ALVIN: Hello, and welcome to (Re)thinking Insurance. I'm your host, Alvin. Thanks for joining us for today's conversation, the Role of AI in Financial Reporting-- Hype or Reality? Artificial intelligence has become one of the most talked about forces shaping the way we work, from data analytics and model building to how we interpret and communicate financial results. But beyond the buzzwords, how is AI actually changing financial reporting today, and where might it take us next? To help unpack that, I'm joined by two fantastic guests who bring both to technology and consulting perspectives.
Mark Brown is our global proposition leader for life financial modeling. He's based out of the UK and leads WTW's innovation in embedding AI across our reporting ecosystem, including within RiskAgility Financial Modeler or RiskAgility FM for short. Esther Huang is our Singapore market leader who works closely with insurers across Southeast Asia to modernize actuarial and financial reporting processes to unlock the practical benefits of emerging AI capabilities. Together, we'll explore how AI is reshaping financial reporting workflows, what's real versus hype, how consulting practices are evolving, and what insurers can do to kick start their AI journey.
But before we dive in, let me set the scene with a question. Someone recently described AI as an overly enthusiastic junior analyst, blazingly fast, very confident, and occasionally completely wrong. So Mark and Esther, does that resonate with you? And have you seen any memorable moments where AI has gone completely off the rails? Mark, I'll start with you.
MARK BROWN: Thank you, Alvin. Yes. I'm reminded, oh, a little while back when we were first building RiskAgility Financial Modeler's AI assistant as one of the very first AI-based products within our organization, the company was very keen to show it off and arranged a show-and-tell with around 600 executives. And we showed them very nicely how this prototype would do model design, model coding, model debugging, model explanations, and they asked whether it was safe. So we showed them the built-in controls that we'd been working on.
We asked the AI a non-work question, expecting it to refuse to do it. So I asked it, would it please draw me a picture of a Christmas tree. I was hopeful that it would pop up on the screen. I'm sorry, this isn't a work-related activity. You can't do that. But no, it proceeded to try and draw a picture. Not only did it not obey its instructions, it also failed to draw a Christmas tree. And I'm left explaining why on my screen it's drawn me a picture of a Victorian vampire coffin.
ALVIN: That's hilarious. Oh my god. And so how did that meeting end up?
MARK BROWN: With some apologies, a journey back to the developers to tighten the controls, and a product that we launched to great success.
ALVIN: Fantastic. I'm glad you guys caught that in testing. Esther, what about you?
ESTHER HUANG: Yes. Thanks, Alvin. I actually also had an interesting encounter with AI. So I’m a parent of a two-year-old toddler who refuses to sleep early. So I've actually tried to ask AI how I can get my toddler to rest early. And interestingly, AI advised me to explain calmly to him about how important sleep is for his long-term emotional regulation and to even ask him for some views about how he feels about bedtime. So in that context, AI has become my overly realistic friend, maybe one that doesn't have kids yet.
ALVIN: I'm sure every parent listening just laughed, because that advice works just as well as negotiating with the tiny, irrational CEO. That's a great way to kick us off. Now, let's zoom out and look at the bigger picture with our first question. So there's a lot of buzz around AI transforming how financial reporting is done. From your perspectives in product development and consulting, where are you genuinely seeing AI make a difference today? And where is it more of still a future promise? Mark?
MARK BROWN: We've seen a broad range of AI over time. We started off with the linear models that have been around for many years. We've seen that grow. We've seen that turn into language models, from there into reasoning models and into agents. And all of these are helping in very different ways. What we are seeing is the evolution and use of this technology, is seeing compound improvements in efficiency of teams of around 30% each year. At the moment, we're seeing assistants that can make humans more efficient, either using software features or as actuarial accelerators that can undertake one-off tasks.
A common example, for example, might be translating open source models, and these components can be wrapped together to undertake larger projects, such as a model documenter. We've built one of those that takes actuarial code and turns it into natural language documentation. And where we're doing that, where we're wrapping these into larger projects, we're seeing around a 75% saving in the effort on those one-off projects. But the future now, the recent technology, is agentic AI, and that's capable of undertaking big projects rather than individual tasks, and no longer working as an assistant, but as an autonomous member of the team.
ALVIN: Wow. Thanks, Mark. I love that point about automated documentation. Kind of reminds me of a joke I once heard. Developers have two natural predators, bugs and documentation, and AI seems to be taking care of at least one of them. Esther, what trends are you seeing with clients?
ESTHER HUANG: We've actually seen AI genuinely making a difference in the insurance industry, but AI seems to be applied in still somewhat very narrow areas at the moment. Many of our clients have actually taken a rather serious posture towards AI, particularly in tasks such as chatbots for customer service, data validation, document parsing, financial report preparation, and trend identification. Traditionally, such tasks are carried out by humans, but companies are actually exploring the use of AI to carry out these tasks in a more systematic and structured manner, and thereby freeing people's time to actually do more value-adding analytical work. But that said, we have not seen AI replacing seasoned insurance professionals to make judgment and not in the area of relationship management services as well. So I believe in the insurance industry, there are still many roles that still require communication, trust, and definitely human touch.
ALVIN: That's a great perspective, Esther. Now, let's shift gears to another area that's top of mind for many insurers, and that is that financial reporting is a highly regulated function where accuracy really matters. So, Mark, how do we balance the creative potential of generative AI with the strict control and audit requirements of financial and actuarial reporting?
MARK BROWN: Great question. We're lucky to be focused in a domain that doesn't really process individual data and doesn't have a direct social impact, and that limits the impact of specific AI regulations on what we can do. But nonetheless, our domain is highly regulated in other areas. Reporting today is still constrained by guidance from regulators and from the actuarial bodies, that AI should not be used to produce results unchecked. Ultimately, the humans are still accountable. The human is the control and governance layer, and the AI has to be an assisting tool for them.
So with that in mind, we can ask the AI to do pre-review tasks, like coding, to do precursor tasks, like high-level sense checks on calculations, and to ask it to draft content for a human to complete, such as a report. But we can also ask the AI to improve the quality, both of yourself and of itself. You can ask it to help you write better prompts. You can ask it to check its own output, or to review the outcomes of a process and suggest process improvements. The secret here is to interact with them as if you were training up a new member of your team.
But most important on all of this is your corporate governance. Engage with them, and don't fear them. That team will help you understand and manage the risks associated with AI, but will also help you access the latest tools within your organization to get the best value.
ALVIN: That's a really thoughtful view on the governance side of things. Thanks, Mark. Now, Esther, let's talk about the people side of this. As AI takes on more of the data preparation, the formatting, and even the first draft of narratives, how do human roles evolve? Is this about replacement, augmentation, or something in between?
ESTHER HUANG: So I do think that human roles will definitely evolve towards judgment, accountability, and making ethical decisions. We do not merely produce financial reports, but we are truly the owners of important corporate decisions. We get to decide what truly matters, much of which will be based on authentic experience that we bring. This is something that AI would not be able to replace, and I believe that humans would need to upskill and learn how to work with AI in the same way as how humans have learned to work with various tools, like spreadsheets and PowerPoint.
Well, the challenge is in using AI is really thinking too small. So instead of asking AI how to clean data, perhaps ask AI what it can do with your data, and ask AI how your financial reporting can improve, and consider some of the wildest ideas that AI could bring. We probably may not need better tools here, but we certainly need better imagination.
ALVIN: And what changes do you see in how financial reporting teams will be structured and trained in the future?
ESTHER HUANG: I believe financial reporting teams are likely to see a change of organizational structure in the coming years, possibly like a top-down, top-heavy, and bottom-light structure. Traditional production work is likely to be very compressed with the use of AI and traditional automation tools, but there are actually likely to be fewer roles that are focused on pure mechanical production of numbers. Graduate hiring may slow down. Entry roles may shift quickly from producing financial reports to reviewing financial results. And interestingly, as AI helps with the preparation of financial results, I believe regulators are likely to expect senior management to take greater accountability, to make sure that the numbers do make sense and are well interpreted. We have seen some large organizations downsize their financial reporting teams due to the adoption of AI, but yet equip their smaller staff base with AI training.
ALVIN: Thanks, Esther. Really great insights here. Perhaps we can bring this to life with some real examples. Mark, how are you already seeing AI improve financial reporting and practice?
MARK BROWN: We've seen a lot happening here. Preparation work is already significantly using AI, and by that I mean the work that goes into the reporting cycle. As an example, RiskAgility Financial Modeler has an AI assistant. This is a co-pilot that can explain an actuarial concept, suggest how it will be coded in the model, write the code for each of the various variables, debug the code, and produce the unit tests. And equally, I mentioned the model documenter. That can take an actuarial model, convert it into documentation within a day, and then perhaps a week on top of that for a human review. And then using that documentation in an agentic AI, you can explain the model to new joiners, to the team, to managers, or to auditors. Reporting today, though, is still constrained by the guidance from the regulators and the actuarial bodies about not producing unchecked results. But we've certainly integrated AI activities into our unified automation workflows, such as high-level reviews of results and drafting of the initial analyses, leaving the humans off the critical path, but not out of the loop. And with the learning power of agentic AI, there's a lot more exciting improvements we expect to be coming here soon.
ALVIN: Very cool. Esther, any examples you'd highlight from your projects?
ESTHER HUANG: We have some companies use a combination of WTW's workflow automation tools and AI to achieve faster month-end financial closing, so manual reconciliation reports and business planning forecasting are also delivered on time, with fewer surprises. For financial reporting teams that adopt AI and automation, they seem to no longer work overnight or on weekends to satisfy regulatory reporting deadlines. But that said, I think potential gaps or weaknesses, or even aspects that require management override may become more visible earlier rather than later.
In terms of our projects, we actually go beyond financial reporting. At WTW, we have actually worked on other interesting client projects in insurance business operations. So we have used our workflow automation software to call upon the AI tools that our clients have chosen in order to enhance their existing customer service, underwriting, and claims management processes. There have been a wide range of applications of AI that have proven to be very beneficial.
ALVIN: What strikes me from both of your examples is just how quickly these capabilities are moving from experimentation to day-to-day impact. So with that in mind, let's look a little further out. Esther, starting with you, what excites you most about AI this year?
ESTHER HUANG: Well for this coming year, I'm excited to see how agentic AI can augment the current way that financial reporting teams interact with their IT systems. Agentic AI is capable of allowing non-technical people to access technical capabilities much more easily. Updating dashboards can be done in a calmer and consistent manner, and CXOs will be able to receive real-time responses to their what-if questions. It would be a dream come true if financial reporting can be less laborious and much more agile and dynamic with the introduction of agentic AI.
ALVIN: And what about you, Mark? What excites you most about the future? What do you see coming after that?
MARK BROWN: For me, the goal is greater scope and autonomy, less human interaction, and hence greater value coming from the AI. But that challenges the governance constraints when AI today still feels a lot like magic. Making AI outcomes explainable to humans is one potential step to resolve this. One research area here is neurosymbolic AI, a combination of recognizing patterns and applying strict rules, which would then make the AI more robust and more transparent.
It's very hard here to have a long roadmap when the technology moves so fast. Each year we see new options that are bigger and better than we'd imagined with the previous incarnation. What is clear is that RiskAgility FM and Unify both intend to remain leaders in AI innovation for actuarial modeling.
ALVIN: Thanks, Mark. Listening to both of you today, what's clear is that the pace of innovation isn't slowing down, and that organizations that embrace AI thoughtfully are the ones already seeing real, tangible value. So as we wrap up, it's clear that AI and financial reporting have moved well beyond theory. It's becoming a practical partner that can enhance speed, consistency, and insight when it's used responsibly.
A huge thank you to both our guests, Mark and Esther, for sharing your perspectives on how technology and consulting are coming together to create the next generation of financial reporting. And thank you to our listeners for tuning in. If you'd like to learn more about how WTW is using advanced AI to complement RiskAgility FM, Unify, and other actuarial tools, please check out the resources linked in the episode notes.
Until next time, I'm Alvin and this has been AI in Financial Reporting-- Hype or Reality? Thanks for listening to another episode of (Re)thinking Insurance.
SPEAKER: Thank you for joining us for this WTW podcast featuring the latest perspectives on the intersection of people, capital, and risk. For more information, visit the Insights section of wtwco.com. This podcast is for general discussion and/or information only, is not intended to be relied upon, and action based on or in connection with anything contained herein should not be taken without first obtaining specific advice from a suitably qualified professional.
