There's a moment in every company's AI story where someone in leadership says the quiet part out loud. At BDO USA, one of the nation’s leading accounting and advisory firms, that moment came sometime around mid-2023 when a group of senior leaders looked at ChatGPT and collectively said something like: "we have this hammer now, and there are a lot of nails."
What happened next is actually kind of a masterclass in how big organizations can move fast without breaking things. Zak Fay, AI Innovation Manager at BDO and leader of the company's RAID team, joined The Winners' Circle to talk about building something that actually works inside a heavily regulated, compliance-driven professional services firm. The BDO RAID team recently picked up a BIG Innovation Award in the team category, and once you hear what they've pulled off, you'll understand why.
What RAID Actually Stands For (And Why It Matters)
RAID is the Research and AI Development team, and the name tells you everything about how BDO approached AI. They didn't stand up a committee or hire a consultancy to map out a two-year road. They built a small, nimble, exploratory unit and gave them room to move.
"We built a small team that was full of creatives, full of dreamers, full of futuristic folks, full of learners," Fay said. "We interface directly with the business. We do workshops and hackathons to try to build up what is the next thing that's really going to move the needle."
The team includes Gavin Martin, a remarkably optimistic developer who brings steady enthusiasm to every build" Nate Hodges, a longtime friend with a rare ability to translate business needs for technical teams. Steven Wangler, the developer manager and ringleader who consistently pushes the group forward. And Jordan Sepetys, a relatively new addition who joined as a business focused on process improvement and bringing organization, transparency, and helping the team move work from idea to pilot.
Together, they sit at the tip of the spear for a much larger Data & AI team that's now around 20 to 30 people. But RAID is the group that identifies the opportunities, builds the proof of concepts, and drives the relationships that make adoption actually happen.
The Trust Problem Nobody Talks About
Here's something worth sitting with for a second. You can build the most technically impressive AI tool in your industry and still have it die in the hallway outside the conference room because nobody trusts it. According to recent data, 70 to 85% of AI initiatives fail to meet expected outcomes, and the number one reason isn't the technology. It's adoption.
BDO took this problem seriously from day one. Fay talked about how AI adoption can feel scary inside a large firm. Is this going to take my job? Is this something that's going to do things I don't understand? Those are real concerns, and dismissing them or just doing another all-hands call doesn't make them go away.
What actually worked was a combination of top-down buy-in and grassroots trust-building happening simultaneously. On the leadership side, Mike Gerhard, BDO's Chief Data and AI Officer, was consistently in the room with the C-suite, making sure that executive stakeholders felt informed, collaborative, and bought in to the direction. On the other end of the spectrum, RAID was meeting individually with service owners, premier teams, and what Fay calls business line stewards, the folks across the firm who actually do the work that AI is being asked to support.
"Instead of just saying it, we're showing it," Fay said. "That's been pretty massive. We're making tools for them, not for me. So we make sure they have creative control over the direction."
They also joincompany-wide fireside chats where Zak and the team get on a call with the AI stakeholders and anyone in the firm who’d like to learn more. They talk about what they're working on, answer questions, and model transparency in real time. The result? From 2023 to 2025, 66% of BDO's 14,000-plus employees now regularly use Chat BDO - the firm’s internal generative AI platform.. That's not a test group or a pilot. That's the majority of employees, actually using it.
Nine Solutions in Under a Year: How the Sausage Gets Made
Rolling out nine AI solutions in under a year inside a firm with hundreds of services, heavy regulatory constraints, and multiple business lines is not a small thing. To put that in context, McKinsey research shows only 23% of organizations are currently scaling any agentic AI system across their enterprises, and most of those are doing it in just one or two functions. BDO's RAID team did it across nine significant high-impact use cases in the same timeframe. So how exactly does RAID take something from idea to production?
It starts with the business line stewards. RAID doesn't sit in a room and imagine what would be useful. They go out to the firm, run workshops and discovery sessions, gather real data through interviews, and do root cause analysis to understand what the actual problem is, not just the surface symptoms. Fay referenced a five-why approach, which will sound familiar to anyone who's worked in process improvement. You keep asking why until you find what's actually broken.
One of their most impactful tools is the Consultative Selling agent, which Fay says has now logged over 35,000 uses in the first year. The idea came directly from the business. BDO professionals were spending hours researching clients before meetings, digging through LinkedIn, piecing together context, trying to understand what services might actually be relevant to a given prospect. The RAID team built a tool that does that research in about 10 minutes and packages it up in a way that's genuinely actionable.
Fay framed it in a way that really clicked. "We aren't interested in making a robot salesperson or a robot relationship person. Can we just make it so the research is easier?"
That framing, fixing the research so the human can be more human, runs through everything RAID builds. The goal isn't to replace the professional. It's to remove the stuff that's in the way of what makes that professional great.
Once a proof of concept shows it works, the BDO’s Data and AI team goes through a rigorous governance process that includes legal review, data security and privacy checks, and sign-off from the AI governance team. Nothing goes to production until it's been properly vetted. For a firm serving clients with sensitive financial and business data, that's not bureaucracy. That's responsibility. And it's the kind of discipline that separates the teams actually scaling AI from the ones still stuck in pilot purgatory. Research from Box found that leading AI adopters using process automation see 37% productivity gains over their less disciplined peers.
AI Adoption in Enterprise: The Lesson BDO Is Teaching the Industry
Here's the thing about AI adoption in enterprise settings. Most companies are either moving too slow because they're scared, or moving too fast because they're excited, and skipping the trust-building that makes the whole thing sustainable. BDO is doing something genuinely different.
Russ Fordyce, who spent years inside major companies like Comcast and Windstream and knows the corporate AI hesitation cycle firsthand, pointed this out during the conversation. "I stepped out of that corporate world a few years ago. At that time, it was halt. Do not touch ChatGPT. Obviously the world has radically changed."
What changed it wasn't just the technology getting better. It was organizations like BDO proving that you can do this responsibly. You can build an internal team that operates with transparency, moves quickly, governs carefully, and actually gets adoption instead of just delivering a tool that collects digital dust.
The million hours saved by Chat BDO alone is a remarkable number for a firm of BDO's size. But Fay made a point that's even more interesting. That number is going to compound. As more employees get comfortable, as more use cases get proven, as more agents go live, the time savings will accelerate in ways that are genuinely hard to predict. OpenAI's State of Enterprise AI report found that 75% of workers say AI has improved either the speed or quality of their output, and the gains get steeper as organizations move from casual use to integrated workflows.
"We're a force multiplier," Fay said. "My team loves it. We're very interested in just making it so that you can get home to see your kid's soccer game."
That last line landed in a way that a lot of AI conversations don't. The most ambitious AI strategy at one of the country's biggest professional services firms is ultimately about getting people home to their families a little sooner. That's the human case for all of this.
What They'd Do Differently (And What To Steal)
Fay is refreshingly candid about where things got hard. He named the biggest pitfall for any team trying to build something like RAID: expecting the first iteration to be the right one.
"The first pass at this is going to be a draft and you need to be comfortable going, all right, rebuild, relook, make an adjustment," he said. "We've been really successful because of that."
That agile mindset, reviewing results, inspecting what's working, tweaking the structure even when things look pretty good, is something a lot of organizations resist. Eighty percent success in month one might feel like a failure when you're used to measuring in years. But for exploratory work, 80% is a strong starting point if you know how to learn from it.
If you're thinking about standing up your own version of a RAID team, Fay offers one clear piece of advice that goes beyond the org chart. Get buy-in across the business before you start building anything. Not just from the executive team and not just from the technical folks, but from the people who are actually going to use what you make.
"It becomes a relationship problem," he said. "If we have folks on our team who are just fanatical about AI, it makes the conversations much easier. The business starts to speak to us in a way that we can actually use."
BDO and its RAID team, Zak, Gavin, Nate, Steven, and Jordan, are doing exactly what the BIG Innovation Award is designed to recognize. They're not just experimenting with AI. They're doing the harder, slower, more important work of actually making it work inside a real organization with real stakes. The award is well-earned, and frankly, the rest of the industry has a lot to learn from how they got here.
Congratulations to the whole team.