Generative AI and Finance – Transcript

October 24, 2023
Season 5, Episode 3

Vin Kumar:                  

What I think finance executives should be looking at in the next maybe three to five months is planning what the impact is going to be for finance. For that, we need to use education with the finance leadership team on what generative AI is and what it’s not.  

Start looking at what could be the size of the prize that they have to go after. What is the approach they should be taking? What do we need from an investment on who’s going to be tasked with looking at this? How are we going to be addressing this? What are sort of the tools that we should be looking out for that we can start implementing in a gradual manner from next year onward?

Announcer:                 

Welcome to The Hackett Group’s “Business Excelleration Podcast.” Week after week, you’ll hear from top experts on how to avoid obstacles, manage detours and celebrate milestones on the journey to world-class performance.

Tom Willman:              

What impact will generative AI have on the world of finance? How can finance leaders get started using artificial intelligence to improve their operations? On this week’s “Business Excelleration Podcast,” we’ll answer these questions and more as part of our ongoing series covering artificial intelligence and the enterprise.       

My name is Tom Willman. I’m a principal in the Hackett’s Finance Executive Advisory Program, and I’ll be talking today with Vin Kumar, principal and AI and Digital Operations practice leader. Welcome to the podcast, Vin.

Vin Kumar:                  

Thank you, Tom.

Tom Willman:              

AI is a technology solution we have been talking about for a while now. And until recently, it’s been something that companies have been exploring and piloting in the back office, but it doesn’t seem to have really taken off yet. But it does seem like we are at an inflection point with the introduction of generative AI. Vin, what do you see different between the AI that we’ve been talking about for the last couple of years and generative AI?

Vin Kumar:                  

Tom, that’s a great question, right? AI itself is not new, but what is the difference between generative AI and the AI that we’ve been talking [about] is really important for finance executives and enterprise executives to understand that.

Till today when we spoke about AI, what we were referring to was we were providing the machine instructions. It could be extremely complicated instructions. We would have written an algorithm. We would have used a data scientist to write an algorithm. And we would feed that algorithm to the machine to run it at extreme speed, and it can cross multiple sorts of data – significant data. It could analyze and give it to us. But fundamentally, it was us providing a set of instructions for the machine to do and give us input and assist us in our decision-making.        

Whereas, generative AI, we do not need to spell out every instruction to the machine, and the machine is able to make decisions on its own, and it’s able to make these decisions because it has gone through what we call training. It has been trained on a particular domain. It could be in an audit. It could be in legal. It could be in a regulatory filing. It’s been trained on that particular domain, so it’s got the ability to make its decisions on its own without us giving it all the instructions.    

That is what has fundamentally changed between generative AI and what we call … previously, we call it cognitive AI. And that’s the biggest difference between that. And what we see the impact of generative AI is going to accelerate the automation and give new skills and be a digital assistant to our finance professionals in making their decisions. So that’s the big impact that generative AI is going to have in finance.

Tom Willman:              

That’s really helpful, Vin. Thanks for clarifying that for us. Now, there’s a tremendous amount of hype around generative AI right now, and it reminds me a lot of the conversations we were having about robotic process automation five years ago. And while companies had some success with it, it didn’t really deliver the kinds of benefits that many expected it would.        

And I was reading an article the other day that recent data is showing that the most well-known generative AI solution, ChatGPT, is seeing its accuracy decline – usage statistics are starting to drop. Should we be tempering our expectations for generative AI, or do you see this being different than what we saw with RPA?

Vin Kumar:                  

A couple of things to look at. One – one of the drawbacks of the RPA and why it didn’t meet up to its expectations was the level of specificity which we have to define exactly what the machine had to do. Every action had to be scripted, and we had to have complicated decision trees to decide which set of actions – what was common knowledge for us as humans to process. We had to codify it and script it out for the RPA, and that took a lot of effort. And the benefit we were getting was not in a line with the effort it was taking to write that script. So that was the biggest gap to do it. Now some of that is going to be addressed, and we’ll talk a little maybe in our conversation about how generative AI is affecting the impact of RPA or is it accelerating that use of RPA using generative AI.       

But coming to your point, and what we are seeing in ChatGPT usage dropping, there is more hallucination happening that’s giving incorrect answers to do that. So the first advice we would give the CFOs and the finance organization is you should be using the enterprise version of generative AI solutions – not the public versions – precisely for these reasons. Where there is a deviation, there may be a quality issue on the responses you’re getting. There may be an increase in hallucination. We cannot control that when you’re using the public version of a generative AI solution like ChatGPT.  

We strongly recommend enterprises to use the enterprise version of generative AI solutions. So the one example of an enterprise version is Microsoft’s Azure OpenAI services. Another example is IBM’s watsonx.ai. These are enterprise versions of generative AI solutions that we should be using in an enterprise. Primarily for the ability to protect from a liability, from a risk management, from the deviation in the quality of the output you’re getting from the generative AI solutions, protect your own IP and your data. And that data is not used to train the generative AI solution that can be used outside of your enterprise. So, for these reasons, we believe that you should use enterprise versions.   

Now, coming to the hype cycle, yes, we are on the hype cycle, and we are going to hit the traditional life cycle of these type of technologies. But we do believe strongly that this is fundamentally going to change how we are going to be processing and how we are going to be servicing our enterprise from a finance organization, but it’s going to take some time. It’s not something we are going to see immediately significant changes. These domains-trained generative AI solutions are going to take some time to be delivered, and we think that within the next five years, we are going to see the significance of how this is going to impact and how it’s going to affect how finance operates.

Tom Willman:              

Excellent. Thanks, Vin. Good advice on public versus enterprise versions, and appreciate the perspective on the hype cycle and the promise for the future.           

I’ve talked with many CFOs and senior finance leaders about generative AI, and there’s no shortage of ideas or potential use cases being talked about in areas ranging from planning and forecasting, external reporting, investor relations, even transactional processes like collections. But how do you envision finance leveraging generative AI, and are you seeing companies actually designing or building anything yet?

Vin Kumar:                  

So the way we are seeing companies exploring this … I should say, Tom, that when we do see clients thinking about it, they’re all in the exploratory stages. As I said, the enterprise versions of the generative AI solutions have been released on a very limited basis in the industry. So it’s not accessible to every enterprise that wants to use it. So it’s available only on a limited version by the providers such as Microsoft or IBM to do that.    

Given that, we are seeing companies explore and coming up with proof of concepts predominantly in two domains. One is what we call embedded generative AI solutions. So there are solutions available today where they’re using it … a generative AI solution is embedded within other applications and that you could use to do that. For example, ServiceNow is coming up with an embedded solution in managing case summaries. There is Microsoft’s Copilot, which is going to be embedded in your Office products that you could use to summarize a PowerPoint or a slide into an email, or take an email and convert it into a PowerPoint slide or an Excel table to be summarized in an email, which these tools are available. We are seeing some of that happening.

The other area where I think you’re seeing more is in the what we call native generative AI use cases, and that’s predominantly being used for external research, summarizing of market data information, writing some quick scripts to do certain work. I have a client who used this as a proof of concept and pilot stage where they took and wrote a VBScript. A visual basic script was generated by the generative AI solution to go and update a particular spreadsheet table, rerun the pivot tables, come and update 30 presentations that they had to send it out to their clients – internal customers from the FP&A team. It could have taken maybe a couple of hours for the FP&A analyst to do it. That visual basic script was run. It took 30 minutes to get script run, pilot it, test it out and execute it, update it on the presentation, then email it. 

So it is possible these are all used, but mainly as a digital assistant for the unstructured work or knowledge work that we see that the finance teams are working on. So they are all being done in the POC in a pilot, not in a full-scale deployment and significant impact. That’s going to come when you start using domain specialized generative AI solutions – like in legal, like in audit, like in tax, like in regulatory filings or intercompany reconciliation of fixed-asset processing. That is still in stages where companies and solution providers are developing the solution.

Tom Willman:              

Excellent. That’s helpful, Vin. So we’re still very early on in terms of companies figuring out where this is going to fit – what role it’s going to play in the finance organization.    

So for those that are just getting started, where would you suggest finance leaders start and getting their arms around generative AI and the potential role it could play in his or her finance organization?

Vin Kumar:                  

So what I think clients and finance executives should be looking at in the next maybe three to five months or till the end of the year is planning what the impact is going to be for finance. So there are a lot of companies and finance leaders – [they] should be thinking of what is the impact of generative AI going to be on finance? For that we need to use education with the finance leadership team on what generative AI is and what it’s not. How do you separate from what’s there in the public versus how do you adopt it for the enterprise version?           

Start looking at what could be the size of the prize that they have to go after. What is the approach they should be taking? What do we need from an investment on who’s going to be tasked with looking at this? Is it going to be in the finance team? Is it going to be in your shared services team? Is it going to be part of your IT organization? How are we going to be addressing this? What sort of tools we should be looking out for that we can start implementing in maybe a more gradual manner from next year onwards? So that’s what I would advise the finance leadership team should be focusing on in the next three to five months.

Tom Willman:              

That’s helpful, Vin, thanks. And I suspect that to really be able to fully take advantage of something like generative AI, most companies probably still have a fair amount of foundational work to do, correct?

Vin Kumar:                  

Absolutely. I think there has to be some foundation work. Otherwise, you can’t just leapfrog across all your tech debt. Certain tech debt you can leapfrog, but the foundation needs to access … how do you access your data? Where are you going to train that data? Where are you going to host these solutions? What will you build and buy? Those foundation things needs to be in place.

Tom Willman:              

Yeah. This has been tremendously helpful for me as well, but I talk to clients every day that almost seem in a panic because they haven’t yet done anything with generative AI. But what I’m hearing you say is that given where generative AI stands­ – given the limited availability of the enterprise versions of the technology that we really recommend finance and other functional leaders use – they shouldn’t feel bad about where they are. But I think you’ve provided some very helpful and prescriptive guidance on where they should get started to get their arms around it.

Vin Kumar:                  

That’s right. And this is an accelerated learning exercise, understanding the impact and how to go about it. That’s what they need to focus on – not get too carried away that they don’t have generative AI implemented and deployed. That’s happening next year as more companies are planning for it.

Tom Willman:

OK. Fantastic. Well, thanks very much, Vin, for joining us today. And listeners, if you’d like to check out other episodes in our artificial intelligence series, we’ll put links to those in the show notes and on our website. Thank you very much for listening.

Vin Kumar:                  

Thanks, Tom.

Announcer:                 

Thanks for listening. You can find the audio, helpful resources and a transcript of each episode at podcasts.thehackettgroup.com. If you liked this episode, please share it. You can also subscribe at Apple Podcasts or your favorite listening app, so you never miss an episode. We’d welcome your feedback by tapping the rating on this or any episode, or send us an email at podcast@thehackettgroup.com.

The Hackett Group is a global leader in defining and enabling world-class performance. Learn how we can assist with your improvement journey at www.thehackettgroup.com.