Good ideas can come from anywhere, especially when informed by real world experience, a fact that some firms are leveraging to source new solutions from field staff, which serves to both rapidly expand their available tools and upskill their professionals.
While the specific approach varies from firm to firm, those that empower staff accountants to take an active role in technology development report they have found new versatility and flexibility against the practical client engagement challenges, as well as increased engagement and enthusiasm.
However, much like any other project or initiative, success is not just a matter of throwing together a program and calling it a day. Regardless of how someone decides to pursue it, taking a deliberate approach is essential to ensuring the effort does not become a monumental waste of time and money. While different firms do so in different ways, they all stressed the importance of intentionality, oversight and a willingness to recognize when something's not working.
Accounting Today is looking at how accountants are being empowered to contribute new solutions at three different firms. Today we will examine the collaborative approach of Top 10 firm Grant Thornton.
Grant Thornton sees its staff as the true source of ideas for new products and solutions, and has developed a structured process to harness them and realize viable tools and offerings. While rough concepts and prototypes originate from many parts of the organization, they all filter upward through a series of channels that ultimately lead to a centralized development team that oversees refinement and deployment, as well as long-term management and maintenance of the final product.
"We do concept sprints," said Mike Kempe, the firm's chief information officer. "We do a lot of product incubation work and ultimately we come up with a set of use cases that goes into our pipeline. Our professionals are bringing the ideas forward. This is not us in the back office thinking of what we should be doing. It's all coming directly from our professionals."
While the firm always had room for staff innovation, he said the true formalization of the process came with the launch of its AI enterprise strategy last year as part of its efforts to become an AI-powered firm that can move quickly through experimentation to scaling to adoption. By bringing disparate efforts into a standardized process, the firm was able to bring clarity and efficiency to what previously might have languished or, even worse, become a governance liability.
"In the past, before we had our AI strategy, there was a lot of work going on, but it was decentralized. It was informal. There wasn't a whole lot of governance and risk management over it," said Kempe. "By implementing this program and putting in place the strategy, we really have a central way to take our ideas in, evaluate them looking at risk, looking at data, looking at scalability, applying all our governance standards, and then try to go as fast as possible for the high-value use cases from concept directly into product."
Ideas begin with the aforementioned concept sprints, incubators and other staff-level innovation programs. The results of these efforts are routed to a centralized intake where they are then evaluated in terms of specific business outcomes. Will it improve audit quality? Will it increase productivity? Will it help risk management? And how will it scale? These are the kinds of questions the team considers when deciding what to prioritize.
"What we're trying to do is be a little bit more strategic and pragmatic about this, saying, let's look at the entire workflow and see how we can optimize that and put AI enablement directly inside of the workflow, as opposed to a standalone tool or a bolt-on solution," he said.
Once the team decides on an idea, it then moves quickly from concept to experimentation to an actual solution that can be scaled and deployed not just in the U.S. but globally throughout the firm. Kempe noted that the recently launched
A similar process happens when it comes to AI-guided "vibe" coding. Someone develops a great idea for a solution, and a member of the product development team sits with them and side by side they develop a proof of concepts, mock-ups and even, to some extent, working systems; this close collaboration is possible because AI allows them to do this in a matter of hours as opposed to weeks or months. Kempe said this is an area they're looking to further open up in the future by setting up more sandbox environments where people can safely experiment and, if things work out, have their ideas further developed by the professional team for firmwide deployment.
"That's an area that we are looking to democratize more so we can get the power in the hands of our people [by] setting up sandbox environments and giving a certain amount of users access to those environments so they can do a lot of that themselves. We want to move as fast as we can — obviously, with the right governance, the right oversight and risk management, and the data aspects — so [we want to give staff] access to a controlled environment for them to be able to vibe code. And then, if this thing becomes a real concept, then we want to turn it into a real scalable product that we're going to deploy globally. The centralized product team will then take over and build it up from there," he said.
This is similar to how Grant Thornton currently handles AI agents. While the product development team releases hundreds of its own agents throughout the firm, individual staff are also empowered to make their own within a safe environment cordoned off from sensitive data. However, much like other things, those with a good business use case can then be filtered upward for refinement and conformance with governance standards to be deployed across the firm.
Kempe said the firm does its best to balance between empowerment and safety.
"We want to strike the right balance between democratizing AI, because we think there's so much demand for AI, with the proper governance and oversight and risk management. Obviously we're in a heavily regulated environment, so we have to make sure that everything that we do has the proper guardrails associated with it," he said.
Beyond helping with governance risk, guardrails also manage costs. Kempe noted that these AI solutions are not free, especially if they use a consumption-based pricing model. Monitoring is required to not only ensure compliance with the right governance frameworks but make sure their costs don't suddenly spike because of unintended use or sudden model changes.
"Every AI solution will use tokens with ChatGPT or Anthropic and we have to watch that, because if there's some unintended usage that's happening by a user, it may spike your cost, so we are constantly monitoring the cost aspect of it with the LLM providers. … There's a cost associated with the ongoing governance and risk oversight because the models are changing underneath, so you may get one outcome today, and six months later you may get a different outcome for the exact same use case. Monitoring these models is also a cost," he said.
He said any firm considering a similar program needs to take governance seriously from the start. Data standards, risk management, professional ethics and cost effectiveness should be baked in from the beginning. "You have to have the right governance, but you don't want to slow things down to a trickle where nothing can be implemented for another year."
