Expense AI had to leave its first idea behind

During the COVID-19 lockdown in 2020, Precious Arikeri noticed something about her spending that she could not easily explain. She was at home more, but she was not necessarily spending less. There were online orders, data purchases and transfers between accounts. The money was moving, but keeping track of it was difficult.

“I was spending more than I would usually spend even while I was going to the office,” Arikeri said. “I was ordering a lot, spending more on data because we were home, and buying a lot of things online, but I couldn’t track.” She tried finding an app that could do the job. The products she came across either involved too much manual entry or did not fit the way she managed her finances.

That eventually led to a conversation with Wahab Balogun, a software engineer she had worked with at IT company Vascon Solutions. The pair would spend the next few years trying different ways to solve the same problem. The first attempt was Trackfundx. The second became Expense AI. The difference between the two is largely a story about what the founders discovered did not work.

Trackfundx started with bank accounts

Arikeri and Balogun launched Trackfundx in June 2022. Their starting point was that people often have money spread across several bank accounts and cards. Rather than checking each account separately, Trackfundx would bring the transactions into one dashboard.

The founders did not want users manually entering every purchase. “If I have to come here and write, ‘transfer money for this, transfer money for that,’ it’s really not very efficient, and also time consuming,” Arikeri said. So they looked at open banking. Balogun integrated Mono to pull transaction data from users' bank accounts. At the same time, the founders brought in a machine learning engineer to work on categorising transactions using bank statements provided by users.

They initially offered Trackfundx for free while testing the product and collecting feedback. Paid weekly and monthly plans followed. There was a problem with the data, though. Knowing that someone had spent money was not the same as knowing what they had spent it on.

A bank transaction could show an amount, date and merchant information without giving Trackfundx enough context to categorise it in the way the user wanted. Mono had its own categorisation feature, but the founders wanted more control. They wanted to create categories themselves and allow users to customise them.

“That category didn’t really work with us, because we also wanted to be able to customise ours. We wanted users to be able to customise theirs as well,” Arikeri said. The founders kept working on their own categorisation system. Then the infrastructure underneath the product started creating bigger problems.

The numbers did not always work

Trackfundx depended on open banking connections. That meant the founders were relying on another company's systems to keep their own product running. Arikeri said one problem they encountered involved users being repeatedly logged out of their main banking application after linking an account to Trackfundx. The cost of pulling data was another concern.

Mono was charging roughly ₦200 per API call at the time. As Trackfundx retrieved transaction information, those charges accumulated. For a small startup, the founders could not ignore that cost. “We just felt like we couldn’t build our business on another person’s business, because their pricing could change,” Arikeri said.

The product also had a geographic limitation. Trackfundx had been built around Nigerian bank accounts. In September 2022, Arikeri moved to the UK to study for a master's degree. She suddenly found herself in the same position as before: she wanted to keep track of her finances, but the product she had helped build was not designed for the country where she now lived. “I may have changed my country, but I’m still particular about being able to track my finances, and I’m not able to do that here,” she said. It was another indication that the bank-account approach was limiting the product.

Then they tried a receipt

By late 2023, Arikeri and Balogun were considering what else they could do. Balogun was looking at generative AI. Arikeri was working in the UK insurance industry, where she had seen technology being used to extract information from documents. That gave them another option. Instead of asking a bank for transaction data and then trying to work out what the transaction meant, they could ask the user for the receipt.

The receipt contained the information they needed: the merchant, amount and items purchased. The software could read it and turn it into an expense record. There was a practical benefit too. The product would no longer need to depend on a particular bank or open banking provider. That was important if the founders wanted to sell it outside Nigeria. At that point, adding receipt scanning to Trackfundx no longer made much sense. They decided to build a new product around it.

Expense AI was a reset

Expense AI launched in 2024. Balogun took care of engineering. Arikeri handled product, operations, customer support and user feedback. The company has largely remained a two-person operation. Users can upload receipts, organise expenses and generate reports in PDF or Excel. The product also includes income tracking, budgeting and savings features that were carried over from some of the work done on Trackfundx.

The company says it has more than 10,000 users across more than 100 countries and has processed over 80,000 expenses. Subscriptions start at $8.99 a month. In Nigeria, the price is ₦2,500. Those numbers give the founders something to build on, but they are also still relatively early for a consumer finance product.

The next market they are looking at is business users. Arikeri said the company wants to offer a version that allows business owners to upload multiple receipts and manage their expenses. The founders have also set a target of one million users and more than $100,000 in annual revenue.

For a company that began with a personal frustration during lockdown, the product now looks quite different from what Arikeri and Balogun first built. Trackfundx tried to solve expense tracking by getting closer to the bank account. Expense AI moved in the opposite direction. It asks for the receipt instead. That change came after the founders discovered that the hard part was not simply collecting financial data. It was getting data they could actually use without making the business too dependent on banks, APIs or one country's financial infrastructure.