Articles
March 26, 2025
Beyond Buzz: A Treasury AI Expert Separates Fact From Fiction
# AI
# Technology
James Kelly, co-founder of treasury AI consulting firm Your Treasury, dives into the ways AI can help finance teams immediately.
A simple realization convinced James Kelly, founder of AI treasury consultancy Your Treasury, of the true potential of AI: “I suddenly realized that things I’d been doing in Excel that had taken days and weeks, I could now do in a matter of minutes,” Mr. Kelly explained in a recent interview with NeuGroup Insights.
- “I realized there were operational issues that I’d suppressed because I had no way of actually solving them, without embarking on a complex IT project. All of a sudden, I was in a position where I could solve these headaches simply and quickly.”
- In the interview, Mr. Kelly discussed how AI is already delivering practical benefits for treasury teams—while cautioning against solutions that promise more than they can deliver.
- “You can add an AI chatbot to anything—but I joke that, in much the same way as you can strap an engine on the back of a raft, that doesn’t make it a motorboat,” he said. “What you’re looking for is something that’s really been designed with AI in mind, rather than having it badly bolted on.”
James Kelly, Co-Founder, Your Treasury
Small tasks, big impact. Mr. Kelly believes that the most promising applications for treasury in the short term will streamline routine processes, including KYC, pre-populating forms, drafting routine emails, automating validation processes and creating weekly reports.
- One pragmatic approach he recommends involves using AI to document and share institutional knowledge. He suggests recording meetings where treasury processes and challenges are discussed, then using AI to analyze the recordings and create a chatbot that can answer questions based on that information. “Do that recording exercise every couple of weeks, and quite quickly, you build up a fairly large knowledge bank,” Mr. Kelly said.
- This approach has already yielded significant benefits for teams that Mr. Kelly has advised. “I’ve had quite a few conversations with treasury leaders who were highly dependent on one team member because they had specific knowledge or skillsets.
- “Having the chatbot takes that dependency away to some extent, which is a relief for the leader as they have cover when that person is unwell, or on holiday. But equally, the specialist is also freed up from all those requests and can concentrate on more meaningful work.”
Separating hype from reality. Many AI solutions in treasury are being oversold, Mr. Kelly warns. “Practically everything under the sun is being labeled as AI,” he said.
- AI still requires human oversight. Machines don’t “magically” understand treasury operations, he said; they need proper integration with treasury management systems and data ecosystems.
- As an example, he said that fraud detection is a commonly touted use case, but he argues it’s irrelevant for most treasury teams. “Most companies don’t have anywhere near enough data to make this work. It’s ultimately a waste of their time to try to build that in-house,” he said. “It would be much better to buy that capability, because they’re going to get a far better return.”
Why forecasting isn’t that simple. Even the much-discussed use case of AI for cash flow forecasting requires a reality check, according to Mr. Kelly. Instead of trying to predict everything for the forecast using AI, he advocates for a hybrid approach that focuses on the causes of surprises and uses AI to identify patterns in regular, predictable transactions.
- “When people say that they’re not happy with their cash forecast, they’re not generally saying ‘it’s off by half a million pounds, and I’d like it to be less than £100,000,’” he said. “They’re really thinking ‘I don’t want to be surprised. I want to make sure that I’ve got confidence in my numbers.’”
- AI’s real value may lie in smaller tasks, such as automating forecast submissions and follow-ups. He said that ensuring the right data is in the right place at the right time is more valuable than AI-driven predictions.
Perfect data can wait. Despite the potential benefits of AI, one question Mr. Kelly still finds many treasurers asking is whether their data is “good enough” to make the most of AI. Here, he offers reassurance that the notion that you need perfect data before starting with AI is outdated.
- “Many of the AI tools are fantastic at quickly improving data quality. You can actually use AI to clean and tag data,” he said. “So, for instance, you could take three years’ worth of bank statements and get everything categorized within half a day, thanks to AI.”
- In fact, there’s a huge amount that you can do with AI to facilitate further use of the technology, he believes. “So, never let perceived poor data be a reason for not getting started on an AI project.”
Getting everyone on board. While the data and tech side of AI is one challenge, success also requires careful consideration of two critical human factors: buy-in and transparency. He suggests reaching out to teams to brainstorm and come up with use cases that will make their lives easier to encourage adoption.
- “When you’re deploying AI, it’s vital that everyone is involved and feels like part of the process,” Mr. Kelly said. “There’s also a great opportunity for teams to come up with their own number one use case for AI within treasury rather than be told what it is by vendors, banks, consultants or treasury publications.”
- He added that transparency is key. Some “black box” AI tools provide answers without explaining how they got there—a red flag in treasury. Instead, treasurers should prioritize automation that enhances visibility and control.
James Kelly’s top AI tips for treasury teams:
- Challenge the hype. Don’t automatically assume that popular use cases like fraud detection or complex cash flow forecasting are right for your organization. Evaluate each application based on your specific needs and data availability.
- Focus on practical applications. Look for ways AI can solve real, everyday problems rather than chasing cutting-edge applications that might not deliver value.
- Start small, think big. Begin with specific, manageable use cases rather than attempting to revolutionize everything at once. Focus on areas where AI can augment existing processes rather than completely replace them.
- Explore process automation. Look for opportunities to use AI and ML in streamlining routine tasks and administrative processes. The biggest wins often come from automating multiple small tasks rather than solving one big problem.
- Build a treasury chatbot. Consider using AI to capture and distribute institutional knowledge, reducing key person dependencies and improving team efficiency.
- Don’t wait for perfect data. Start working with the data you have. AI can actually help improve data quality and preparation for more advanced applications.
- Prioritize transparency. Ensure you understand how AI systems make decisions, particularly in areas that could affect financial operations or reporting.
- Foster team buy-in. Involve your team in identifying and implementing AI use cases, rather than imposing solutions from above. Brainstorm together. And roll out discrete use cases first, helping the team feel comfortable and confident along the way.
Dive in