Accountants share how they're using AI

Data preparation: "We are trying to come up with ... use cases that can potentially go cross-functional, where doing one or two things can bring in almost 30-40% efficiency in their day, whether that be tax or audit or ESG," explained technology advisory practice leader Suma Chandar. 

"So, how do you scrub data from a PDF or Excel or from a Word document and put it in a format and analyze it quickly?" she asked. "Everyone wants to do that for different reasons. They want to scrub it for financial statements or for an audit or for an analysis. If you want to call it a use case, scrubbing data from a PDF and putting it into an Excel and being able to give the end user the ability to analyze it without going through so many invoices is very simple and you're saving 30-40% of your time on this, so this is something that cuts across the firm. It's a pretty simple solution but that is one very powerful use case."

Cashflow analysis: "We've got 13 Week Cash Flow that has RPA, some machine learning and some analytics," said Carmel Wynkoop, the partner-in-charge of AI, analytics & automation. "The RPA goes and gathers information around spend and cash, the machine learning analyzes this information and then predictive analytics can explain what your cash will look like in 13 weeks. We're seeing a lot of traction there on a client basis." 

SOC 2 preparation: "Audit Ally is our SOC 2 platform we created in-house and within the tool itself we're leveraging Anthropic's Claude [family of large language models] to help auditors fill out the information they need," Wynkoop explained. "So think of an evidence request: You have to write these things who knows how many times. There's no need to key all that information in over and over…. You don't have to open up Copilot or GPT, you're working directly in it. … It saves a paragraph or two of writing, it saves a few minutes, but over the course of a year that makes a big difference. … You can't just automate the entire CPA's soup to nuts — that would be a terrifying liability — but for the little things that you get along the path are so much easier now than they used to be." 

Bespoke digital "workers": "This was 15 hours a week reconciling funds across 70 entities with hundreds of bank transfers. We created a digital worker that interacts with these web-based systems and does all the consolidation and reconciliation and provides a full audit trail and then sends it for approval," Wynkoop said. "We also created a digital worker that goes into 100-plus investment statements, pulls information on a monthly and quarterly basis — they're spending 50 hours a month doing this and creating journals — so the digital worker now goes and does all this and creates a journal entry, points out the anomalies and routes it for approval. They save about 500 hours a year utilizing the RPA." 

Answering internal staff queries: "Everyone wants to know, 'What days am I off?', 'How much can I get in my 401(k)?' … Using Copilot Studio we've experimented with things like that, or directory information: If someone needs an AI expert, they'd quickly find Carmel," explained OJ Laos, director of Armanino's AI Lab.

Financial statement analysis: "We take a client's financial statement, export it to a PDF, take all the client information out … and we upload it and ask, 'If you were going to analyze this, what would be the key factors you'd alert the client to quickly?' Within seconds, it does it," said GWCPA managing partner Samantha Bowling. "[This is] something we would do anyway but it would take us a while to analyze it. … We're using it for clients who do bookkeeping work and outsourced CFO type work. We see this bring a quicker tool to get answers." 

Generating content with consistent tone: "We went through this whole rebanding thing," Bowling explained. "We've been able to generate stuff that would have taken us forever to do, so it's been a huge time-saver. Let's say we communicate with clients, I put in 'This is what I want to say, how do you change the wording to illustrate these values?' and it does it. It's unbelievable how fast it works." 

Exploring hypotheticals: "We asked, if you had a client who had a bunch of leases on equipment, and if we gave you those leases and gave you comparables for the leases in the [Washington, D.C.] metro area, could you tell us if it was below or above market?' And it said, 'Yes, I can do that,' and even put it in table format," Bowling said. "It's kind of crazy … We're having these Foresight Fridays — foresight is one of our values — in the summer where you have four hours to expose yourself to something new. I'd ask questions like, 'How would you capture this information and what would be the benchmark and put the results in a table for me?' and it did it in less than five seconds." 

Finding what you forgot: "I had a client ask me what other executive bonus programs can [someone] offer, like life insurance?" Bowling said. "And so I could say to ChatGPT, 'I know these benefit packages are good for executive officers, what else am I missing, can you analyze this for me and tell me how I could deliver that to my clients?' and it answers. It's like we finally have the technology that can really provide value and we can really charge for value now as opposed to time.