Toronto-based FinTech firm Float Financial wants to save bookkeepers more time with an agentic AI offering for corporate credit cards that it’s releasing inside of its new product suite.
As of Tuesday morning, Float customers on certain service tiers can now access Float Intelligence. The AI automation layer bundles Float’s existing automation tools with a new transaction coding agent that automatically assigns general ledger codes and Canadian tax codes, like HST, GST, and PST, to transactions made on Float corporate cards.
Ruslan Nikolaev,
“Finance is not like programming or design, where you can sort of vibe code things together; you have to be really precise and accurate.”
Float
Ledgers operate on account lines, which sort purchases into distinct groups like mileage, office supplies, or food. Typically, when a business makes a purchase, bookkeepers have to log the transactions and taxes into the appropriate account lines, which Float says can take hours.
“It’s not often well understood if you’re a business owner: whether you run a big finance team or a big company or not, the volume of transactions is immense, and the defects are huge,” Float co-founder and CEO Rob Khazzam told BetaKit in an exclusive interview, alongside co-founder and head of product Ruslan Nikolaev.
Founded in 2019, Float offers a suite of products designed to simplify expense management for small and medium enterprises, including corporate cards, accounting services, and expense-tracking software. Before Float Intelligence, the company most recently launched transactional business accounts for small businesses seeking an alternative to Canada’s big banks.
Float’s new transaction coding agent aims to eliminate as much of the manual work as possible. During beta testing across more than 350 Canadian businesses, Float said its transaction coding agent was accurate more than 90 percent of the time on auto-coded transactions, and when the agent’s confidence threshold isn’t met, the transaction is flagged for human review.
“We’ve actually been testing it in a beta for quite a few months now, because we want to achieve that high precision,” Nikolaev said. “Finance is not like programming or design, where you can sort of vibe code things together; you have to be really precise and accurate.”
RELATED: Float secures nearly $100 million in debt to expand credit products for Canadian businesses
The key to this, according to Khazzam and Nikolaev, is that Float trained its large language model (LLM) on hundreds of thousands of transactions from Canadian vendors with Canadian tax codes and Canadian general ledger codes. The model is also custom-trained with each business’s historical transactions, so it has historical data for how a user’s books are set up.
Float hopes this offering ultimately shifts small business bookkeeping from line-by-line data entry into a review-and-approve process.
“You get the benefit of all … the other finance teams’ efforts, not just your own, and the system just keeps compounding and gets smarter,” Khazzam said.
This past January, Float secured nearly $100 million CAD in debt to expand the scope and flexibility of its financial products. More than 7,000 Canadian businesses, including fellow startups Cohere, Neo, and Jane Software, are Float customers, the company says. Khazzam told BetaKit that his company’s customer base has grown by around 65 percent year over year.
Nikolaev said that Float has always brought new tech to Canadian businesses in a “very localized way,” and that the company is looking to do the same with AI and LLMs.
“We started Float because we saw nobody was really building for Canadian finance teams and Canadian businesses,” Nikolaev said. “That’s what we’re trying to do in this domain and really bring the future of AI-first finance to Canadian businesses.”
Feature image courtesy Nicole Richard from Wax Pencil Imagery.