Bench, a Vancouver-based online accounting firm, has transformed its bookkeeping business with artificial intelligence.Bench
The series: We look at decision makers among Canada’s mid-sized companies who took successful action in a competitive global digital economy.
Whether someone grows up wanting to be a photographer, a writer or to partake in any other skilled freelance or self-employed craft, they generally don’t get into it for the love of filing taxes.
But that’s exactly where Ian Crosby sees Bench, a Vancouver-based online accounting firm, fitting into the work landscape, helping small businesses to stay on top of their financial obligations.
“They start bringing in money and all of a sudden the government is like, ‘Hey, you have to do this very specific math thing or you’re in big trouble,’ ” says the co-founder and chief executive officer of one of North America’s largest bookkeeping services.
After launching in 2012 with Mr. Crosby as the company’s sole bookkeeper working off an Excel spreadsheet, Bench quickly transformed its business practices. “The really big advantage was that we built Bench to be automate-able in the first place.”
In 2015, the company transitioned to an internal app that built in auto-categorization of expenses based on keywords. Earlier this year, Bench built on that with the launch of its artificial intelligence-based auto-categorizer. The AI was trained on five-million historic categorizations that Bench’s bookkeepers had made for clients in the past, and now it is able to identify and categorize 91 per cent of client transactions without fail.
Working in tandem with the app, the two pieces of technology can successfully oversee 95 per cent of transactions, and even in the case that they can’t recognize something, they are able to provide at least one recommendation for a human bookkeeper to definitively categorize.
“All the time we’re automating different kinds of tasks very intentionally and after we automate a task, humans no longer have to do that task and we’ve become more automated,” Mr. Crosby says.
The key to its successful automation is Bench’s consistent, across-the-board approach by its bookkeepers to categorizing expenses, allowing the AI to learn from the humans and replicate their approach. As a result, the company doesn’t do individual categorizations in the way that individual bookkeepers working for themselves might do.
'All the time we’re automating different kinds of tasks very intentionally and after we automate a task, humans no longer have to do that task,' Bench CEO Ian Crosby says.
“We do it the same way every single time and then the AI can learn that because that’s the same for any client,” Mr. Crosby adds.
That level of automation has so far allowed Bench to process more than $32-billion on behalf of small business across North America, as well as realize a 4,373-per-cent revenue growth over a four-year period. The company recently came in at No. 7 on the Deloitte Fast 50, a list of the fastest-growing companies in Canada.
While automation has allowed Bench to crunch more data, and handle more clients, Mr. Crosby says his firm will not be reducing its staff count of about 350. In fact, the opposite is more likely true. As the use of AI increases, he says each bookkeeper becomes more important, not less.
“Theoretically if you’ve done it right, one person should be able to handle more clients. Because there are fewer tasks per client, they can fit more clients into their schedule and that’s where the real savings come from,” he says.
So instead of one bookkeeper handling 25 clients, that person can now handle 50, because the menial tasks of inputting and categorizing data have been automated. It also means that an individual is also able to apply more of a human touch to the customer experience, figuring out exactly what they might need.
Mr. Crosby says that until some “crazy AI” can automate empathy, that’s not going to change.
“Until then, robots aren’t going to be stealing any jobs because especially in services, the job is the person on the other end getting what they need and AI can’t make that determination,” he adds.
In addition to each bookkeeper taking on more clients, one of the other advantages is that Bench can pay its bookkeepers more, as well as invest in greater technology and keep its prices down. While 95 per cent of its new business is still in the United States, where it has operated since 2012, Bench has had an increasing focus on Canada since last February, and dropped its rate to $145 per month if paid annually from $199.
Gunnar Carlsson, founder and president of predictive technology company Ayasdi, which works with financial clients such as HSBC and Citi, says that AI can be especially useful for mid-sized businesses where manpower is sometimes at a premium.
“The goal is always to take everything that can be automated and automate it, so that you can free people up to be better at and do other things,” he says from Palo Alto, Calif.
Within something like accounting, he adds that AI has particular promise for anomaly detection – looking for irregularities. Mr. Carlsson adds that Ayasdi helped HSBC get on top of a particular problem it had with money laundering, using AI to take on a big chunk of that work by reducing all the false positives that its investigators previously had to chase.
He also notes that HSBC hasn’t felt the need to reduce the number of investigators that it employs to look into money laundering.
“They still have the same number of investigators, but they’re able to function at a much higher level because they’re not being confronted with all this suspicious activity or these reports that suggest that someone is performing money laundering, when they’re in fact not,” he says.
Mr. Carlsson also says that AI is a good fit for many different situations in the banking and financial sectors, and could well prove useful for things such as risk modelling and to assess the risk of loan defaults.
And while some companies might not be quite as gung-ho on AI as Bench, Mr. Carlsson says he can see that changing in the near future.
“There have been a couple times before where people have talked about AI and it didn’t really pan out, but I think this time it’s clear it’s going to because the scale of things is to where it’s really needed,” he says. “There’s still some people who are dipping their toe in the water and aren’t fully jumping in yet, but they will soon, in my opinion.”