our listeners did as well. Achille, I'd like to thank you for your valuable time and your input today. I hope our audience appreciated the discussion. I'd also like to thank our listeners for tuning in today and to all of our episodes. I'm Anne-Marie Henson, and this has been BDOs Accounting for the Future. Please let us know if you found the topic interesting and useful, and remember to subscribe if you liked it. We'll see you next time.
Narrator:
Thank you for listening to BDO Canada's Accounting for the Future. Past episodes and related insights are available at www.bdo.ca/accountingforthefuture. Or you can go to Apple Podcasts, Spotify, or Google Podcasts to subscribe. For more information on BDO Canada, visit bdo.ca.