The rise of apps coded by artificial intelligence is changing the dynamics between software vendors and the accounting firms that they have traditionally served. While the impact so far has been small, as only a minority of firms have truly embraced so-called "vibe coding," it could be the beginning of a truly disruptive event for the accounting software industry.
While firms have been making their own custom software for years, the recent introduction of AI coding platforms has significantly lowered the bar for their development, opening the process up to even non-technical staff with no engineering training at all. This has served to save firm leaders like Randy Nail, CEO of Top 100 Firm Hogan Taylor, hundreds of thousands of dollars on at least a few occasions.
At one point his firm needed a new audit methodology due to a change at a third party they'd previously relied on. While they eventually settled on a new platform, they found it lacked sufficient financial reporting capabilities. He said they could have found a point solution that would bridge the gap, likely costing them about $200,000 a year — or they could vibe code.
(Read more: See
"We said we don't want to spend that kind of money if we don't even know if it's going to work. Why don't we just try to solve the problem ourselves? So our internal people, interacting with LLMs, [said] 'Here's our problem: can you take this tool, Word and Excel, and produce something that would give us confidence in financial reporting?'" he said. "It worked."
Kacee Johnson, fintech innovation leader with AI-specific business consultancy Radical as well as the co-founder and executive director of the AI Native Accounting Foundation, said she has been hearing stories similar to Nail's throughout the profession. She has begun asking firm leaders about their experiences with vibe coding and, increasingly, has heard them choosing to let AI build the solutions they once might have bought from a vendor. She brought up a conversation she had with a small firm leader as an example: His firm had already decided the best-in-class solutions they needed were too expensive and so were planning to just "deal with the headaches."
But now, he told Johnson, "We're just vibe coding the solutions we need."
This has led some firm leaders, such as Nail, to feel more confident pushing back against their current vendors, and more comfortable saying no to new ones. He said lately some of their vendors have just announced price hikes and "Just say 'Tough.'" But now, he said, the firm is more comfortable saying "No, we're not going to do that." He caveated that they generally do not let themselves get pushed around by their vendors and so they'd be saying no regardless, but the fact that this is happening at the same time that they're getting better at vibe coding is very convenient.
"You and I might have to suck it up when it comes to our Netflix subscription going up, but firms don't have to do that. We have options," he said.
A see-saw for software developers
This new landscape has led to some vendors losing business. Donnie Hampton, co-founder of audit solutions provider
"We talked to this really awesome firm, it was tech-forward and was honest about needing this help, but they just kind of ghosted. And then I saw them at [the AICPA's Engage Conference] and they said, 'We're testing out [Microsoft] Copilot and [Claude] Cowork and they've been a huge unlock on certain things,'" he said.
But at the same time, he said the opposite has also happened: firms that thought they could develop a solution internally and then, after running into the unique challenges this can present, reached out to his company because the DIY approach failed. He said he recently had this very experience with a large firm that called him.
"The rollout of what was going to be internally developed still hasn't taken the value and efficiency gains to the next level," he said.
What's more, his company still gets plenty of business from the majority of firms that have not been into vibe coding, or are only using it on low-stakes applications that don't interact with client data. Many professionals recognize that it's one thing to get AI to build a slide deck, but it's another to get it to build a client-facing application that draws on sensitive financial data. Firms want convenience and value, yes, but they also want security and safety and many are willing to prioritize the latter over the former. He noted this seems to be case particularly among firms with risk advisory practices.
"When a firm thinks seriously about risk, when the stakes are particularly high when you're handling very sensitive client data, you're probably a bit less inclined to try and trust this to an AI solution. On the other hand, if it's something like taking notes during a meeting, or scheduling an event, or I talked to someone who made a tool that calculates their cloud compute cost, that's purely internal stuff and that's a little bit more where people are willing to trust AI," he said.
Despite being many times larger than Roz, Thomson Reuters has also observed this dynamic emerging. Stuart Cobbe, senior director of product management, acknowledged that AI-generated apps have indeed affected the competitive landscape, and there are many firms that have successfully built simple internal tools or prototypes. However, he does not view these tools as one-for-one competitors but, rather, as a force raising customer expectations in the market. Competition is shifting from feature-by-feature comparisons to broader questions of value, which is not necessarily a bad thing.
"Customers now have a clearer sense of what modern technology can do. They are experimenting with horizontal AI tools, they are asking smart questions, and they are less willing to accept slow innovation or rigid workflows. That creates more pressure on vendors, but it is also healthy for the industry," he said.
Regardless of how one feels, every software company needs to understand that vibe coding is now part of the landscape and, as such, it is raising expectations around speed, customization, and who can create software, according to Cobbe.
However, he is not especially worried about competitive impacts. While firms are doing impressive things with vibe-coded apps, their various issues limit the degree to which they can pose a threat to vendors like Thomson Reuters. Firms may be able to quickly generate an app, but this does not provide the infrastructure to support client work, regulated workflows, confidential data, compliance obligations or firmwide scale. For that, he said, firms need trusted systems built around security, accuracy, governance and authoritative data. So far, he does not see AI instantly producing that.
"The distinction is important. AI coding tools are good at helping people create things quickly. But speed is only one part of the equation. CPA firms also need confidence in the underlying data, the workflow, the security model, the testing, the auditability, and the long-term reliability of the system. That is especially true when the work touches client obligations or professional judgment. While coding agents are great at producing prototypes, we have seen them struggle to produce enterprise-grade software," he said.
