240. AI in Finance: From Complex Workflows to Scalable AI Solutions with Tariq Munir

240. AI in Finance: From Complex Workflows to Scalable AI Solutions with Tariq Munir

Hannah Munro 11:04

So, hello everybody, and welcome to this episode of CFO. 4.0 as usual. I'm your host, Hannah Munro, and with me today is Tariq Munir, who is the managing partner of Finspyr. So, welcome Tariq. Thank you for joining me on the show.

Tariq Munir 11:19

Thank you, Hannah, for having me on the show. Really excited about it,

Hannah Munro 11:22

and we've got a lot to talk about today, but before we jump into sort of the detail, it would be great for our audience. Understand, you know your background in finance transformation. Where did you, I guess, learn your trade, as it were?

Tariq Munir 11:40

Yeah. So, like many finance people out there, I started off with their typical big four. I started off with PwC, and then, you know, from there, I moved on to the industry. I've worked a couple of years in exonobel, which is a global speciality chemicals company. It was a fortune five. It is a fortune 500 as well. And then I moved into PepsiCo somewhere around 12 years ago. So, over past 12 years, I've worked in different areas in finance, or in fact, over past like 20 years of my career, I have worked like from audit to controls, reporting, FPNA, So, pretty much all across finance and whenever, when you are working across So, many roles, you can very well realise that there is So, much potential that is there to improve a lot of finance processes, a lot of the ways we do things. So, throughout my career, which is something which has always been the front and centre of how I would work. So, even in the roles where I was not officially, officially a transformation lead or something, I would still go in and try to transform that role. Then my law, I did around towards, last five to seven years, I have been moving from finance into more integrated planning, supply chain finance. And then my last role at PepsiCo was the finance transformation. This finance, digital transformation lead for APAC, and that is where I was. I had the mandate to actually do the transformation as well. But that does not mean that that was the only time I was doing transformation. I was pretty much doing the transformation and improving our ways of working throughout my career, and I believe that is something which resonates with me a lot. And I'm sure today's finance professionals can also acknowledge the fact that we can truly do better than what we are doing today, in terms of our manual processes, still relying on multitude of Excel files systems not talking to each other. So, this is all frustration, frustrating, but yes, out of that frustration comes out an opportunity to transform and an opportunity to do things in a better way. And that is what effectively pushed me towards that as well.

Hannah Munro 13:51

And I think that's a really important to note, because I go into a lot of companies, obviously, when you're having conversations about change, and they say, Well, I've never done anything like this before I was like, and I think it's, I think the reality is, is that transformation should be a part of our day to day working life in finance, because that, you know, there's no, there's no process I've ever seen that is perfect because, you know, six months later, think changes within the business, and then they it needs to change. So, So, why do you think there's, I guess, a mindset around finance transformation is a particular role, rather than necessarily being incorporated into the day faith. Because, as you said you lived and breathed it in your day-to-day role.

Tariq Munir 14:32

I believe one of the biggest reason is that we are managing a packed agenda, right finance people, anywhere I go, anyone I talked to, including myself and I was, I was working in corporate, I would just not have enough time to actually step back and look at what I am doing. I would just be busy in, you know, somehow get them, somehow get the month and done. Somehow get this budgeting season done. I will look at my files once I come back So, stepping back. And then getting that time to actually improve things. That was one of that I do believe is one of the reason I do believe that, again, you know, finance people are smart people, where we are like, I mean, the way we are trained and the way we have been working. I mean, we have our pulse in the entire business, pulse of the entire business. We know pretty much what is happening across the business. We have good commercial acumen as well. We have good understanding of the business also. But I believe it is just that the way some organisations are set up as well, and the way our whole organisational culture, or the way or the functional cultures are set up that, okay, you know, I mean, this is finance, this is how I'm doing some specific work, and this is how I need to do transformation. As you rightly pointed out, is a mindset, right? It is not something that is going to be done by a transformation team. So, for example, in in in my role, when I was doing transformation, I was just one person. I didn't have a team, reporting interview. All I was doing; I was working with different people across the organisation to help them transform their ways of working. At the end of the day, it is the transformation needs to come from the inside, right? That okay, you know, this is what we really, really need to and I always encourage everyone who comes to me and they say, oh, I want to grow my career in transformation, because it's, it's, it's a happening thing. Everyone wants to get into digital transformation, AI and whatnot. I always tell them, it’s not about you are getting a role in a transformation, or whether, I mean your company has a transformation function as well or not. It is all about how you improve your current phase of working. It's a journey. It's a mindset change. It will take some time. I do see people getting there. People are now challenging and questioning things. With all these AI and all these different technologies coming up, people are questioning thing, but I do believe a lot goes back to our ways of working, how we have been doing things in a certain way, and then just not getting enough time to actually transform. And then the good thing is that once you do it, it creates a virtuous cycle. Once you transform something. Once you save up time from something, you get more time to do something better, and then when you do that, you get even more time to do more things better, because you keep on improving things that becomes part of your DNA. And when that happens, it creates a Virtu cycle of productivity for you. So, it's not easy, it's not simple, one plus one. But this is how I this is one of the reasons I would say that, that that probably need a bit more, a bit more mindset change.

Hannah Munro 17:48

Yeah. I call it the transformation snowball. Once you get it rolling down the hill, it starts to grow and have more impact that successful first small win to kind of build the momentum and the buy in from the team once you've got that. I won't say it's plain sailing from there, but it's a lot easier once you.

Tariq Munir 18:09

it is, it is. I mean, you get some time back, you start funding the transformation internally, right through your time, through your productivity, through your through your improved ways of working.

Hannah Munro 18:20

And when you look back at your time in PepsiCo as finance transformation lead, obviously you mentioned that you're working alongside the team. So, there's obviously a challenge there, because without you're not directly, got any direct reports within those teams, you're almost influencing from the side. So, you know, how can we influence and support people to have a transformation mindset. And how did you approach that? PepsiCo,

Tariq Munir 18:46

So, I believe the biggest, the biggest barrier, first of all, to transformation. And again, I mean, if we step back a little bit, what is transformation, transformation is nothing but a change, right? So, effectively, and I always, always say this, that effectively today, we are all change managers. Yes, with So, much happening around us, we are all change managers, and we need to go back to the basics of change management. We need to make sure that we are we are communicating a clear picture of what the future is going to look like, what it means to transform. And that is how I would, I would influence as well, in my influence, my broader teams to sort of sell them the transformation, to sort of show them that okay, if they make these changes, this is how their work is going to get impacted this much time they are going to get back. This is the value addition that they can do. This is how things are going to improve for their teams. They are. They are So, for example, in one of the roles and I mean, sorry, one of the projects, transformation projects, which, ironically, I was not transformation, man. Major at that I was not leading the transformation, but it was in a different role. Was around accelerating the book closure, right? Apparently, it seemed like a no, like a something which cannot be done, or, you know, which will take a lot of our time. People initially thought that it means that we are now going to cram five days’ worth of work in two days, our life is going to get more miserable, etc. But once you very clearly identify that, okay, this is what the future state look like. This is what the vision of the future is. You paint a very clear picture of the impact it is going to create. How we are going to improve our performance reviews, how we are going to create visibility of our reporting function across the organisation. Because now we are not just someone sitting at the back end doing just the number crunching for four days. And you know, everyone just coming to us and asking about, what is the number? What is the number? We have it ready on day two or day three, we accelerate the things people start to understand, okay, this is the benefit. What is in it for and again, it's, it's all basics, right? What is in it for them? We have been, you know, learning about this, influencing 101 and or influencing 101, finance, business, partnership, 101, class that you know, what is in it for them, right? So, So, that needs to come out very clearly the moment we start, and especially when you are in a role. Now I'll come back to the point around when I was in the in the role as a transformation lead, when we, when I'm approaching a problem with a mindset or give an impression that this is something which is coming down as a top down mandate, take it or leave it right. And that happens. That happens in a lot of scenarios where transformation is just a mandate, and you just have to do it. That is where it doesn't work. You need to make sure that transformation there is a mandate, of course, which is a great thing to have, because mandate means you have an author. I have an authority to actually drive things. But at the end of the day, people need to be able, should be able to drive the transformation from the bottom up, because that is where the real execution is going to happen. So, it is important. It was very important for me, for example, whenever any such project that came up, which was, of course, sort of like, you know, maybe a global project, or a transformation we needed to do, or things needed to move in a certain way, a process needed to change in a certain way, we need to, I needed to demonstrate to the team how, how and why they need to do again, you know, the basics of why they need to do it. What is that future, how the future looks like, and why does it even matter for them to change? What is the cost of inaction? And most of the times, I would just let the Team Drive it, let them instead of me dictating, I would work as a facilitator for them to connect with them, with the resources that are needed, provide them with the right frameworks, with the right mindset, a separate sort of us of an external consultation or advisory kind of a thing to them, support to them, and they would own The transformation. They would own the results; they would own the credit. They will get the credit. They will own the results, and they will get the credit as well, and then that will be showcased as well. That, you know, yes, this is something a team in Thailand did, or this is something team in ANC did, that's how it created momentum. That how it helped me create momentum.

Hannah Munro 23:41

And I think that's a really good insight, because there's a really interesting stat actually came out from one of the big consultancy firms that said that if you outsource, you know, a lot of your transformation efforts, you don't achieve, achieve your objectives, because there is that whole buy in piece from the finance team that's critical to, I guess, not just delivering the change, but ensuring that it stays. Because, again, you know, you look at all of the change methodologies, it's, it's now become less linear and more of a lean. Yeah,

Unknown Speaker 24:16

absolutely. So,

Hannah Munro 24:17

when you're, you know, when you think about, obviously, you're in a you were in a massive company. There's a lot of moving parts and perhaps some politics. How did you, I guess, clear the way for the team to actually create the space for transformation? How did you help find balance? Because resource is always a challenge when it comes to change and transformation and finding space. So, how did you support the teams to find the space for change?

Tariq Munir 24:45

Again, this comes down to if the team is really, really sold, between the why behind the why, they will find time and they will find the resources, and that is how it will. It. It has to work right again. We have certain hours in a day, right and at the end of the day, of course, we need to find we need to do transformation. Is a job which no one external can come in and do for us, even if we have an external consultant, for example, still the process, documentation and understanding diagnostics team, they will always be the time that team has to take out, but as long as we are making clear that, why that? Why do they need to transform and why does it matter to them? They will find time, and they would actually help you do that. And then, of course, you provide support through, though, making sure that if there are any decisions that need to be made, you're escalating them. I'm escalating them accordingly. I'm making sure that if there are any roadblocks, they're happening. And I guess the biggest one of the biggest elements of or an enabler of any transformation, is its governance, which often comes as an afterthought in many projects. Governance, help you not just stay on track. When I say governance, it doesn't mean that we are trying to audit something, because, again, as a as finance people or audit people, it's that's the first thing that comes to our mind that okay, governance, okay. It means we are going to stop. It's going to be a hindrance, as opposed to something which is of an enabler, but I believe it is enabler, not just from tracking the progress, but ensuring that. What are the projects that needs to be shut down, for example, what are the projects where we need to if we are looking at a portfolio of projects, if there are something which are lagging behind, why are they lagging behind? What support can we provide? Or, as I mentioned, is there a need to, actually, you know, consider shutting down those projects who need to take that call. Do we need to take that tough call and, you know, call it a failure and move on? All those things are enabled through governance. And that, I believe, is a big role that any transformation, any transformation organisation within a company can play, and that is how you enable it.

Hannah Munro 27:07

So, what does good in your opinion? What does good governance look like, you know, from a practical perspective, and what are some of the do's and don'ts around?

Tariq Munir 27:17

that for a good governance first of all, the most important thing is to clearly identify roles and responsibilities. Be very clear on who is accountable for what, who is responsible, who is who needs to be consulted, who needs to be informed that? So, I have found racy matrix, a lot of change managers. They would, they would, or project manager know about it. It's called ran ici, responsibility, accountability, consultation and inform information informed So, very clearly define that who is going to be responsible for what? Who is going to be accountable? Make sure that, for example, there are not more than one person responsible for one task, for example. Because the more we disperse responsibility, the more it becomes it hinders the progress. Because then no one takes accountability, no one takes really, really the response, as I say, that if it's everyone's responsibility, it's no one's responsibility, right? So, So, that's, that's, it's, it's very important to make sure that we are clear in that we should be able to identify that. Okay, if the Pro, I know it's a hard thing to say, but if the project does not work, who is responsible for that? Yeah, right, yes, it sounds a little bit policing kind of thing, but it's important, because otherwise we will not be able to identify who is going to going to be accountable and responsible. Then having a very clear reporting mechanism in place where we are tracking progress, where we are making sure that not just financial measures like return on investment, which we are good at that, right? I mean finance people we are good at that. We can right away, calculate ROI. We can make ROI right away. But it's important to track the non-financial measures, things like forecast accuracy, sorry, things like employee engagement or organisational health scores, how much people are spending time on how much time people are saving now, for example, how much? What is the net promoter score of the of the different tools and technologies that we are implementing, things like that? And then, of course, non a bit of operational as well, like your speed to market and how your how your operational matrices are getting impacted. So, this is what I call a balanced scorecard view. So, having that balanced scorecard view of your transformation journey is critical for a good governance mechanism to have a good governance mechanism in place. So, So, this is, this is something that I. I would, I would say, are some of the most critical or most important factors of any, any, any governance, and it works as an enabler. It and I will, I will again, like I will iterate this, this thing that governance is, governance might not be the only reason our transformation will succeed, but it can very well be a reason it can fail, because the lack of governance can result in massive disasters. So, treat this as an as an as an enabler, and show you one more point this, which is around the having the right escalations, So, having the right steer goes, having the right governance structure, what we call the pure governance structure, around who is who has the responsibility to So,, for example, if there is an if in a certain scenario, we have decision making responsibilities, There should be, for example, 70 to 80% of decisions being made within a certain within the teams, right within that project teams, then only 10 to 20% going up to the steer co then only 5% may be going up to the apex body, whatever you call it. We can call it digital Council, or board of Innovation Board, or whatever we call it. So, we have need to have that hierarchical structure, but we need to make sure that we are also empowering people to make decisions, and not just every decision goes all the way up to the up to the up to the Innovation Board, So, it can very well act as your enabler, right? So, that's, that's, that's some of the, I think, I hope I'm able to give some idea around, around what the good what good governance, look like. And maybe that one as well,

Hannah Munro 31:45

I think. And I think your last point was an incredibly important one about the number of decisions and that should be made at almost the grassroots level. So, I think you said 70 to 80% and I, and I completely agree with that. I also think it's I liked your piece around the layers, but maybe just for because obviously in our audience, we have a mix of big corporates, but also smaller businesses. So, if you're a team of, say, five to 10 people, you've got a CFO and FC, you've not necessarily got the big multi, multi jurisdiction, corporate level. How you know, what are the key things that you would empower the team with, and what are the top-level decisions that you would make sure that you have oversight of a CFO,

Tariq Munir 32:36

now, So, So, So, So, Hannah, that's, that's, that's a great question. And when it comes to governance or the oversight, I do believe it is, it is size, agnostic, geography agnostic. You don't just need a CFO for the decision making. Yes, CFO is one of the persons who is going to be highly involved in this one and most problem and in all likelihood be leading all of this. That is where I think CFOs role really comes into play, where her thinking around governance and risk management comes into play, how a CFO can influence the rest of the C suite to bring them on board and build these, these, these tier core structures. So, for example, in any given steer co you must have So, for example, if you are looking at your demand planning, revamp, transformation of your demand planning. Imagine you are trying to, you know, create a bit of predictive analytics for your demand planning. It is not just CFO, who is involved in this project. Demand planning would mean your supply chain directors, or your supply chain executive are involved in that. Your sales directors are involved in that; your IT team is definitely going to be to be to have a big role to play. So, your governance structure will simply be, irrespective of your size, will simply be your project teams. On top of that is sitting your CFO, your CIO, your sales directors, whoever is the executive within those because, again, governance is more toned from the top, as opposed to just something that will happen at a at an execution level. It's very important for that impression to come from the top management that, yes, we care about governance, and we want to make sure that if that we succeed in this project. So, we are here to help you clear all the roadblocks and support you in these projects. And we are not just here to come to just to progress tracking or budget tracking or something like that. Yeah, that can happen on an email as well. Right? That's not something which for which we need a committee to, or we need we need a steel. Go to do that. So, irrespective of the size of the organisation, we always need to have a multi, multi-functional, cross functional team of executives who are sponsoring the projects, who are making sure that there will be sponsors of the project, and then there will be the people who are the key stakeholders in the outcome of that, of that project, or the outcome of that, that transformation. So, irrespective of your size, I do believe we can always create this, this, this governance structure.

Hannah Munro 35:30

And we've talked a lot about what good governance looks like when you think back through your work history, where have you seen examples of bad governance, and how did you deal with the challenges that that posed.

Tariq Munir 35:43

So, I guess the biggest reason any governance would go bad is, again, comes down to that point around the accountability part. So, for example, it and it happens that we have a project right, or we have a transformation agenda that we are trying to run, and then there is just not any clarity around who is going to do what, who is responsible for what output, and who is going to be eventually accountable for the project, who is The Project Sponsor, when these, all these different things do not play well together and not clearly defined, that is a sign of a bad governance and that that really, really impacts the process, the output of your transformation agenda. Whenever there is, there is an there is, it happens that you are trying to get an output So, trying to get a progress tracking from the team, people would not know whose responsibility is it to report, whose responsibility is it to So, everyone becomes a little bit of, not a little bit, actually, a lot hands off that. Okay, you know, this is not my responsibility, So, I'm not going to talk about this. So, that is, that is one element, then the other element is not having a reporting mechanism in place. Now, again, this sound very, I would say, unglamorous, I know. But unfortunately, in transformation, the biggest impact happens by doing most of the unglamorous work. It does look like, it does look like, you know, transformation, you know, I might be, you know, working with AI and, you know, God knows what. But under the hood, there is a lot of unglamorous process, understanding, and, you know, workflow, understanding documentation that goes in to actually work that, and governance is one of those pieces as well. It's, it's, it's unglamorous. It doesn't look like; you know that something is happening. Its people see that. I see it as a hindrance. But having a reporting mechanism in place is very important, because without that, we will not be able to know where we are, what, where are we actually putting our resources? Or, in fact, are we putting right resources for the right projects? So, it's important to have a proper balanced scorecard view, and then, plus, you know, a portfolio, I always advocate the portfolio review, that we should have a brief, one page portfolio review, that this is the project, simple, green, red lights where we are, and then we drill down into different projects, key areas that Okay. These are the KPIs which are missing. These are the objectives we defined at the beginning with these are these are we are not able to do? This is something which, which, which is lacking. Again, I do see a lot of companies, a lot of work that I do. The first thing we do that that is always there is a steer CO, right? So, there is always a steer co present, right? I mean, no matter what project, there will always be someone. Because somehow people just assume that if there's a steer CO, it is getting governed. Yes, that is an important piece, but the input and output of steer CO is even more important. Lot of times, tier co in my in my experience, I have seen steer goes becoming more of a good conversation where you are. Just have you come in, you listen to a few people, a decision gets bounced off around the room. The high, the hypo effect, which is like highest, highest paid officers, in fact, comes into play. A lot of people who have a bigger voice. They then, you know, drive the agenda of the of the entire meeting, people then start getting into that group thinking, and all those issues around group thinking comes into play. This is something we really want to avoid, and for that to happen, we have to be very objective with us with us with our inputs, our outputs of the end, what is the expectation? What is. The terms of reference of a steer CO as well. So, yeah, that's something I would say is has been, has been my experience as well.

Hannah Munro 40:09

Yeah. And I think that communication channels are So, important, people like you say it's not the glamorous part sending an email out just to inform everybody about what's happening. Wave up to you know what's going on in the project. But it's, it's So, important to make sure that the wider business is aware of what's going on. And you mentioned key, you know, if you're working across a portfolio of product of projects, this is what we should look at. So, say you're, you know, you're, I don't know, project manager, or, you know, internal project manager for a finance transformation project internally, what sort of things would you report on a regular basis to keep the wider business and your project team on board? What does good reporting for an individual project look like?

Tariq Munir 40:57

So, it depends on from a project to project. But of course, if it is, we are talking about, say, for example, we are, if we are talking about doing an AI implementation, for example, right? So, a good scorecard would look, would have, of course, your, as I mentioned, you know, I mean, we are good at financial piece, So, I will not touch that a lot. I mean, your ROI is and everything. But yes, there is, and there is a point on ROI, and I will come back to that in a bit. Those, those financial measures are going to be there. But then, for a typical AI project, we need to make sure that we are also looking at the before the model accuracy, both in the development and in the production. So, because we want to make sure that when we are developing the model and it is showing a high level of accuracy. But when in production, it is not showing a high level of product of accuracy. This might mean there is something issue with the over training, or under training, or something like that, right? So, we want to get to that root cause. Also, of course, we will do the financial tracking, as I mentioned, around what is the budget that we have already spent and how we are going to do that. Another important piece is around making sure that we are also reporting on once the implementation is happening, once the project is once the AI is being used. How are people using it? What are the adoption rates? What are the, as I mentioned, the net promoter scores, what is the overall engagement of the team? Because, through use of those different tools and technologies that we are we are building because, again, if we just look at one thing, which is project is implemented, AI is implemented, we are good done with that. But if a person is imagining an AI model is giving us say, 90% accuracy, right? Great. Everyone would love that. But if my team is still sitting in the office, staying back because they need to now do five extra steps to get the AI output to be more accurate than what it was, I mean, my team will be still frustrated. So, would I call it a win or a failure? Right? So, it's very important to understand the pulse of the business, pulse of your team as well. That how what is the organisational engagement? What is the team engagement that is happening? And we miss that part. And I must say, I mean, this is it's not easy, but we though it's not easy as well. But still, we, not a lot of people, take an effort to actually go down that road and build that scorecard. Now, from our typical AI project point of view, since I gave that example, it is also the financial becomes even more important, a little bit more important than your conventional projects is because AI has a cost. AI is not a kind of a project where you just implement, and you forget you implement a project the first cut would likely to be 60% accurate. Say, hypothetically speaking, you will retrain the model, and then it becomes 65% accurate. You retrain it again, 70% it starts. People start using it, all those retraining, all those usages have a cost associated with it. So, the throughout the life cycle of an AI project, you will have a cost. Your cost will keep on increasing. So, your return on investment will keep on changing. So, how do you capture that? That is a very important piece. That is, that is, if, in if in the financial in my years, it's very important that we are, we are we are tackling that. And then, apart from that, I mean, depending upon, of course, your project, if it's said, for example, demand forecasting project, you will look at your forecast accuracies your preferred supplier scores. For example, because the better you predict your demand, the better you are able to supply your product as well. Again, it's important to have all these measures identified at the beginning of the project, at the beginning of your transformation, when we are. Building those roadmaps, and we are building those, and I always recommend that doing it in a certain systematic way. But again, it's, it's there is there is no one size fits all. But some of the, some of the things that I would say will always be there. Are the ones that I have mentioned. They've in one form or the other, they will be there.

Hannah Munro 45:22

Yeah, I think that's a really, you made a point that I always, I always think is really key is to focus on the outcome that you're going to be delivering. And I call it shiny toy syndrome. There can be a tendency, especially with technology projects, to focus on we've got the technology in, it's being used and actually, what I love about what you were saying there is that no the output, it doesn't matter if, if the AI is being used, if, like you say, the finance team are having to do more work to get the accurate results out in the first place. So, I think that's a really important point for people to note that it's about the outcome on the business, not just the true the outcome of the of the implementation or the project itself.

Tariq Munir 46:59

positioning around, around my thinking, my positioning around my transformation work is that any initiative that we are taking, any work that we are taking, we are doing, must be tied to a specific business need. Yeah, if it is not, it is should not be part of your roadmap. It should not be part of even you should not be even considering it, unless it is, it is solving a business problem. It is creating an avenue of growth, creating a better way of working. Yes, go for it. If not, you're just doing it for the sake of doing it. Because, you know, competitors are doing it, because everyone is doing it. There is something wrong. We need to really think, step back and think about it, that we really need to do it.

Hannah Munro 47:40

And you mentioned the ROI you know, you need to think about it slightly different with AI and versus traditional technology project. Are there any other aspects of traditional project or programme management that you see as needing a little something extra when it comes to AI projects?

Tariq Munir 48:00

Oh, absolutely, absolutely. AI needs to be created in a very different way. In our typical scenario, when we are implementing an ERP, even, or implementing any software tool, what happens is that we have a certain we pretty much there is, like a business requirement phase, where, as a finance person, I'll get involved, I'll tell the IT team or the or the vendor, that these are my business requirements. They will go away. They will, you know, develop the product, come back, implement it, do a UAT, UAT, user acceptance testing, or different forms of testing. We'll say, yes, go. No, go. Okay, let's go ahead. We are fine with abroad, with the with the software in AI, it's an entirely different if we want to get AI right. My argument, his is that, and this is what I have found working is involving SMEs. And that is where, in fact, finance becomes So, So, critical and So, So, important. Involving SMEs from the start of the AI development life cycle. And

Hannah Munro 49:05

just to clarify there SMEs for those perhaps not used to consult consultant terminology, subject matter experts, right?

Tariq Munir 49:13

100% yes, yes. Subject Matter Expert, finance domain experts, needs to be involved in the AI development life cycle throughout, sorry, throughout the AI development life cycle. So, I will not go into the details of the entire development life cycle, but it involves things like starting off with your with your business requirement, understanding, of course, problem definition, data, preprocessing, model training, every single step requires finance teams or the subject matter expert teams, domain experts to be involved, because only then you are able to create and build an AI model which is according to what we want the output to be. Then even once you do that, you. So, in fact, I will, I will, just to clarify a little bit more, I'll give an example of the training part of the of the AI development lifecycle. On the face of it, it might seem that, okay, model training is a very technical piece. What does a finance person can do? Why would we need to get involved? We need to get involved to make sure that the training data that the vendors are using, the training data that the development team is using, is representative of the population that we are trying to build a model for. Because if that is not happening, you will train a model which will perform great on training data, but as soon as it will be exposed to the unseen data, the new data, it will, it will, its performance will go down. And we do not want because that, because, number one, there is time involved. There is a lot of cost involved in the in training of an AI model. And we want to make sure that we are tagging with the technical teams to make sure that this is happening. And that's where, of course, you know, building digital literacy, AI literacy, all that comes into play as well. And then once, even when you have done your AI implementation per se or deployment, that's not the end of the end of it. That is where we call the AI flywheel, or the AI, the virtual cycle of AI. AI comes into play. You implement a first version of AI. It will create people will start using it. It will generate data. That data will then it will then use that data to get trained. So, your model will theoretically improve its performance, because now, as more and more people use it, more and more data gets generated, and more and more training happens. So, if theoretically, it will keep on improving its output as and when people are using the data more, and people are using that model more and more. But of course, you need to keep track of that also. You need to make sure that, as I mentioned earlier, that you know, there is a cost involved with retraining. We need to make sure that the objectives that we set at the beginning, that these are the forecast accuracies, for example, model accuracies, or different OKRs, or different objectives that we set at the beginning. Are we meeting those or not? So, it's, it's, by far, it's not a linear, linear process at all. It is very agile. We need to embrace an agile mindset. We need to embrace the minimum viable product failing fast mindset if we want to really work with AI, because AI has an ambiguity associated with it, and we should be able to embrace that ambiguity. So,

Hannah Munro 52:38

what? And again, you know what I heard from that conversation, I think there's a couple of critical points where you need to make sure, especially for finance specific AIS, that you have that involvement. So, it sounds like in the definition, making sure you're really clear at that beginning stage about what you're defining, So, that we've got, and we've got measures to track that all the way through, as we go into that, making sure that the subject matter experts are involved in the training phase, particularly around validating the data before you start training, to minimise the cost of that training aspect. And then I think the third thing you said is almost like a, there's like a, there's a, there's a recheck phase post that you know, the actual implementation or deployment piece, which is where you need to then go back to those measures that you've defined beautifully well in your initial phase to make sure that you're still delivering against that. Because of the danger with the retraining, I guess, and more data coming through, that it might go slightly off course, potentially.

Tariq Munir 53:38

100% and even Hannah. It happens that your again. I mean, no one predicted covid, for example, no one predicted supply chain disruptions, for example, in 2022 23 So, I mean, a lot of this happens suddenly, and your model needs an entire retraining. You cannot stick, get stuck with that old model. So, you have to come back, and you as a subject matter expert, need to bring out and say that, okay, you know, gas prices have increased by 100% or electricity prices have now increased, or commodities are going in this direction. The network costs have increased in this, direct in this, in this way. How does that impact your model assumptions? A lot will change. So, it's an iterative process. And

Hannah Munro 54:25

there's an interesting piece there, isn't it? So, we, we live and work in an incredibly volatile environment, right? And we're designing AI yet. So, how do we build in, I guess, the capacity to identify what a one-off volatility versus something that is needs to be assumed or taken into account. So, you know, are those the sorts of things that we need to think about with the data that we're giving AI to work with? To be clear on what is, I guess, normal, or, at the moment, perceived normal, I don't think. Any of us thought normal, as we're in there, but how do we, I guess, the ability to keep cope with volatility into this, into these learning models.

Tariq Munir 55:11

So, again, from an outlier point of view, which is where, again, the subject matter expertise will come into play. So, your treasury team will know that, okay, these are the things which, these are the cash payments, or these are the payments that happened because of an M A transaction, and these, not these, does not need to become part of my normal cash flow, overall cash flow, or Treasury forecasts. And that is, again, where I am a big, big advocate of this, that subject matter experts will be the ones who will be able to identify, number one, that, what are these true outliers? And then again, it's, it's, it's, like, anybody's guess that if that is an outlier, it is going to become sometime I mean, I mean, if you have like, two data points, you can draw a line So, that can be a trend, right? So, again, I mean, it's, it's anyone's judgement that it is, it is a one off. There will, of course, always be one off. But again, as finance team, or as any team, for that matter, who is working on AI, there is always this level of ambiguity and uncertainty that we need to need to be comfortable with. Any a model in the world will not give us a 9590 900% accurate results, right? I mean, they are there. I mean, there will be models, of course, depending upon you know what the complexities are. But in a in a complex environment, if we are operating, we are not going to get a 90 900% accurate, accurate model, except for what a vendor demo shows us, right? But, but in reality, there will always be things which, which would be, would be, would be a new data. It would be a nuance. So, we will need to either call it out as a noise. And again, it's a very important point you have raised Anna. And when we are training a model, technically, when we train a model and there is a lot of noise in our data around these outliers, and we over train our model, that is where we again create a bit of an overfitting issue with the model that there that model is now actually reading the exact data pattern, as opposed to it's actually getting trained on data, as opposed to recognising the patterns. So, it is actually looking at what is the exact pattern of the data, as opposed to just looking at the general patterns and then performing as it as it gets exposed to new data. So, it's important to again, that's were subject matter expertise, that where involvement of your, your even your senior executive team, would come into play, because a lot of information in the market. They might have, your sales director might have a lot of information, which you might need that insight to then treat it as a whether this is your baseline or this is something which we need to overlay on top of our baseline model. So, again, no, no, I wouldn't say there is a there is a one exact answer for that that you know you can find. You know you can do these three things, and then you can identify outliers. It will always be a lot of judgement call in this one. But of course, you need to get involved with it, with your developers early. You need to stay on top of the model early.

Hannah Munro 58:32

and very aware that we're, we're, we're running rapidly out of time, just as we get into my favourite topic discuss, which is AI, which is always the way. So, yes, if I was maybe, you know, because I think there's been some great learnings, not just about AI, but also about general projects. So, my first question, my first of two questions to you. So, the first thing is, when you're looking at AI projects, you know, maybe you're a CFO, maybe, you know, or a financial transformation person that's been asked to get involved in an AI piece. What are some of the things that you should look for in you know, either an existing project plan or an existing brief that will tell you that that project you know is show will be successful, has a high likelihood of being successful. And I guess the reverse of that, which is, what should you be looking for before a project starts to see whether it's got a good chance of succeeding, particularly when it comes to

Tariq Munir 59:36

AI I would, I would always say, look at your look at your and no one would know better than yourself. Look at your process First understand when you are looking at, you’re in any AI solution to resolve to automate something, the process that you are trying to automate. How complex is that? Yeah, is it something? That involves a lot of decision complexity, where, you know, I mean, you are doing, you are raising a purchase, acquisition. It goes through 20 different steps, and then, you know, it goes back someone changes it. Then it goes to someone else who changes it. Is there a lot of manual intervention that is happening. Is there is a lot of judgement calls that are happening again. AI is smart, but it is dumb as well, right? Yeah. So, I mean, it depends on how you use it, or which process you are using it to automate. If you will expect that, if someone will expect that, you know, I'm I will just put an AI on my purchase to pay process, for example, and it will somehow magically transform it. That's not going to happen. Your complexity is going to be amplified with So, So, I can, I can pretty much assure you that the project is not going to work, right? I mean, if your if your underlying process is very complex, if it has a lot of handovers, and I also do not say that first simplify everything, and then start looking at AI, yes, of course, that is also not practical, right? We need to move fast. One hand, we are saying we need to Agile, and on the other, we need to say, simplify everything, but go in in bit of a modular way, look out your project and look out your processes end to end, and identify areas which are really complex, which needs a bit of streamlining. Identify areas where you can actually just simply use basic automation to make them even smarter or more streamlined or optimised. And then look at those areas which you do believe have some kind of judgement involved, which have a certain pattern, which have a lot of data, requires a lot of data analysis and something that AI can help you do, yes on chat, GPT, it seems like it can do everything, but trust me, it cannot. It will break. So, it will not work if our processes are too complex. So, I would always, always, always start with, what is the workflow again, at the end of the day, what, no matter what we do, an organisation or a function, is a sum of different workflows. And when I say workflow, sending an email is a workflow as well, right? Sending a message on Zoom, for example, which is a kind of a workflow as well, as long as of course, it's not just a joke and it's related to your work, right? But this is this, is this how complex your workflows are will determine whether your AI project is going to succeed or not, and you can tell that right away if your projects or if your workflows are too complex, too dated, you have multiple systems not talking to each other, no, AI in the world is going to come and solve it for you. You have to do again, I'm really sorry to say, but the real output comes in that unglamorous work, yeah, to break it to you, but I mean, that's where the real work needs to happen. We have to roll up our roll up our sleeves and get into the grunt of it, and then, you know, or simplify our processes. Use basic automation to simplify the workflows and then use AI. That's where the value will be amplified otherwise, AI is already So, complex that it is hard to explain its output. Yes, right? Today, if you are doing Imagine you are inputting a complex process into something which is already very complex, and then you are making decisions based on an output of which you have no idea how that gets generated, yeah, because it will set overly complex. The complexity will be exponential. So, it is not just a breakage or a breakdown of a process. It will have real impact on your business, your decision making.

Hannah Munro 1:04:01

And I think that's a great point to finish on, is that the principles of, you know, implementing AI are very similar in some ways, to the principles of automation, right? If you've got disparate systems that aren't talking to each other, if you've got data sets that aren't clean and validated, if you've got processes that are overly manual and not been thought through with a lot of complexity, like you're going to spend a lot of, you know, potentially, a lot of money and a lot of time with maybe not the guaranteed success piece throughout the other side.

Tariq Munir 1:04:36

100% 100% Hannah and you have, you have, you have summarised it really well. A lot of my time actually goes in helping educating people in this space, right? And this is a great platform. We are, we are we are having this discussion on as well. Is that AI is not a rocket science. It's not, it's not something that is going to come and just somehow solve all you. Problems. We need to go back to the basics. We need to stick to the basics in order for us to move forward. So, yeah, greatly, very well summarised.

Hannah Munro 1:05:10

And obviously we're not just here, like I've been picking your brains for a good hour now, longer than a lot of our episodes, but I think that shows the validity of the content. But you've also got a book being released and potentially will have been released by the time this podcast comes out. So, you know, if people want to learn more about you know how to do transformation well, and you know, and you to learn from your experience. Tell us a little bit about that book and where they can find it.

Tariq Munir 1:05:40

Thank you So, much. Hannah, yeah, absolutely. That book is something I have written over last it has taken a good, good time. Over last two years, I have been, I have been working on that and my, that's my, I would say, a purpose and a passion project that was, it is getting released on September 9, or if by that time, the podcast is already out. So, it is available, of course, on Amazon. It is available on violet. Is available on a lot of different Barnes and Nobles. So, I mean, if someone will search reimagine finance by Tariq Munir, they will, they will find the book. Now, the premise of the book, the entire book is around and like the title of the book is reimagine finance and the subtitle is the CFOs leadership playbook to lead in the age of AI, data and digital So, effectively, the premise, or the through line of the entire book is that CFOs or finance leaders, being at having a unique vantage point In the organisation, are uniquely placed are in a in a in a position, in a in an advantageous position where they can actually lead the entire transformation, digital transformation for the organisation, whether it's demand planning, it's supply planning, or it's, you know, because any process that goes cuts across the goes through the organisations finance has a role to play in it, So, finance can lead that, and this book actually helps them with busting a lot of these myths that we were talking about that you know, what are different technologies? What are the real-world applications of those technologies? How AI, cloud computing and different process mining, digital twins are actually helping organisations transform their operations. How can finance do that? But where it becomes really powerful, is that, how do we re architect the finance operating model around data, digital hubs and talent for enable to enable us to lead the transformation, to enable us to wire the transformation. And then it goes into a there's a playbook, a five-step playbook, which I have built based on my experience and based on a lot of research that went in. And of course, it is built on a lot of lot of work that has already been done in this space as well. So, full credit to those, those thought leaders as well on how they can actually lead the transformation across their organisation, how CFOs can actually influence that. So, I do believe that this will help a lot of finance professionals, a lot of finance leaders and CFOs to really take their digital transformation to the next level.

Hannah Munro 1:08:19

Amazing. And so, for those that are interested, we will, of course, put the link to the book in the show notes for you to click on and go have a look. Thank you So, much, Tariq, for joining me today. It's been an absolute pleasure to talk to you.

Tariq Munir 1:08:33

Same here. Hannah, thank you So, much for having me. And it was, it was pleasure having this conversation.

Hannah Munro 1:08:39

Fantastic. And thank you to thank you to our listeners as well for joining us. If you have enjoyed this episode, please do not forget to share it, to leave us a review on your favourite podcast platform. As we help spread the word of the CFO 4.0 podcast, thanks for joining me, and we'll see you next time.