In an era of widespread automation, it’s hard to believe manual data entry still constitutes a sizeable chunk of employees’ chores. According to a recent Salesforce survey, people at small and medium-sized businesses spend 23% of their workday manually inputting data. Another study — this one conducted by SoftwareAdvice.com — found the time sink to be particularly acute in sales, where around 71% of reps say they copy data by hand each day.

While a silver bullet remains elusive, San Francisco-based Bill.com believes its new AI suite, which formally launches today after a lengthy pilot, will eliminate manual entry in at least one domain: payments. The company today took the wraps off the Intelligent Business Payments Platform, an end-to-end financial workflow automation toolset designed to streamline payment processes for Bill.com’s more than 3 million members.

How? Well, in part through an AI agent dubbed Intelligent Virtual Assistant, or IVA, that automatically tenders invoices and kicks off the approval process, expediting it by a factor of two to three compared with manual methods. The company says the machine learning algorithms underpinning IVA, which were trained on more than 52 million bills and invoices handled over the past year, can automatically capture data from invoices and spot human errors. Moreover, IVA is capable of recognizing bill approval and thresholds for payments, routing workloads, and automatically creating business rules personalized to customers and payees.

In the course of invoice approval, IVA’s machine learning algorithms attempt to predict with “high confidence” all required fields, including the invoice number, invoice date, due date, vendor name, and amount. It presents predictions for review prior to the bill’s submission, with colored boxes indicating auto-filled data.

Bill.com AI

Above: Bill.com’s revamped payments dashboard.

Image Credit: Bill.com

IVA currently has a few limitations, chiefly an inability to recognize foreign currency or create separate bills from multipage documents. But Bill.com founder and CEO René Lacerte expects it will still save customers thousands of hours of accurate data entry — the equivalent of over 35 business days per year, on average. Lacerte cites a Bill.com survey indicating that 80% of the $58 trillion paid between businesses each year in the U.S. involves paper checks. A separate report published by Concur found that it takes an average of 20 minutes to complete a single expense report and about 18 minutes to correct an erroneous one.

“Bill.com has focused on developing new technologies that help SMBs grow. Automating the back office is a huge industry-wide need,” said Lacerte. “Our new intelligent platform, which is the most significant update to Bill.com since its inception, is built on 10 years of experience managing business payments and hundreds of millions of bills and invoices to train the AI. Increasing the speed and ease of payments will help businesses get ahead.”

The Intelligent Business Payments platform also allows customers to set up autopay for recurring bills and supports new payment methods, including international payments and virtual cards. These typically shave one to two days off a vendor receiving payment, Bill.com claims.

“Any company that’s still making bill payments the old-fashioned way really needs to innovate their back office,” said Gary Hornbeek, vice president of finance at Quicken. “The Bill.com Intelligent Virtual Assistant saves a great deal of time for our finance department. When invoices come in, IVA automatically processes and codes the invoices in the system with vendor name, invoice number, amount, and due date. This can drastically reduce hours spent on manual entry.”

Bill.com — which now moves more than $60 billion on an annual basis — isn’t the only payments company that’s tapping AI to save employees time and labor.

AppZen, which raised $35 million in venture capital last October, offers an expensing tool that integrates with Workday, Oracle, Coupa, Concur, and other reporting services to assign employees a score tabulated from each report, receipt, and travel document they’ve submitted over the course of a trip. Taking into account hundreds of internal and web data sources, it recognizes and authorizes reimbursement for items within clients’ approved list and highlights possible infractions that warrant a closer look.

Startups like Conciliator, WorkDone, CumulusPro, Xtracta, and DocParser also offer fully automated invoice services, as do well-established players like Google Cloud (in partnership with Taulia) and Automation Anywhere. Research firm Markets and Markets anticipates that competition will drive the AI in financial tech market from $1.27 billion in 2017 to $7.28 billion by 2023.