Josh Robinson, Vast Space | Workiva Amplify 2026

Welcome back to WAKA Amplify 2026. We're coming to you live from Las Vegas. I'm Allison Caic alongside Christa Case and we're about to get into a great conversation about how AI agents, they're moving into the workplace. They're doing work that humans used to do and it's kind of leaving the question, are we going to lose control here? >> Absolutely, Allison. It's I think something that's keeping a lot of people up at night. Um and I think when it comes to establishing that control um you know auditability um is something that's really important. I know we were talking about that in our keynote discussion um earlier this morning but being able to make sure that again we have that audit auditability especially given that you know agents change the game versus kind of when it's just humans taking action agents can operate much faster at much greater scale. And so I think that has a lot of ripple um a ripple effect on on their audit processes. >> Yeah. Yeah. All right. Well, let's get into it. I want to introduce Josh Robinson. He's the chief audit executive of Vaspace. Welcome to the cube. >> Hi, good morning. Thanks for having me. I'm really happy to be here. >> So, talk us through first what Vast Space does. >> Yeah. Yeah. Uh so, Vast >> and your role with them. >> Absolutely. So, um as you mentioned, I'm the head of internal audit at Vast Space. Uh it's an aerospace manufacturing company that's based in Long Beach, California. Um and we are uh trying to make human space travel relevant for the future. It you know the future in this area is changing. It's becoming more commercial and VAST is at the forefront of making humans go back into space. So it's a really really exciting industry to be in right now. >> Oh. So let's let's kind of start with the big picture here. AI agents as I said they're increasingly you know making decisions and taking actions that humans once did. I'm curious what worries you the most from an audit and from a risk perspective. >> Yeah. So there's a couple of things. I think first and foremost as you alluded to AI agents are now doing the work that humans really didn't want to do in the past. So I want to start with the positives in in so far that you now can have you can have tasks that were redundant, repetitive, mind-numbing or boring that a computer is now doing. So what I to answer your question what I like to think how I like to phrase this is mitigating the good with the bad. So there is a lot of positivity because we're re we're reallocating human brain power to something that makes more sense to to utilize that human brain power. Now from a risk perspective and as an audit executive as someone who's reporting to boards who's dealing with external audit firms what we're trying to understand is how do we understand the probabilistic nature of that AI and making sure that okay with humans we could understand they were using their judgment they were exercising probability but we could see it there was either handwritten notes or there were tick marks or or there was email chain or slack messages that has all been removed from the equation. So for me the risk is in the unknown what's happening underneath the covers that we need to look into more and the good news is particularly some of these agents the workflow capability the audit trails are becoming more and more robust and complex over time that allows us to look under the covers but to be clear that risk still exists and and I think the profession of full cander is still trying to understand how to mitigate it but like I said because of the ROI on the human factor um it's an acceptable risk and I think that's why companies are leaning leing into it so much. >> Yeah, absolutely, Josh. I'm hearing that as well from both, you know, security practitioners as well as risk practitioners that the need for enterprises to embrace AI has changed the whole appetite for risk for the business. But as you're alluding to, Josh, you know, we still need that observability and auditability. So if you take a step back and think maybe conceptually, how do you think about you know laying the groundwork to be able to have that level of assurance given the fact that with these AI agents we might not be able to have that kind of you know that that full observability and control over every action that they're taking? Yeah, that's a great question. And for me, I I like to just go back to foundational fundamental auditing. And one of the things that's been we've been harping on in our profession for a number of years now is this concept of completeness and accuracy. If you can demonstrate completeness and accuracy as an auditor generally that uh allows you to reach conclusions about the assurance of data sets or or audit tests you're doing. So there are ways and again we're still developing these in the profession to demonstrate completeness and accuracy around the work that agents are doing. So for instance, if an agent is a multi-step workflow, you're it is potential there is excuse me, there's possibility to chunk it out into let's say there's five different steps and validate the completeness and accuracy of each one of those steps along the chain that allows you to paint a bigger picture as to okay the five things that accomplished were indeed complete and accurate representations that allows us to reach a level of assurance. Similarly, there are the concepts of testing like LLMs to be like, okay, let's let's stress test this or sanity test it. And if we put in data and it spits out what we'd expect it to, we can then extrapolate that to a a more traditional audit test. So, and again, we're still testing the waters on these things. But I do think these are different ways for us to still reach that level of comfort. And of course it still takes time and effort but you're just you're reallocating it in such a way to still maximize how these agents are able to maximize these processes. Yeah, absolutely Josh. I think you know to your point you know definitely we're still early days and I think in navigating you know all of these challenges one thing so as you're kind of talking through this audit trail you know human auditors are already strapped for time right so how again conceptually how do you think about beginning to you know approach that problem can we maybe use AI to start you know making taking some actions on our behalf um and if so you know what what are you looking for in terms of you know trusting the AI to potentially take some action? >> Sure. Sure. Yeah. >> So I think so going to the root of the question auditors being strapped for time and this is a big answer but in my humble opinion you need to think about the entire entire audit life cycle and figure out where you can use technology to remove redundancy or time syncs. And so for for me and my current organization, what we talk about is how do we introduce AI into operational processes that then have that trickle down effect of making your audit processes more simple because the evidence is is easier to gather. The uh information around the IP the information provided by the entity is is validated for completeness and accuracy. That way auditors even if they're not using AI spend more time doing you know human you know brainpowered work rather than oh I got to tick and tie and put put evidence together. So that's one thing. Now in the audit life cycle itself and again still very much testing the waters on these things. But for me as an audit leader, what I'm what I will be doing when we implement these technologies is making sure that my auditors are still exercising judgment where it's necessary and not blindly relying on AI or aentic AI to solve problems for them. And then I think at least in these early years and I'll say years who knows how many years it'll be. It'll be having leadership do detailed reviews evaluations of audits work of junior auditor's work of Agentic AI's work to make sure that we would reach the same type of conclusion before going to that reasonable assurance model just solely via the the AI. So I it's I think it's a heavy-handed mix right now of getting the entire life cycle in a better place to make it more efficient, but then making sure humans are still inserting themselves appropriately until such time we get that level of comfort we'd expect. >> Digging a little deeper into the governance piece of this, one of the big questions is AI accountability. You know, if it makes the wrong decision, you know, who owns the data? Is it the technology team? Is it the business unit? Or is it the person who is kind of overseeing the process? Wow, that that's a really loaded question, Allison, but here here's the thing. As an audit leader, I I love to use the analogy. It's and I I say this to people and it feels so old school, but it's the old the buck stops here analogy. So Harry Truman, when he was president of the United States, had a big placard on his desk and he's like, "Decisions come to me. I own those decisions." That's what I tell my team. So yes, there there are going to be inputs whether it's a business user, whether it's an IT user, but if we're using data, I my team, you need to make sure it's right, it is complete, it's accurate, you've vetted it, and the decisions that we're making, the workpapers that we're initiing, you know, digitally initially these days, you stand behind that. So from my perspective, and again, there's probably a lot of conflicting thoughts about this. As an audit leader, if it comes to my desk and I'm going to put my signature on it, I I it pretty darn well better be the right answer. But and of course it that would be a collaboration. You know, we can still look to it like, "Hey, by the way, this didn't solve right." We can still look to the business. You applied this agent wrong or you didn't do diligence on the agent. That's okay, but I'm going to own making sure that happens. >> Wait, so is that does does the AI factor of this wind up creating more work for you because if if others who are creating this who are in the process in the process of creating whatever that document, they rely on AI too much. you've trusted them and didn't and you don't realize that they're relying on AI, not checking, you know, dotting the eyes and crossing the tees. Is this just because it's causing you to to look over everything more? >> I think that's Alison, I think that's a great question and I think it's a very astute observation. I think in the early days the answer to that is resoundingly yes. However, I humbly believe that internal auditors as a profession are very well suited to take on that level of effort because we've got the professional skepticism built in. We've got the the knowledge of companies at a very holistic level built in. We've got the relationships with people to help solve for that built in where we can do it. But, you know, at the end of the day, if that were to happen, if if I was to start billing a bunch of hours like, "Oh my gosh, we're we're telling people what they did wrong. We're double-checking. We're we're blowing budgets on audits." I think fundamentally comes down to how are organizations handling AI governance at like their like entity level making sure that's appropriate because again I think the knock-on effect of that would reduce the types of challenges you just said. So again very complex answer in today's like immature environment but I humbly believe internal auditors are up to that challenge that can make businesses better as we solve for that. Yeah, and kind of as I'm reflecting on what you were just um you know talking about Josh, I'm kind of thinking about the role of audit in terms of actually with time allowing the organization to move faster in terms of adopting AI. So, you know, would you agree with that in terms of making sure that those kind of guard rails and those audit capabilities are in place up front? It might take a little time to get it going, but again with time >> Yes. So couldn't agree with you anymore. And and I think well-governed organizations, AI governance has been top of mind for them for for years now. And to your point, and it's interesting, as you might imagine, everyone who's in an audit space has this battle. But good governance is a facilitator of good business. Whether that's business that goes with high velocity, whether that's business that treads with caution but makes the right decision, it doesn't matter. Good governance b is baked into that. There's a lot of leaders who I've dealt with over my career who would say that's not true. I I respectfully disagree with them, but exactly to your point, you bake in good governance, it facilitates success, even if it takes time, and it does take time, even if the the ROI is two years of of pain. I I do think you can have very meaningful conversations after that two-year period to say, look what we built together and how it's facilitating our strategic success. Absolutely. Yeah. >> Absolutely. And Josh to circle back, we were talking a little bit about kind of, you know, the context and and the knowledge that humans bring to the table. And you know, we heard at the keynote here, we're obviously here at Work Amplify. We heard at the keynote this morning about kind of work even knowledge and kind of what work is doing to be able to, you know, connect some of these different systems and applications um and maybe help to facilitate some of that contextual knowledge that AI will um will need. My understanding is that you've been aware customer for almost 10 years I think since 2017. Yeah. So I guess maybe if you could comment a little bit on um you know the role that you see potentially playing in that regard especially with announcements like knowledge. >> Yeah. So I have been a longtime customer and obviously advocate for for the platform. I'll tell you when I saw the keynote both yesterday and then it was continued this morning I I'm really excited by what they're bringing to the table and I'll let me tell you why before I get into work specifically you guys know this because you meet so many folks every technology firm around the planet is figuring out how they can in implement or you know it bring inhouse AI to help solution for create solutions for their customers however in my opinion I haven't seen any firm really try to step outside of that box and say, "Let me try to solve problems for you that are created by having disperate technology solutions with disperate AI solutions." And I think what work has demonstrated through their their announcements from yesterday and today is that they want to be a thought leader in that space. So what I'm really excited about is seeing how their technology can take disperate data inputs, it can interact with LLMs that companies are using, no matter which one it is, and then help create a gentic AI that solves business issues drawing from those various sources. So again, I haven't had a chance to use it yet, but I've had a chance to see it. And I think I'm actually genuinely excited because I think we're finally seeing a company that's going to help solve. I can go to my CFO and be like, okay, we use the likes of a Netswuite and an HS and this and this and and this is our data lake and you know we're using Claude, but now look at Worka who can connect to those five different things, bring it in and create agents simply that are is going to solve XYZ problem for you. I think he'd be just as excited as I am. So I think it's a really cool thing that they're they're introducing right now. >> Yeah. >> So streamlining certainly super important. >> Absolutely. Yeah. I think because look, yeah, the best way we want humans to to better utilize their brain power to make strategic decision-m and the best way to do that is to streamline these processes or functions that simply don't require that level of effort. And I think what we're demonstrating is we're getting there. We're we're we're finally getting there. So, yeah. >> So, Josh, I'm curious to kind of double click on so obviously VAST, essentially a commercial space station company like you were talking about earlier. We've been talking a lot about, you know, audits and I think there's obviously compliance implications there, but especially in an industry such as, you know, what you guys are in and what you guys are doing, there's safety concerns potentially. There's bigger risk beyond just complaints. So, can you talk to, you know, I guess some of those risks as a result of potentially control gaps, you know, in this process? >> Yeah. Yeah, that's a great point. So, you're exactly right. this industry that that VAS operates in is highly regulated and of course there are like and like many industries there are human lives at stake. So we take at our organization risk management incredibly seriously. I I think and one of the things is again I'm newer head of internal audit this organization is trying to bring together disperate risk leaders to create more holistic enterprise level risk assessments and risk management programs to help solve for what you're just saying. But to your point, at least and look, I can't speak for those risk leaders cuz they're not here. But I would say that particularly when it comes to safety, occupational safety, safety of astronauts that are eventually going to be flying, the company takes a more conservative approach. As you might imagine, we are trying to we're trying to win contracts. We're trying to get humans back into space, but when it comes to that aspect, it's very conservative. So AI, agentic AI or or LLM, it's part of the process to help us build, but to be clear, there are very much like let's let's hit pause and make sure we're doing the right things to keep our astronauts safe. So but to your question, I think enterprise level risk management programs help bring together all the different players to make sure we're addressing those things holistically instead of something getting missed or or you know overrelying on AI and then all of a sudden a safety concern does bubble up. >> Yeah. Yeah. >> Absolutely. And I think for all those reasons, but also Josh, I would imagine even in resilience for, you know, core business functions, you know, like you say, getting astronauts into space, especially as, you know, AI agents are, you know, taking action. So there's things like we need to make sure that we can reconstruct, you know, certain decisions that were made, you know, we make sure that we have the logs that are founded in them. So again, it even has kind of a resilience implication. At least that's what we're saying. >> No, I I think you're absolutely right. And I would say that at my organization, we're I think we're using AI tools to help that. Now, I wouldn't consider it agentic AI or probabilistic AI, but to to your point, the technology is so versatile, but but yes, absolutely. I do think that's helping us be more efficient in in keeping our astronauts and our our team who's building a space station safe. So, yeah, absolutely. >> Wonderful conversation, Josh. Thank you for stopping by the Cube. We appreciate it. >> Thank you. It's a pleasure to meet you guys. Thanks so much. >> And you're watching the Cube, the leader in live tech coverage and in-depth tech analysis. We'll be right back.