AI hype is easy to find, whether it’s worries about software agents replacing accountants or hopes for a super-powered finance function with unprecedented predictive abilities.

But what isn’t so obvious is how companies are really using AI in finance in 2026. More than three years after ChatGPT catalyzed new interest and investment in AI, many companies are still learning how the latest AI iterations might fit into their business models.

A recent report by Gartner found that while close to 60% of finance teams are piloting or fully implementing AI projects, only 7% of CFOs are reporting a strong impact from that investment.

“Right now, AI is struggling with an identity crisis,” said Mohit Sharma, ACMA, CGMA, who recently launched two AI finance startups after a decades-long career in finance.

So, what are companies trying? And what’s working?

Three leaders from across the business world share how they’re using AI and automation in finance. (See the sidebar, “3 AI Implementation Tips for Finance Professionals,” at the end of this article.)

GETTING PAID ON TIME

Sharma co-founded Pinaka AI in 2023 to address a common frustration he heard from finance leaders in his globe-trotting career.

“There was one common, consistent problem,” he said. “Around 60% of the customer invoices that are generated in a B2B environment do not get paid on time.”

A late payment can cause a cascade of consequences. The sales team might have to negotiate a new contract. The company might need to take on debt.

“It is an entire cycle,” Sharma said. “This is such a vicious problem.”

Pinaka AI, which employs about a dozen people, has developed a software product that predicts which customers will be late making a payment and the specific reason they appear likely to miss it. It’s now being piloted by two large manufacturers in India, he said.

“I’m using intelligence that is scattered all across my landscape in order to determine who’s going to pay it and why,” Sharma said.

The company’s algorithm can make predictions with 96% accuracy, he said. Hosted on Oracle Cloud, the platform can also use generative AI to recommend actions for the user to take and can draft and send custom-tailored emails to help collect payment on time.

“It gives recommendations on what to do to resolve it today, weeks ahead of time,” Sharma said.

It’s an ideal use of AI, he said, because it’s tapping into a broad spectrum of data to deliver an actionable insight at a scale that would not otherwise be possible. It uses what Sharma describes as four types of AI — a recommendation engine, decision intelligence, a classification algorithm, and generative AI.

“The payment behavior of a customer is scattered across the systems,” Sharma said. “That’s the job of AI: integrate, create a single version of truth.”

Pinaka AI’s predictions rely on data from a customer’s customer relationship management and enterprise resource planning sources, as well as external sources such as credit agencies and news sources, according to the company.

The startup is still in its infancy, having recently completed one accelerator program. It is also working through its test deployments with the major manufacturers.

It is a task that automation alone might have addressed in earlier years, Sharma added, but the rise of new AI tools has made it far faster and more economical to develop the product.

“When we are solving for complexity, we need intelligent systems,” he said.

UNIFYING MESSY LEDGERS

Janice Stucke, CPA, CGMA candidate, took a new job last year as CFO of CREW Network, a trade association connecting more than 14,000 women in commercial real estate around the world.

Stucke took over a finance department that still wrote paper checks and needed a thorough overhaul, she said. It was right in her “sweet spot,” as she had implemented robotic process automation (RPA) and generative AI in her previous jobs.

“We needed to completely automate the entire infrastructure,” Stucke said.

But first, she and her team had to confront accounting data that was spread across some 50 entities’ charts of accounts. Some were subsidiaries, others were adjacent entities, and they were spread across several countries — not to mention various formats.

The fragmented data was slowing down the organization’s payment processing. It also required constantly updated custom coding to maintain — a lot of technical debt for an organization with just 35 employees.

“With so many entities and each having their own chart of accounts, we could generate over 10,000 lines of [general ledger transactions] per month,” Stucke said. That’s why she moved to create a consolidated chart of accounts across all entities to be able to automate systems. But the process of updating all the historical chart-of-accounts activity into the new structure for historical comparison “has been a massive project,” she said.

AI’s time dividend

It would have taken weeks for Stucke’s team, even with help from consultants, to transform the data from the old general ledger structure to the new general ledger structure. The “old way,” she said, would be to painstakingly implement spreadsheet macros or to implement RPA. However, she thought both those solutions would require customization for the countless variations in data format and definitions.

Stucke had a different plan: She tried using her enterprise ChatGPT account to transform the charts. She was able to input the charts and ask the AI to map the data to the new, unified format.

Besides merging data by time period, it could also iron out the variability in how each entity labeled transactions — using generative AI to parse the many ways the charts might label “event income.”

It worked — to an extent. But Stucke also ran into common frustrations with generative AI. After executing the transformation perfectly on 10 charts in a row, it might implement a new and unwanted method on the next.

Stucke struggled, in effect, to convince the software to cooperate 100% of the time. There was also the question of hallucinations and errors, which are common in generative AI software. Stucke hoped to have ChatGPT include formulas that would show and verify its work, but it simply wouldn’t do it correctly.

Instead, Stucke implemented her own formulas, just as she would with human or RPA products, to check the new, unified ledger against the old data.

“My internal controls process hasn’t changed,” she explained.

Even with the frustrations, Stucke said, the AI approach delivered faster and still-reliable results in days — enabling her to upgrade and continue moving forward with automation of the finance function. And it all happened with a relatively generic, accessible, and inexpensive product.

“I think it’s empowering small and medium businesses to give it a try,” she said of the new wave of AI products. “A lot of small and medium businesses just don’t feel like they have the talent to implement some of these systems. However, you can take a generic product and get automated in a much easier lift than ever before with the right vision behind you.”

FIXING MANUAL PROCESSES

The most valuable AI and automation advancements can’t happen without a unified and accessible source of data.

That’s been a key focus of Lawrence Amadi, ACMA, CGMA, a partner and head of the technology risk business for KPMG Africa.

His clients include some of the continent’s largest telecoms — one with more than 85 million subscribers. Amadi has been working with them to transform their Subscriber Identity Module (SIM) systems, which authenticate and manage the identities of users and their devices on the network.

The company had relied on manual processes to manage that voluminous data. About once a week, staff would download data and check for incomplete or anomalous records.

Those manual checks raise the risk that the data won’t be exported completely or that staff will suffer from “audit fatigue,” Amadi said.

“It becomes very stressful and daunting,” he said.

KPMG Africa’s goal was to automatically export, analyze, and raise exceptions from the same data — ultimately a seven-month project involving people from across the organization.

“They should have product understanding skills. They know the business rules, they understand the product, they understand how the data is produced for this product,” Amadi said. “You need very strong data analysts who can make sense of understanding data, unpacking data, analyzing data, and repacking data.”

The project was built with Automation Anywhere, which describes itself as an “agentic process automation system” that “combines the power of AI, Automation, and RPA.”

“Now, instead of the periodic manual downloads … there’s an automatic export happening on the system. There’s an automatic analysis happening … There are also automatic exceptions raised on the system,” Amadi said.

The new system has led to “reduced errors, greater efficiencies, and better reporting to the board and to the audit risk committee,” he said.

And now that it has sorted through and centralized a major source of data, the company is ready to apply automation and AI in more areas, Amadi added.

---

3 AI implementation tips for finance professionals

Ready to launch your own AI project? Here’s what three finance professionals have learned from their own efforts.

Understand the costs

Mohit Sharma, ACMA, CGMA, has learned plenty of lessons with his startup, Pinaka AI. One is just how much money AI costs to run.

A poorly configured product can consume tokens — a practical proxy of needed computing power and cost — at an alarming rate. Running an AI product also introduces new risks: What will it cost if a generative AI product makes an error or causes a controversy — perhaps by using abusive language with a customer?

“There are layers of cost, direct and indirect,” Sharma said. “The moment you lose visibility, you lose money.”

Those costs and risks must be weighed against a realistic prediction of revenues and efficiencies to be gained from a project, Sharma added.

“Think in terms of finance,” Sharma said. “It doesn’t matter whether it is OpenAI technology, whether it is Google. You need to think, what is the return to my business? When does the solution break even?”

Build buy-in

Especially at a large organization, an AI or automation project might require combining skills and data from across the company. It can’t happen unless people understand and accept the mission, according to Lawrence Amadi, ACMA, CGMA.

“You need the buy-in from all of them,” Amadi said. “The why is crucial.”

That “why,” he added, might range from improved coverage of controls to better visibility of the business. No matter which, “you won’t get any progress” if the purpose and means of the project aren’t clear to everyone.

Enable others to learn

Janice Stucke, CPA, used generative AI to complete a daunting data transformation in just a few days, all by herself.

“I was able to accomplish in four to five days what would have taken a whole team of consultants and my staff two to three weeks,” she said.

But that efficiency came at a cost: “My whole team didn’t get taken on that journey,” Stucke said. In other words, her team didn’t get to practice using AI or to see just how and why it was so effective.

Mastery of AI will continue to be a crucial skill for finance teams, especially as automation consumes more work — so getting everyone involved now could help to future-proof their careers and to accelerate the transformation of the finance function.

---

About the author

Andrew Kenney is a freelance writer based in Colorado. To comment on this article or to suggest an idea for another article, contact Jeff Drew at Jeff.Drew@aicpa-cima.com.

---

LEARNING RESOURCES

AI-Powered Excel: Leveraging AI and ChatGPT for Supercharged Productivity

This course will have you streamlining your Excel work, research, and documentation, saving time and effort in your day-to-day tasks.

WEBCAST

Agile Finance Transformation Certificate Series

This training is a result of the partnership between Oracle and AICPA and CIMA and is built around the agile finance operating model developed by the Association and implemented by Oracle: operational excellence, digital intelligence, and business influence.

COURSE

The AI Advantage: Leveraging AI for Efficiency and Impact

In this fast-paced session, discover how AI-driven tools and strategies are revolutionizing workflows, improving decision-making, and driving business impact.

WEBCAST

For more information or to make a purchase, go to aicpa-cima.com/cpe-learning or call 888-777-7077.

---

MEMBER RESOURCES

Articles

“Report: AI Speeds Up Work but Fails to Deliver Real Business Value,” JofA, Jan. 14, 2026

“Transformation-Focused Companies Are Outpacing Others in AI,” FM magazine, Nov. 10, 2025

“Professionals Worry Their Organizations Could Fall Behind in AI,” FM magazine, Aug. 8, 2025

“Leveraging Gen AI to Maximize Performance,” FM magazine, July 7, 2025

“Why and How Finance Leaders Need to Steer AI Adoption,” FM magazine, June 26, 2025

“What Agentic AI Will Mean for Finance,” FM magazine, June 20, 2025

Report

Leveraging Generative AI: AI-Human Co-creation for Tasks Requiring Social Intelligence, AICPA and CIMA, March 2, 2025

---

AUDIO RESOURCES

Listen to Mohit Sharma, ACMA, CGMA

Listen to Janice Stucke, CPA

Listen to Lawrence Amadi, ACMA, CGMA