At Bench, we’re deeply passionate about client experience. While most people believe “client” refers solely to the entrepreneurs whose books we take care of, at Bench we also associate the term with our teammates (or “Benchmates,” as we like to say) who deliver the technology-supported bookkeeping we provide. After all, we build software products for both groups. This passion for client experience drives us to be continuously exploring ways to improve everyone’s Bench experience, and we’ve found that one of the best ways to do this is through automation. What’s more is that the automation that was introduced early on at Bench eventually evolved into a powerful example of how machine learning can disrupt brick-and-mortar roles such as accounting.
In choosing to automate we tackle repetitive tasks that are error prone when done by people and less so when done by computers. Generally these are tasks that don’t require human judgement or creativity. When done well (and it isn’t always), automation improves both operational efficiency and employee happiness. Furthermore, we take care to involve all users in the changes we make from inception to deployment so that no one’s user story gets left behind.
Automation is a lot more accessible than one may think! At Bench, we’re pretty big fans of trying new things that allow us to shift and adapt quickly based on user feedback. Things like Excel and Google Sheets are incredibly powerful, low-cost tools for automating workflows. Many codified pieces of automation at Bench started as spreadsheet-assisted process improvements. Part of what I do at Bench is build Macros on Excel and Google Sheets that allow our Operations team to turnaround to-do’s more quickly.
Looking at it through a business lens: by keeping automation low-cost to begin with, we’re able to run lots of tests to help us find areas that would most benefit from it. Validation of good ideas in turn makes for compelling business cases for us to dedicate our more expensive Engineering team towards solidifying and fine-tuning these workflow automation tools.
A great example of this is our internal bookkeeping workflow management tool, Flo. It’s pretty complex, with lots of bells and whistles, but at one point was a just a simple Google Sheet.
Automation is exciting still because it doesn’t necessarily stop at the delegation of routine tasks.
At Bench, a core part of what our bookkeepers do to prepare a client’s financials involves processing and categorizing bank statement transactions. In the early days of Bench, the process of categorizing these transactions to their appropriate Income Statement ledgers was painfully manual. Our Engineering team created a tool called Categorizer Rules which enabled the Operations team to create logic that we used for the manual categorization of transactions. For example, we were able to create a rule that would categorize any transaction that read “UPS” to “Postage & Shipping Expense” or all “Stripe” transactions greater than $0 as “Sales Revenue.” Before we knew it, we had assembled a wealth of information on how to categorize transactions that was pretty unanimously agreed upon. This is when things became really, really fun.
Last year, we built a Deep Neural Network based categorizer using the millions of transactions that we’ve already categorized in the past and trained the model to predict how future transactions should be categorized. The model is able to provide at least one suggestion for over 98% of all transactions processed by our team. In essence, Bench has used this technology to remove the part where bookkeepers have to think about what ledger any particular transaction should go to.
Let’s bust a big myth: robots are not going to take away our jobs. When I gave this presentation at the UBC BizTech Tech Talk conference, I was met with several concerns about the future of accounting, and what exactly it is that’s left to do once we automate all the routine stuff.
Our experience at Bench is that automating well takes care and a lot of stakeholder involvement. At it’s best, automation that engages stakeholders in changes to workflow is responsive to feedback from users and can provide the feeling of being able to do more with better tools. This is what we as a company strive to achieve in our approach to automating workflows.
Accounting firms of the very near future will use automation to become more efficient, communicate key insights to clients in closer to real-time, and continue to build their reputation as trusted advisors. By thoughtfully eliminating rote and error prone tasks from workflows, bookkeeping and accounting are evolving into people-centric roles that emphasize relationship-building and strategy.