Workiva Amplify 2026 | The CFOs Blind Spot

Clips
More from Workiva Amplify 2026

Steve Soter

VP & Industry Principal

Workiva

play_circle_outline Executives Trust vs AI Reliability: 84% Would Trust AI-Only Annual Reports, 25% Report Errors Reaching Boards

play_circle_outline Workiva Knowledge and AI Guardrails: Defensible, Traceable, Governed, Integrated Governance for Trusted AI Adoption

play_circle_outline Human review remains critical for financial statements and disclosures

play_circle_outline CFOs must demand demonstrable trust and embedded proof before critical AI use

Info
Transcript

The CFOs Blind Spot

Steve Soter

VP & Industry Principal Workiva

In this interview from Amplify 2026, Steve Soter, vice president and industry principal at Workiva Inc., joins theCUBE's Krista Case and Alison Kosik to discuss the gap between executive confidence in AI and the governance needed to trust it. Soter, a former SEC reporting practitioner, unpacks a new Workiva survey showing 84% of executives are at least somewhat willing to let AI generate an annual report without human review, even though a quarter have already caught AI-driven errors reaching the board or external audiences. He attributes the disconnect to or... Read more

explore Keep Exploring

What did the survey ask, and what were its main findings? add

Is Workiva helping to solve the data-trust and governance problem around AI—specifically, will it provide the guardrails (traceability, provenance, governance, and integration) to bridge the gap between how much customers should trust their data and overly optimistic expectations when using AI for tasks like completing filings? add

Where in AI-enabled financial reporting processes is human review essential? add

What is the first question CFOs should ask before deploying AI in critical financial workflows? add

bolt Powered by CUBE AI

The CFOs Blind Spot

search

Alison Kosik

>> Welcome back to Workiva Amplify 2026. We're coming to you live from Las Vegas. I'm Alison Kosik alongside Krista— it's Krista Case. We are after lunch and the brain is still sort of going with it. We're about to bring in a guest who's going to talk about the risks that leaders may be overlooking and what it takes to use AI with confidence now that we've really got AI in the room.

Krista Case

>> Yeah, Alison, it's an important topic. I think there's a degree of kind of fear, uncertainty, and doubt around AI, but I definitely think that we are getting there in terms of developing some confidence and really looking forward to this guest.

Alison Kosik

>> Yeah. All right, let's get to it. I want to introduce Steve Soter. He's the Vice President and Industry Principal with Workiva. Welcome to theCUBE.

Steve Soter

>> Thank you so much.

Alison Kosik

>> Talk me through your role.

Steve Soter

>> So I'm a former practitioner, meaning that I used to have roles in accounting and financial reporting, SEC reporting in particular. Now I get to spend time doing things like we're doing here at Amplify, which is talking to customers, hearing about their problems, learning about the ways that we can solve those problems in ways that are going to build confidence. I love my job.

Alison Kosik

>> Oh, wow. That's amazing. I'm so glad you love your job. Yeah, it's all over your face. You look excited. I want to start with some research that Workiva has out, a new survey. Walk us through what that survey asked and what it found.

Steve Soter

>> Sure. So quite a few highlights. I would recommend people go check it out. Honestly, those could, I think, foster some very productive and probably thought-provoking conversations within their own organizations. We asked a lot about how companies are thinking about AI adoption, the way that they're communicating with investors and other stakeholders. The thing that stood out to me the most was a stat that actually surprised me significantly. It was that 84% of executives said that they were at least somewhat willing to trust AI to generate an annual report without human review. To me, that's astounding. What was also interesting, though, is that a quarter of those executives said that they had found instances where AI had managed to get an error out to either the board or to an external audience. So to me now, we see a massive gap between readiness and willingness versus the reality of what the tool is doing.

Krista Case

>> So Steve, that's concerning. Based on your role and how you're working with customers, why do you think executives are trusting AI more than they arguably should based on your research?

Steve Soter

>> I think because, number one, there's tremendous pressure. AI could present significant opportunities for organizations to move faster, to move more efficiently, to do better in a sense. And I do think that there is that optimism that executives are feeling. I also think that part of it is that maybe there is a bit of a disconnect between an understanding of what is actually happening behind the scenes. I don't mean that in a negative way. Executives have a lot to pay attention to. But I think when you really get down into the details about how the work is done, that disconnect might be part of what's causing, again, the willingness and maybe the aspiration versus the reality.

Krista Case

>> And so, Steve, I'll make a little plug here for Workiva from a product perspective. we heard at the keynote from your Chief Product Officer Deepak this morning a lot about, Workiva Knowledge and the insights that Workiva is bringing to the table. Based on, your context around the data that customers are using and based around the guardrails, that Workiva is also putting around how customers are potentially using this information to do things like complete filings. So, do you think that— is this kind of an issue that Workiva is helping to solve? Do you think Workiva is going to help to bridge that gap between, the reality in terms of how much customers should trust the data versus maybe the rose-colored glasses, if that makes sense?

Steve Soter

>> 100%. And the way that I would describe it is people have always trusted Workiva for, let's say, 4 things. It's defensible, meaning that when you have some kind of an output, you know how you came up with it. It is traceable, meaning that you know where the data came from. It's governed, meaning that it's subject to your internal controls and it's integrated, meaning that work is happening from different parts of the organization, all in a single platform. Now, that was before AI, but those 4 things become particularly important as you think about AI. And one of the things that I think is happening is that people are adopting solutions that aren't governed and built upon those 4 things that I mentioned that I think is going to be the difference between that 84% and the 26%, the 1 in 4 that we just talked about.

Alison Kosik

>> And this has been talked about. It's been a theme that I've heard about how important it is to show where the data come from. If a company is using AI without clear controls for validating and documenting the output, does that become a broader control problem rather than just an isolated AI mistake?

Steve Soter

>> Oh, 100%. Now, the severity of it is going to depend a little bit on the process. But let's use an analogy here. let's say you hired an intern or a staff accountant. They were really good at analysis. They were very persuasive, very confident in the outputs that they gave you. But somehow they were finding a way to get errors out again to your board or to external audience. Well, you're probably not going to want to really continue within that process with that individual. What you're going to want to do is to support what sounds like an otherwise great employee and start to put guardrails around that to give them the protections. I think to me what's happening is that organizations are maybe putting a little bit too much confidence in these off-the-shelf tools that can be tremendous productivity resources, again, without thinking about what needs to— the foundation that they need to be built on.

Alison Kosik

>> Yeah. Yeah.

Krista Case

>> So pulling on that thread a little bit further, does AI create new risk or is it more exposing potentially weak data, data controls that we already have in place?

Steve Soter

>> It's such a good question. To me, I think it's maybe a different flavor of the same risk. So when I think about it, back to the days when I was a controller, it was really important for me to know where the data was coming from, who touched it, what happened to it, how did it get reviewed and approved. And that also starts to speak to the other thing I needed to know, which was the process. How did it all come together? Did the right people look at it? Did everything happen sequentially in the right way? To me, that is really no different whether we're talking about an old manual process or whether we're talking about something that's generated by AI. To me, the difference is that inherently with AI, we're talking about less human touch. And so how do I get the right foundation and how do I get the humans involved in the right place? I don't know if we're talking about anything new here. I just think, again, it's a slightly different flavor because of the promise of the technology.

Krista Case

>> So is there really at the core of it any difference between providing assurance for an AI system which is more probabilistic versus a more traditional financial system, which is more deterministic? I'm hearing at its core they're very similar, but are there any differences that we should keep in mind?

Steve Soter

>> Well, I think that when you get back to what your core finance processes are going to look like, I don't think there are differences. Now, can AI do things that we never thought possible? Of course. Is that going to maybe challenge some of the boundaries? Of course. But to use an example internal control over financial reporting, We have those frameworks. Those frameworks have been out there for a long time. To me, AI doesn't change that. It actually makes it even more important because accelerating a process, if you don't have it grounded by those things that we discussed, those 4 things, then speed doesn't become an asset. It really becomes a liability. It becomes a risk. And that, I think, is the thing that we're really figuring out. How do I go from speed, which was always the goal, to speed that I can trust?

Krista Case

>> And I'm glad you brought that up, Steve, because I also wanted to kind of get your take on, as we do start integrating AI into these processes, where do we really need to make sure that human review is critical? To your point on we need to balance speed gains that we're getting from AI versus the risk that that might create.

Steve Soter

>> I think anything that is going to touch sensitive financial information, to get back to that example about something making it to the board or an external audience, does a human need to look at it every time?

Alison Kosik

>> 100%.

Steve Soter

>> An AI tool isn't signing off on the financial statements, a human is. And if that human signs off on it, but yet trusted that AI had, done everything that it was supposed to do and done it correctly, again, if that's not the case, that could be a big risk. And it's not just the existence of an error that could be the risk, it's even if you got to the right answer, but it was a bad process or a bad tool that was used, even that by itself could be a potential liability.

Alison Kosik

>> For boards and audit committees then, what questions should they be asking management right now about AI in financial reporting?

Steve Soter

>> Well, it's such a good question, and it actually gets back to something that you already brought up. Where did the data come from? What did the process look like? What assumptions were made? What review existed? It's all of those things. Because if an executive or an audit committee is asking those questions and they're not getting sufficient answers, and I don't mean just verbally, here's how it worked, but no, let me prove it to you. Let me show it to you. There's likely some risks in the process that they need to address.

Krista Case

>> Yes. Yeah. Yeah. And I was, just kind of looking at some of your research here. And I believe the stat is that only 11% of executives actually believe that their data quality was sufficient for AI use. Just to sort of underscore that point, can you talk a little bit about when you saw that? I guess maybe your reaction to that data point.

Steve Soter

>> Well, I'm so glad you brought that up because that was the other thing, like 11%. Yes, that is really, really low. And then it makes you wonder, how bad was the data before we were even having this AI conversation? And to me, that just underscores, honestly, the opportunity for AI, because I think AI actually has a role in potentially helping to clean that up, like maybe boosting that 11%, but AI is only as good as the data that it is using. And so if that's the baseline that you're starting at, these executives should not be surprised that errors are making their way out there. And it also tells me that there are probably significant manual processes that are happening in order to clean that up. And that could provide, I think, significant upskilling and upleveling opportunities for their team to be able to shift from spending their time on tying all that stuff out and getting it fixed and corrected to maybe be able to do higher-value type of work that will really help the business.

Krista Case

>> So I've been doing— it's a great point and I've been doing a lot of research around AI governance overall, and there's a number of different best practices, but there's really no standard framework, especially for disclosures, of risk. So, I guess if you were to put your, yourself in the shoes of an investor or a board member, I guess what are they looking for? Are they looking for just signs that the organization is being mindful of these potential pitfalls? Are they actually looking for, some proof that they've implemented certain pieces of technology? You know, can you comment on that a little bit?

Alison Kosik

>> Sure.

Steve Soter

>> I think the first thing is that, whether something is generated by AI or not, boards, investors and other stakeholders, they need to be able to trust the data that they're getting, whether that was a painstaking manual process or whether AI assisted in that. trust is like the bedrock of the economy. So I think that's a good place to start. I think the second thing, though, as I talked about, there's tremendous pressure from boards and from others for companies to be adopting AI. But they need to do that responsibly so we don't end up getting hit with these errors that we've already talked about. And I think what companies need to be doing is having a clear vision and a clear direction of travel for how they are going to do that. And to the extent that that's going to impact the material metrics, whether that's operating expenses, revenue, or what have you, they need to be talking about that because in some ways, while let's say poor AI is a liability that we've discussed, the right kind of AI adoption could present or could represent a significant opportunity for these companies. And that's probably something that they do want to talk about with certainly their board, but with investors and other stakeholders.

Alison Kosik

>> For CFOs that are moving quickly on AI right now, what do you think should be the first question they should be asking before putting it into a critical financial workflow?

Steve Soter

>> Well, the first thing is, can I trust it? And can you, whoever provided me this information, can you give me the proof? And is giving me the proof actually part of the process? Because here's the other thing: if I use AI, but then I end up having to painstakingly check every single output manually again and again and again, it's not a win. It didn't save me any time, and it actually probably would have been easier for me to just have done it myself. CFOs need to be driving that dialog and that discussion. Prove it to me and show me that that proof is actually inherently built into the process. It's back to the 4 things that we've discussed. And that's actually one of the reasons why I'm so optimistic about what we saw earlier today about, the way Workiva is approaching our AI innovations.

Alison Kosik

>> Yeah.

Krista Case

>> Yeah. And so there's definitely an economic tradeoff there, right, between, what the human is checking at the end of the day. And Steve, I'm interested. So as CFOs are kind of leaning on, some of these AI tools, when we think about a potential error, right, that an AI tool could make, from your perspective, what separates something that would be a more significant deficiency in the model and kind of maybe more of a systematic error versus maybe kind of a point issue that would have been, more difficult to detect or perhaps couldn't have been prevented? Is there anything in your mind that maybe separates those two?

Steve Soter

>> Well, speaking not just to the process itself, but actually speaking to the result. So let's say a number was wrong in a set of financial statements or some other disclosure was incorrect. You're going to want to assess that for materiality. And that can often help determine whether or not that is a simple deficiency, a significant deficiency, or— and, I hope this is not the case, but up to and including a material weakness. So that's the first thing. But I think the second thing is actually going back and saying, hey, at scale, what could have gone wrong? How big of an issue could this have caused? And that's one of the things that becomes a little bit scary about AI adoption en masse. Because if you were talking about highly scalable processes that are impacting multiple disclosures and multiple numbers, what I said just a minute ago actually scales really, really quickly. It could be that those errors make their way to where multiple things are wrong in multiple places. And now for sure, you'd be talking about a significant deficiency or a material weakness. So I think it's two parts. It's the magnitude and then the what could have gone wrong. How severe was it?

Alison Kosik

>> Yeah. What's one of your biggest takeaways you have so far out of Workiva Amplify?

Steve Soter

>> Well, it's how optimistic we are. And I'll shift gears just a little bit. We were here a year ago. Of course, we were talking about AI. Everybody was talking about AI. But I think at that point it was a little bit uncertain about what the direction of travel was. Is this real? Is this not? Are we all going to have a job in a year? what are we going to be doing? How is this going to work? I wondered if that was going to feel that way this year. But we announced some significant AI innovations that I think for me as a former practitioner are really exciting. But the other thing is seeing the optimism that practitioners are having, because that uncertainty that I saw a year ago, it feels different right now. They're seeing a clear path, at least a direction of travel. And while they may not know exactly how this is going to end up, like what the North Star is, they're starting to see real ways that AI can not just make you more productive, but actually transform the work that you're doing. for somebody who was in this work for a long time, that represents a significant opportunity. I don't want to be chasing numbers, tying out things, formatting spreadsheets. I want to be providing insight to the business. I want to be providing the story, the rationale behind it. I think we're starting to get to that point. People are seeing that, and that's what's got them so optimistic. It's the thing that has pleasantly surprised me most about being here.

Alison Kosik

>> All right. That's a great note to end on. Thanks so much for your time. Thanks for coming on theCUBE.

Steve Soter

>> Thank you both. It was a delight.

Alison Kosik

>> And you're watching theCUBE, the leader in live tech coverage and in-depth expert analysis. We'll be right back.