AI hype to real impact: A strategic approach to Finance AI with Cox Enterprises

Hey, welcome everyone. Thanks for joining us today for our webinar, AI Hype to Real Impact, a Strategic Approach to Finance AI with Cox Enterprises. I'm Tiffany Mott, Director of Product Marketing for AI and Operational Analytics here at OneStream Software, and I'm joined with Inger Hunter, Director of Innovation and Advisory for Financial Solutions Delivery at Cox Enterprises.

AI is everywhere right now, as we all know, and for many organizations, the real challenge really isn't access to tools. It's knowing where to start. It's knowing how to scale and how to actually drive value.

Today's session is about cutting through all that noise. So again, we're joined by Inger Hunter, who led a thoughtful, strategy-first approach to AI adoption, focusing not on chasing tools, but aligning AI to real business problems in finance.

What makes this story compelling to me is that it's not theoretical. It's grounded in real use cases, real challenges, and real outcomes. So before we dive in, let's level set on a common pattern that we're seeing here within the market in AI.

We're seeing that teams are experimenting with multiple AI tools. There are a lot of disconnected pilots across functions, and there's a lot of excitement, but limited measurable impact. And so what Cox did is they took a different path here.

Instead of asking what AI tools should we use, they asked, where can AI create the most meaningful impact in our business, and how do we get there intentionally? And that's what we're going to impact today.

So Inger, thank you so much for joining us today. Do you mind introducing yourself to the audience and providing an overview of Cox Enterprises? Absolutely. And thanks, Tiffany. And hello, everyone. I'm Inger Hunter.

And as Tiffany mentioned, I serve as Director of Innovation and Advisory within Financial Solutions Delivery here at Cox. My role centers on finance enablement and transformation, really helping to connect strategy, process, data, and technology so our teams can really move faster, stay disciplined, and make better decisions.

Just for a bit of context, because I'm not really sure how many of you are familiar with Cox, we are a privately held company. It's actually fourth generation family owned, and originally started by James M. Cox.

Today, we operate in 15 countries, and we have over 500 employees, and we generate about $23 billion in revenue, really big number. As a family owned business, we are really known for some of our flagships, which includes Cox Communications, which some of you may recognize as one of the largest private telecom companies in the U.S.

And to rival that is Cox Automotive, which some of you may also know is one of the world's largest automotive services and technology providers. Beyond that, we also have a growing portfolio of businesses and have invested in more than $2 billion in sustainable businesses and technology.

That long-term innovation-focused mindset, that's really the core of who Cox is. In my current role, I work closely across finance and our technology partners to really modernize how we plan, how we execute, and how we deliver.

We're not just here to implement tools, but really we're about building repeatable and scalable capabilities within finance. And because Cox spans such a diverse portfolio that includes multiple businesses, multiple systems, and data environments, we're always very intentional about scaling what works and really avoiding fragmented and one-off solutions.

So, Tiff, that is a little bit about me, as well as a little bit about the great organization that I work for. Tiff, that's incredibly helpful context, Ingrid. Thank you. I mean, Cox must be really doing something right.

Being a privately held, family-owned company that's grown into this size and scale with such a diverse portfolio really is no small feat. And it really speaks to the discipline and long-term thinking that you mentioned.

So, let's start from the beginning. When AI first became a priority, what was the initial conversation inside Cox? And how did you avoid jumping straight into the tools? Tiffany, that's a really good question.

So, when AI first became a priority, the very first conversation in finance wasn't what tools should we buy. It was really what's our North Star? And what do we need AI to change about how finance operates? Leadership aligned really early on that AI and ML should help us make finance faster, more scalable, and more decision-oriented, not just automate a few tasks.

At the same time, we were aware of the risk of fragmentation. So, different teams experimenting. We had tons of vendors in our building knocking on doors. And suddenly, you have multiple point solutions that don't share data, don't share controls, and are hard to scale.

So, we intentionally put governance and discipline in front of tooling, clear decisions, rights, early partnerships with IT, legal, and security, and a focus on repeatable use cases where we could truly measure impact.

And that was the mind shift for us. We didn't want AI to become just another layer of disconnected tools. We wanted it to become a part of our finance operating model. And that's what led us into the discovery work and prioritization you'll see next.

Getting clear on the problems, the data readiness, and where should we start small but scale enterprise-wide. Right away, this highlights a critical distinction. We've noticed that organizations seeing real success with AI are not the ones moving fastest to adopt the AI tools.

They're the ones taking the time to align vision, governance, and ownership across finance and IT from the start, just like how Cox did. So, what did that strategy-first approach look like in practice? So, in practice, it meant we treated AI like a business transformation effort, not a technology shopping exercise.

AI is a very huge space. So, our first job was to narrow it down to the highest value problems that finance actually needed to solve. So, that meant we conducted cross-functional discovery. We identified pain points and opportunities across our teams.

And we built an AI use case inventory. Then we created what was called or what we coined an opportunity radar. It's just a simple way to rank those use cases by impact versus effort. And we intentionally favored work that would create enterprise value, not just localized wins.

So, use cases that show up every month, touch a lot of users, and can be scaled with consistent controls. But in parallel, we also engaged IT early. So, before we committed to anything, we wanted to pressure test the practical constraints.

What data can we access? What are the security and privacy implications? How would this integrate into our existing platforms and controls? And we also assess data readiness and process maturity. Because even the best AI capabilities, they won't deliver if the underlying process is inconsistent or data just isn't trustworthy.

So, that's what strategy first looked like for us. Start with the outcomes. Build a fact-based view of use cases. And align with IT on guardrails and feasibility from day one. Once we had that, choosing where to start and how to execute, it became much easier.

Well, it's clear you and your team were very disciplined and intentional up front, not just reacting to the momentum around AI, but stepping back to define where it could truly create long-term value.

So, what stands out to me is that AI was being treated as a business transformation, not a technology experiment or what you described as a shopping exercise, which is interesting. So, that shift from tool selection to problem prioritization and taking a very methodical approach to the transformation is what really ultimately drives measurable impact.

So, question, oftentimes we see there's a gap between strategy and execution. So, how did Cox decide where to actually start? So, we took a while. We actually spent about six months in this assessment journey.

Doing the discovery, watching the market move, and seeing vendors roll out new AI capabilities almost weekly. And at some point, we realized research alone isn't going to close the gap between strategy and execution.

We needed to take a controlled step forward and start learning by doing. So, we defined three non-negotiables for our first move. First, we had to be able to validate accuracy because finance can't afford close enough.

Second, it had to deliver measurable efficiency in a reoccurring process, not just be a cool demo. And third, it had to build trust through transparency, clear guardrails, and explainability so people understood what the AI was doing and when a human needed to stay in the loop.

So, that's what made OneStream. So, that's what made OneStream the right entry point for us. It met the success criteria. And it also was already in a platform that finance uses. So, we weren't asking people to go outside of their normal workflows.

From an IT perspective, it also reduced the need for new infrastructure and a heavy lift. And it was easier to support because it's a known entity or quantity in our environment with existing controls.

So, net-net, it gave us a faster path through governance while still staying disciplined. So, choosing purpose-built AI embedded in OneStream let finance and IT move faster together because we were building on something we already trusted.

And we could prove value quickly without creating a new island of technology. And that set us up well for the partnership with OneStream AI and their product development team. So, as our needs started to evolve, they moved right lockstep with us.

It's reassuring to know that Cox arrived at OneStream as the agentic AI vendor of choice, especially after your six months of due diligence. And for finance, agreed, accuracy and transparency are absolutely non-negotiable.

So, it's great to see OneStream's AI was able to deliver that. Our founder and CEO, Tom Shea, often says 80% accurate is 0% useful for finance. And so, that's very much true for you guys. So, question.

As you started working with OneStream's AI team, how did that collaboration shape your experience, especially as your needs evolved? So, we approached it as a true strategic partnership. And what stood out immediately was just how responsive the OneStream AI team was.

We set up regular feedback loops and we watched that input translate into real product improvements quickly. In fact, there were at least three major releases of the AI agents platform where the development took our feedback and delivered meaningful enhancement.

And that was really exciting just to see how fast the roadmap evolved and really became a reality for us. In many cases, things you might expect to take a year to show up actually manifested in a matter of months.

So, we approached it as a true strategic partnership. And what stood out immediately was how responsive the OneStream AI team was. We set up regular feedback loops and we watched that input translate into real product improvements quickly.

In fact, there were at least three major releases of the agent platform where the development team took our feedback and delivered meaningful enhancements. And what was exciting was how fast the roadmap actually became a reality for us.

In many cases, things you might expect to take a year to show up manifested in a matter of months. A concrete example is meeting users where they work. Early on, finance analyst was promising, but it didn't fully land unless it was available in Excel because that's where most of the analysis happens.

Once it was embedded in Excel's, users didn't have to leave the workflow. And that was a major unlock of adoption for us. So, overall, the team has been very attuned to customer needs and has continued to evolve the product in a thoughtful and a very practical way.

That's so great to hear, Andrew. Thanks for that. And your point about meeting users where they are is critically important. I can't emphasize that enough. You can implement the best technology, but if it creates so much friction that you're forcing your users to change their behavior, then there simply is not going to be any adoption.

So, you know, you also reinforce something broader here, which is when you combine tight feedback loops with willingness to operationalize that feedback quickly, you're not just improving the product, you're driving real usability and value for the business.

So let's switch gears here. I've also seen that with organizations looking at AI, governance is often where things slow down. So tell me about your partnership with IT and how it evolved throughout this process.

Absolutely. So we treated governance like a team sport from day one. Cox leadership had the foresight to stand up an AI council that brought together finance, IT, legal, and security so we could make decisions quickly, but also with the right guardrails in place.

IT's focus within our organization was to ensure we were doing things safely and sustainability, data privacy, secure environments, and making sure anything that we adopted would integrate cleanly within our existing controls and architecture.

In the early stages, I will be honest, it felt a little heavy. More process, more process, more reviews, because this was all new territory for all of us. But as IT got more comfortable with how the technology worked, the review cycles became much more streamlined, especially for embedded AI and platforms we already trusted.

The key insight was clarity of ownership. Finance led the use cases and the change. Where the value would come from and how people would actually use it. While IT, well, they provided the guardrails and the final approval.

That's when it really became a true partnership for us. Finance driving the outcomes in IT, ensuring we did it the right way. I love that. Treating governance like a team sport. It sounds like there were initial learnings of working with IT on this significant initiative.

But after a bit of clarity of ownership, it sounds like this streamlined the value that was delivered. That, to me, reinforces that governance doesn't have to be a blocker. But when finance and IT operate as partners, governance becomes an enabler of scale, not a constraint of innovation.

So let's make this clock story tangible. What were the first use cases of your OneStream agent implementation? And how did they impact the business? Absolutely. So I will walk you through our three use cases.

We had a use case that aligned with each OneStream agent. So in the case of the finance analyst, before we used the agent, we saw ad hoc queries that our finance team would take on, taking an average of 10 minutes per query.

We introduced the finance analyst, and that agent really has enabled self -service analysis that's trustworthy. It shows people not only the output, but also allows them to see how the system created that answer and to engage with the user as they interrogate their data.

So what that looked like for our team was like a 90% efficiency lift, just when it comes to ad hoc analysis, which is very much a routine part of our monthly analysis and reporting function. On the other case, when you look at search, which for us, we like to say is very similar to how our users use Microsoft Copilot.

So for us, we're in a period of change. And so we have a lot of users really being onboarded to new roles, people taking on more. And so what we found was that search could actually be an abler in an environment with a lot of change.

And so where we found search really fitting in as a real change agent is really our team using the search agent to really onboard our new users. And so for things like how to and tier one questions, rather than having to log a support ticket, our users are able to enter a question in layman's term into the chat box within one stream and really get pointed directly to that answer and also to the supplemental resources where that answer was sourced.

What we found is those questions that the users raised during the onboarding process. Our chat bot through the search had 100% accuracy in delivering those responses back to the users. And not only that, it was an efficiency lift of, on average, eight minutes for a user to ask a question.

Now, them asking that question and getting an accurate response back in about 30 seconds. And then the final frontier that we stress tested the agent was deep analysis. And so we align that with the contract and document analysis function within our accounting groups.

So what we see is that on a routine basis, our accounting teams are really responsible for reviewing hundreds of contracts and then making a determination on what type of treatment different activities should have based on how those contracts are set up.

So what deep analysis has done that's really transformative in that space is it's empowered our users to be able to really quickly scour hundreds of contracts, gleam those pertinent key terms and elements that really influence how the accounting treatment.

And we've taken that contract review and analysis time down from an average of 10 hours to 10 minutes. So you can see that's a game changer in our space. Wow. Those are some serious KPIs there. 90% efficiency, 100% accuracy, thousands of potential hours saved.

That's really great to hear. What stands out here to me is the focus on repeatable high frequency processes. That's where we've seen where AI can really drive compounding value over time, especially when it's touching such large groups of users across the organization like at Cox.

So let's switch gears again and see, talk through some of the challenges that you learned from this journey. Sure. Another great question, Tiffany. So the biggest lesson for us was that the hard part isn't the technology.

It's the people side of the change. When you introduce AI into finance, there's real fear about job impact. There's skepticism about whether the answers are trustworthy. And also a wide range of comfort levels with AI just in general.

So we had to invest in enablement early. Show what AI does. Talk about what it doesn't do. And really make it clear that this wasn't about replacing their jobs or their roles. It was really about giving them tools to do their jobs more efficiently.

We also learned that you can't design this in a conference room and then roll it out. We learned that AI with these types of projects, you have to learn alongside your users. You have to watch where do they hesitate? Where do they overtrust AI? And where does their actual workflow or process need to change? Building in transparency and a clear human in the loop moment in all of our integration points of AI is really where we figured out how we could really make AI thrive in our space.

And then as I reflect, I think the final lesson that we learned was really about governance. Early on, we didn't always get that balance right. So sometimes IT involvement, it felt heavy. Sometimes it actually felt like it was a little bit too late.

But over time, we found the sweet spot. Engage IT, legal, and security all before the decisions are made. Align on expectations up front. And really differentiate between net new AI solutions versus embedded AI within platforms that your organization already trusts.

That one distinction really helped us streamline reviews without compromising safety and security. Finally, and perhaps what I would say the ultimate lesson is that communication is a control. Regular updates, shared learnings, and really being transparent about what we're testing really kept our stakeholders aligned and reduced the rumor-driven resistance.

The payoff is that we now have a repeatable playbook. So every new use case doesn't feel like we're starting from scratch. Wow. There's a lot to unpack here. But two really big things stand out to me from what you said.

The first and foremost was that this ultimately is a people transformation, not a technology one. So clearly positioning AI as an augmentation, not a replacement, seems to have been critical to getting traction early.

And that makes sense. The second big thing that stood out to me is the way you co-develop this with users rather than designing in isolation or in a conference room, as you said. We talked about the hugely beneficial feedback loop between OneStream and Cox during development.

It sounds like Cox was applying that same successful strategy of using that tight feedback loop with their users during implementation and go live. So that's really, really interesting to me. So Ingra, I'd like to get your thoughts on the role of finance.

There's so much change externally and internally within finance organizations today between system upgrades, new projects starting. How do you see AI playing a role in finance in an environment of constant change? The truth be told, Tiffany, right now Cox is actually living in a period of change.

So what I would say to all of that is that in an environment of constant change, whether it may be you're implementing a new ERP or your organization is flexing through acquisitions or divestitures, or maybe it's reorgs and leaner teams, AI can still play a really practical role.

It can actually be a stabilizer when the ground is shifting. Something like a force multiplier when teams are actually stretched. And it can also be a huge buffer against knowledge transfer. So much critical context today lives in people's heads.

And AI, it can help capture that context, document it, and make it easier to reuse and scale. Most teams starting with AI should start with, you know, one specific localized problem. And that's the right way to build confidence.

But as you understand what the technology can do, the possibilities expand. And especially for finance teams that are being asked to run the business, transform it, and still deliver all of their day-to-day support all at once.

So in those moments where your organization are flexing through huge transition, AI can really move from a nice-to-have to mission-critical by helping keep performance steady while change is underway.

And then the final point is, I always come back to AI, it isn't the answer. It's a tool that helps you get to the answer faster and more confidently. That's really interesting. What I'm hearing is that AI really becomes the most valuable in moments of change, which is interesting because it's a bit of a shift from the broader narrative out there from what I've heard.

And I feel like AI is painted as disruptive or destabilizing. But what you're saying is through being able to preserve critical knowledge and extending team capacity, it's actually doing the opposite, which is stabilizing.

That's really interesting to me. So let's talk about a big concern regarding AI within the office of the CFO, skill gaps. Within the finance function, skill gaps and talent are often a big concern when it comes to adopting AI.

What's your experience, Inger, when implementing sensible AI agents at Cox? So my personal experience is that OneStream actually made AI easy because they embedded it in the workflows, the day-to-day workflows of our finance users.

If AI is hard to find or hard to use, people aren't going to adopt it. So one very real example for us is finance analysts. It actually allows users to ask questions in natural language and get answers grounded in their finance data.

And the key is that you don't have to be a technology person or developer to use it. The heavy lifting actually happens behind the scenes with this very innovative persona design that's bundled with the finance analysts.

What that persona lens does is it really reflects how finance and accounting teams think and work. It also anticipates the questions that they ask during the course of closing or forecasting or as they're providing variance explanations.

And it translates those questions into the right intent. So the system understands what the user means and not just what they typed. So building smart AI that's accessible for people at all levels, that is how you really win.

Because when that's done, it makes it a lot easier for all those naysayers to really be won over. Yeah, thanks for that, Inger. What really comes off is how critical that idea of purpose-built AI is to driving adoption.

This isn't just generic AI that's layered on top of OneStream. It's designed specifically for finance, embedded directly into the flow of work, and aligned to how finance teams are actually working and thinking and operating.

So beyond the current phase, tell us about Cox's future plans with OneStream agents. Yeah, so we are excited about where we are, but we are even more excited about where we're headed. One of the things that we are very much looking forward to as we traverse our AI roadmap is deeper, more seamless integration across systems.

So agents that can securely connect to upstream finance sources like our ERP and our subledgers and enrich the workflow end-to-end, not just answer a single question. The goal for us is a smoother drill-down experience, keeping the analysis in one place with fewer handoffs and less switching between tools.

That's really one of our North Stars. We also want stronger data connectivity across finance. So we want to be able to link what's in OneStream with our ERP and subledger details so answers can naturally progress into deeper analysis.

And finally, we're keeping the same shared ownership model between finance and IT as we scale. That makes sense. And that plan, that roadmap, you know, is aligned with what we're seeing in that, you know, agents are more powerful as they get access to more data and more seamlessly.

So thank you for sharing that, Inger. You know, three key takeaways from today's discussion with Inger from Cox here is, one, start with the strategy, not the tools. Two, focus on high-impact, repeatable use cases.

And then lastly, adopt AI where users already work to accelerate adoption and drive real value. So what Cox's journey told me today is that it shows that success with AI is not about doing everything.

It's about making intentional choices, starting small and building momentum through meaningful outcomes. Inger, thank you so much for taking the time to share your perspectives, your insights and your experience here with us all today.

With that, we should take some questions. So the first question is for you, Inger, is as you move from planning to execution, what were some early signs that your approach was actually delivering meaningful impact to the business? Yeah, so the earliest signs were very practical.

First, we could see time coming out of the work in repeatable process. Things that used to take eight to 10 minutes to chase down were now taking seconds or a minute. That created real capacity. And just as importantly, it sped up the business conversations because analysts could get to the answers while they were in the meeting rather than having to circle back.

Second, we saw trust from the very start, start to form. When accuracy held up in real usage, like our onboarding FAQs, where we validated that the responses end-to-end were the very same responses we would have provided if the users had called in help desk tickets.

So that level of accuracy, it really built in trust. When users could see the reasoning and they also knew when a human needed to confirm, they stopped treating AI like it was a novelty and start relying on it as part of their workflows.

That's so interesting. Yeah, so being able to see the transparency, the thought process to gain the trust of finance and also seeing that value right away with efficiency gains helps a lot, I'm sure. So interesting.

Okay, second question that's coming in is, what did you learn about getting your teams comfortable with AI, especially for users who may not have had prior exposure to these technologies? Yeah, so I go back to the people side of change.

The biggest thing that I learned is that change management, it cannot be an afterthought. There's a real understandable fear that AI is here to replace people's jobs. And if you don't acknowledge that openly, you'll feel an undertone of resistance, no matter how good the technology is.

So what worked for us was being very explicit about the intent. AI is here to empower and to supplement. Taking the first pass on the analysis, helping people find answers faster, and reducing the time spent on repetitive work.

So our teams, they can really spend more time on the judgment, the storytelling, and the partnering with the business. We also showed what it does and what it doesn't do. And we reinforce, there always must be a human in the loop.

So people understood they are still the ultimate decision makers. So for users with little prior exposure, we focused on making it safe and simple. Start in the tools they already use. Give a few day one prompts that they can copy and try.

And create space for questions. Quick demos, office hours, peer champions who could really share wins in plain language. Once people started to see a couple quick successes and understand the guardrails, comfort grew very quickly.

Because it stopped feeling like hype. And it started feeling like real help. That's great. It sounds like you and the team at Cox really did a great job with enablement, with the office hours and the example prompts to get them started.

Because oftentimes that's the biggest thing is like learning where to get started and what to prompt. So that's really great to hear. Another point I wanted to reiterate that you said was like AI is really helping shift finances work from those tedious tasks of just gathering and collecting data to the meaningful work, which is what we're meant to do, right? Being the partners of the rest of the organization or we're doing analysis.

So that's really great to hear that agents are helping with that. Okay. I think we have time for one more question here. So Inger, once you proved value with your initial use cases, how did you think about scaling AI more broadly across the organization? So Tiffany, that actually was the easier part because scaling, we did all of the work up front.

So when we did the discovery, we didn't just pick a couple of interesting ideas. We built a broad view of the problems that people were trying to solve and captured them in our use case inventory. The opportunity radar that I spoke about earlier, it gave us a shared way to prioritize what would create enterprise value and not just be a one-off single win for a team.

From there, we were intentional about starting with the use cases that could extend across multiple functions. So once we actually proved the value, we didn't have to go back to the drawing board. We could reuse the same governance approach, change playbook and patterns for adoption, and then expand to the next group of users with less friction.

A good example is our self-service ad hoc financial analysis, which was empowered through the AI agent. Financial analysis, well, that's not an accountant-only need. It's extensible across multiple business operations, such as tax, FP&A, and other functions.

So once the capability is in place, each team can ask better questions faster, and the compounding value comes from the repeatability. The same pattern applied over and over by more users with consistent controls.

That's really interesting. That makes sense. All right. Well, thank you again, Inger, so much for taking the time to share your story here and Cox's story with all of us today. Have a great day.