Over the last few years of partnering with finance teams, we've learned that the people actually getting AI projects across the finish line inside their companies rarely have "AI" anywhere in their job description.
They're the AR director, the controller, the head of finance ops. These leaders pick up the problem because they own the number it's tied to. They're tired of running an exception-heavy process on tribal knowledge and a spreadsheet nobody else understands. They're already convinced the potential of AI is real, but they need an approach that survives a real company's constraints and politics.
And they understand the stakes are personal, because when they're the one who championed the pilot, their credibility ships with it.
This perfectly describes the operators who joined our CEO Chris Sperandio for a Boston Tech Week panel last month: Suzie Rawden, a supply chain digital strategy director at Johnson & Johnson; Jason Wotman, a finance leader at PE-backed parts distributor Parts Authority; and Ben Ganzfried, a senior director of data and governance at Hungryroot.
They work in different industries with very different data, but there was a consistent pattern in what made their AI implementations successful. None of it was about the model. It came down to how each pilot was set up before anyone wrote a line of code, and in nearly every case, the person who drove it was an operator who owned the process rather than a specialist who owned the technology.
Here's what these real AI leaders do differently — job description be damned.
Pick a process you actually own
The AI projects that survive tend to be built for and by someone who owns the KPI, can change how the work actually gets done, and has the alignment to follow through.
"People come up with ideas to help another group, essentially saying, 'This annoys me, so I'm going to build you a tool that makes it work better for me'," Suzie Rawden said. "Those types of solutions don't tend to go very well."
That's why finance and operations leaders are often better positioned to lead here than a centralized "head of AI." They already own an exception-heavy process (something like AP, cash application, or order processing), they know which exceptions actually matter, and they have a clear picture of what "done right" looks like. That ownership is the asset.
Scope it as a prove-it pilot
A small, fast, working result is the most persuasive thing you can bring to a conversation about a bigger investment, because it turns "Trust me" into "This is already working." One of our panelists calls this approach the "prove-it pilot", and it's the complete opposite of a traditional, IT-heavy implementation plan.
"Instead of going after the whole thing right away, take something small that still has impact but can be done fast," Jason Wotman advised. "Show some results, and then use that as the ammunition to go back for the conversation about a larger project."
We've seen the same thing here at CoPlane. One customer started with invoice and exception handling in AP, automating the routine end and eliminating much of the team's manual workload. Humans stepped in only on the real exceptions. It was a contained, ownable process where the win was obvious to everyone who saw it — which is exactly why it opened the door to the larger opportunity.
Sell it as math, not magic
Executives don't have a uniform appetite for "AI." Some want it everywhere, and some are wary. But nearly all of them share an appetite for efficiency and profit, and the operators who get pilots funded speak in that language.
Jason described deliberately reframing the pitch away from the technology and toward a plain business case. "Rather than saying we're going to build a great tool and AI is going to power it all, I'll say 'We currently have this many people processing this many invoices a month, we can automate 80% of it, and that'll save us these hours'. It creates an approachable opportunity-cost analysis they can do, much like any other investment decision."
The point isn't to spend more on AI, but to attach a real number to every initiative and work backward from the economic value.
As Chris elaborated, sometimes that exercise reveals a process is quietly losing money and shouldn't be done at all. He described a company with dedicated staff whose entire job was chasing payment disputes that cost more to recover than they were worth. The highest-ROI move wasn't to automate that work, but to stop doing it altogether.
Make governance the reason it scales
For many AI champions, governance and IT show up as the brake. But the operators who get pilots approved treat governance as a tool that helps them scale. A clear, auditable record of who can do what is exactly what makes automation safe to expand, and the smartest operators lean into those constraints rather than fighting them.
Ben Ganzfried, the data governance lead on our panel, called his philosophy "enabling in a smart way": instead of opening up broad access or blocking requests outright, his team builds narrow, governed pathways that answer specific questions reliably and safely.
Chris described how we've built CoPlane around the same idea. It runs inside your own environment so sensitive information never leaves your trust boundaries, and it does the work as the person who already does it, using the permissions they already hold. No one has to hand over the keys to core systems to get started.
This principle gives the operator a way to evaluate vendors: the ones worth backing are those that fit your controls instead of forcing IT to take on new risk. Bring IT a partner like that, and the relationship shifts from adversarial to aligned, with finance owning the rules and outcomes and IT owning the infrastructure.
The pilot becomes the proof
Put these four lessons together, and a pattern emerges. The operators getting real results pick a process they own, scope it small enough to prove quickly, make the case in the language of dollars and hours, and treat governance as an accelerant instead of a roadblock. They design a pilot so the win is obvious to everyone who needs to see it.
From there, it compounds. One proven process becomes the template for the one after it, along with the political cover to greenlight it: routine work automated, people freed for judgment, process logic finally legible, and real-time KPI visibility.
Ready to follow in our panelists' footsteps? Our AI Implementation Toolkit is a step-by-step guide to scoping, justifying, and winning approval for your first pilot, with the templates and ROI calculator you'll need to get started.