Firm leaders are getting contradictory messages right now. Every conference, every vendor, every LinkedIn post says artificial intelligence is about to transform accounting. The implication: If you're not implementing it yesterday, you're already behind.
But the companies actually building AI tools for accounting are working on a different timeline than the one being marketed.
When I talked with Mike Cieri, executive vice president of software at Bill, about implementation timelines, he was direct: "We'll see rapid change over five years, but I don't think in six months you're going to have a complete turnaround on how the accounting industry works."
Five years of steady change — not a six-month revolution.
That gap between the hype and the reality is creating unnecessary anxiety. Firms are making decisions about AI adoption while feeling panicked about being left behind, when the actual timeline gives them room to learn and test properly.
What 'experimenting' actually means
Most firms fall into one of two camps: diving into AI implementations without testing, or freezing because they don't know where to start. There's a middle path.
"Take a couple of associates and say, 'We're going to try this in your area. We're going to try this with a few clients and experiment with it until we feel good that there was value added, and we have the controls we value as a firm in place,'" Cieri suggested.
The word "experiment" matters here. An experiment has defined parameters, a limited scope, and permission to produce unexpected results. A firm-wide rollout has none of those things.
This approach addresses what firms actually worry about: What if this doesn't work the way the vendor says? What if it creates more problems than it solves? What if we spend money and time and see no real benefit?
Testing small answers those questions with data instead of assumptions.
During my years in public accounting and later in C-level roles at companies, the frustration I saw repeatedly was talented people spending hours on work that didn't require their expertise: Transaction coding, data entry, repetitive reconciliations.
AI should solve that problem — not by replacing accountants, but by handling the work that doesn't need human judgment.
In our conversation, Cieri framed it this way: "Human involvement will be high leverage at key moments, where creativity is needed, judgment is needed, advisory is needed. We're trying to amplify the value of human intervention in those moments."
This changes what entry-level work looks like. Instead of spending two years learning to code transactions before getting to do analysis, new staff can move into interpretation and pattern recognition faster. The technical skills still matter, but they're not the bottleneck anymore.
For experienced professionals, it creates bandwidth for work that keeps getting postponed: Strategic client conversations, team mentoring, te advisory work that actually requires expertise.
From the client perspective: "I should be happier that we're showing up to meetings and getting a higher-order thinker on the other side, showcasing for me a better picture of my business than I was getting before."
That's the actual value — not efficiency metrics on internal processes, but
How review processes build trust
The question I hear most often: How do I know the AI did it correctly?The same way you know a new staff person did it correctly: You review their work.
"We think of AI as just another actor in the system. We record those actions, there's clear auditability, there's clear transparency," Cieri said. "You're building trust over time through verification. You see what the AI did, you check its work, you override when needed. The same process firms already use for training staff."
The difference between firms that scale AI successfully and firms that pull back after problems: The successful ones built review processes from day one. They didn't assume it would "just work."
If you're somewhere between, "We should do something about AI," and "I don't know what to do," here's a practical framework:
- Spend time learning what AI actually does. "People tend to fear things they don't understand, so just spending some time on that alone ... to separate some of the fact from fiction in your own mind" makes the difference between decisions driven by anxiety and decisions driven by clarity, Cieri suggested.
- Pick one workflow that's causing pain. Not the most complex one — the most consistently annoying one. Test with a limited scope. One person, one set of clients, defined timeframe.
- Define success before you start. What are you measuring? Time saved? Error reduction? Less end-of-month stress?
- Then pause before scaling. What worked? What didn't? What surprised you? Most firms skip this reflection and miss the learning.
"Firms have a choice. If you want to turn on agents to do some of this work for you, that's a choice you're going to be able to make," Cieri said. "It's not just going to happen overnight."
You control the pace of adoption.
When I work with executives and partners on transformation — technology, workplace culture, business process — the pattern is consistent: The ones who succeed don't move fastest; they move with intention. They test, reflect, adjust.
The ones who struggle try to do everything simultaneously because someone told them urgency equals importance.
AI creates opportunities for firms to reclaim time for work that matters: Strategic advising, client relationships, team development. But only if the adoption process itself doesn't create the burnout and overwhelm that AI is supposed to solve. AI trends come and go. Fulfillment is evergreen.
Cieri said it best: "This is about supporting, not replacing,
The technology will wait. The question is whether you're making decisions from clarity or from the pressure to keep up with what everyone else seems to be doing. Take a beat. Test small. Build trust through verification. Scale when you understand what you're scaling. The firms that will succeed with AI are the ones who test first and scale deliberately.
