How generative AI can help companies by asking questions, not providing answers

Imagine having a “coach” with deep expertise and knowledge of your company’s priorities, guiding you in analysing data and making informed decisions.

That’s the vision behind Sumitomo Mitsui Banking Corporation (SMBC)’s generative artificial intelligence (gen AI) tool, dubbed “CFO Agent”.

First showcased at the Singapore FinTech Festival (SFF) last November, the refined prototype will be featured again at SMBC’s booth this year.

“We started with a hypothesis: Using relevant data to get correct answers from gen AI to provide meaningful insights for CFOs (chief financial officers),” said Mayoran Rajendra, managing director and head of Asia Innovation Centre at SMBC.

It aims to address a common challenge in digitalisation: Data analysis. This challenge emerged from SMBC’s engagements with over 300 CFOs and managing directors across Asia-Pacific.

“Customers (know) what they want to articulate and what insights they want to get, but they don’t (always) have the right team or expertise to analyse the data,” said Mayoran, who is part of a three-man team that has been working on the prototype for over a year.

BT in your inbox

Start and end each day with the latest news stories and analyses delivered straight to your inbox.

Sign Up

Sign Up

But co-creating the prototype with two bank customers proved complex. Data sets fed into the AI model are often incomplete, and AI models aren’t mature enough to solve complex problems. “Users are also not ready to believe the answers generated by the AI; they want to reconfirm,” Mayoran noted.

Today, the CFO Agent has evolved to act as a “coach” – using available company data to ask context-specific questions, providing guidance on the thought process instead of providing answers, he said.

The prototype, which is co-created with customers, can be used with existing software so customers will not need to install a new tool.

Leveraging the bank’s knowledge in supply chain finance and working capital management, the CFO Agent cross-checks data with a company’s defined priorities to identify areas requiring a deeper look. 

Mayoran explained that this approach, where information is curated and defined, helps to lower the risk of “hallucinations” – where an AI model might present inaccurate information as facts.

How it works

For example, if sales figures don’t align with inventory levels, the CFO Agent might ask: “Why are sales low despite high stock levels?”

Mayoran likened it to ChatGPT’s role as a starting point for assignments in education: “It’s a ‘supporting tool’ for knowledge work.”

The bank’s goal is to continue supporting its customers in their digital transformation journey, and co-creating solutions like CFO Agent to help.

“We’re a bank, not a software company nor a consulting firm,” emphasised Mayoran. “But what we have is proven financial expertise, which we are leveraging with technology to support (our customers’) digitalisation journey.”

This was produced in partnership with the Monetary Authority of Singapore and the Global Finance & Technology Network.

For more stories, go to https://bt.sg/sff2024