Machine learning algorithms are nothing new to QuickBooks, which has been leveraging them more and more over the last few years as part of its drive toward greater automation. However Michael Hitchcock, director of product management of accounting and tax at Intuit's QuickBooks, thinks they can do so much more. To this end, QuickBooks intends to lean especially hard into artificial intelligence this year in service of a paradigm described internally as "All-In."
Hitchcock, by his own admission, said it is an ambitious agenda, and one that acts as more of an aspiration than a specific goal. It's what QuickBooks has been calling its general strategy over the past few years of making step-by-step functional improvements in its ability to automate mundane bookkeeping work by leveraging AI to identify where customers and accountants can focus, and then provide the intelligence to be able to work more efficiently.
"So when we say 'All-In,' it's really about our continued focus on providing automation across all aspects of the QuickBooks platform - from accounting to tax to marketing, payroll, payments, and capital. Where in an ideal future state, someone can come into the QuickBooks ecosystem, connect their data, and accounting is done," Hitchcock said.
How a lot of accounting works now, he noted, is there are manual forms that someone then connects to their bank or accounts payable. Users need to go in and manually review every transaction to ensure the books have the right categories, customers and vendors.
"The 'All-In' paradigm is, 'Hey, how can we flip that on its head? How can we make it so that we're leveraging the power of AI and automation to automate the busy work? ... How can we create the correct accounting journals based on the fact [our machine learning models] have seen this before?'" he said.
Hitchcock is confident QuickBooks can achieve what would be effectively "hands-free" accounting that can run with little to no human intervention. Part of this is due to improvements in the AI training model. Much like a person, AI models need to be trained to understand what it is they're supposed to do. Hitchcock said it used to take around six months to fully train a model before it can be used with customers. Now, however, it not only can be done almost instantly, the training can be done by the customer with the "Train QuickBooks" feature. Instead of spending months feeding data into the AI system, users now answer between five to 10 questions about their business and what they need; the clustering machine learning model does the rest.
"Today we learn the mode up almost instantly. … You answer five to 10 questions and instantly train the model on how your business works," he said.
Train QuickBooks was introduced about a year ago in the U.S. market and is currently being rolled out to other locations.
Hitchcock also touted the upcoming Guided Q&A as another major factor in this drive for AI. Just as Train QuickBooks is used for routine processes that don't require a lot of attention, Guided Q&A is for those complex situations that demand human focus. Built around a computational knowledge engine, it walks people through especially thorny scenarios such as getting a loan or accounting for something ambiguous. Hitchcock compared it to how Intuit's TurboTax product works, except it is at the transaction level.
"Instead of asking the customer to create, for instance, another chart of accounts to handle the principal from a loan separately from the interest, we ask a series of very simple questions a la TurboTax to be able to just do the correct accounting," he said. "Think of it more as an easy button: I'm looking at this data in QuickBooks, this one looks hard, let me hit the easy button."
Hitchcock said that with these two things working together, users can automate all their busy work while leveraging AI to handle the critical matters that require their attention.
On top of this, he said in a few months QuickBooks will be rolling out an AI-guided end-of-month review feature, which not only provides a summary of financial information and what tasks QuickBooks has already automated but also generates curated to-do lists covering whatever tasks have yet to be completed.
As an extra sweetener, he said this tool can also help clients provide better information to their accountants. The model can be set to flag not just balance sheet-impacting transactions but low-context transactions as well. Once flagged, the program can ask the user to provide additional information, which the accountant can then view.
"Anytime we know that an accountant will send an email, a text message, and say, 'What is going on with this?' we can finetune the AI engine to work for both that small business and that accountant so at the end of the month, when they get ready to prep the financials, it has all the documents they need, all the contextual data they need," he said.
End-of-month review is currently available to accountants and will be broadly available to small businesses later this year.
Hitchcock believes that moves like this align QuickBooks with where the future of accounting technology lies. Just as the profession itself is increasingly intertwined with machine learning and artificial intelligence, so too must be the products they use.
During this time, issues of trust and control will start becoming more prominent, he predicted, as these will represent AI's biggest challenge to the accounting profession. Unlike other applications, such as image generation or copywriting, accountants generally demand to know what is happening in the system. They will not accept a black box with no audit trail. This is why, he said, QuickBooks will be aiming to use automation, not to take control away from accountants but to give them even more.
"The challenge is we think of AI as being this black box," said Hitchcock. "You give it an input, it spins for a second, then spits an output. The big challenge we have is we know accountants want ultimate control. They want to be able to leverage the power of AI but to do it in a way that builds trust. So that is what we're really focused on. How do we make sure we get accountants control? How do we make sure they get visibility and feedback? How do they get a peek under the hood so they can finetune this because I want to give them confidence and make life easier for clients."
