What artificial intelligence, RPA, and other automation technologies mean for the CFO.
When it comes to automation and artificial intelligence (AI), the future is now for the CFO. That’s not because AI is upending the structure of the finance labor market; that effect won’t likely be felt for many years, as we point out in our AI predictions for 2018. And it’s not because finance functions are struggling to hire computer science Ph.D.s with machine learning expertise. That’s a deep challenge for many corporations, but it’s not necessarily on the CFO’s plate.
The future is now because all finance staff, not coders, need to make a big adjustment. Automation is the future, but professionals are the present. If finance teams stick with the familiar tools they know and love, they won’t be fit for purpose for a finance function built with next-generation technology, such as small automation, AI and beyond. So the finance team is on the critical path to the future of the profession.
Not your father’s spreadsheet
For finance professionals, the transition to new automation tools and AI will be like the adoption of spreadsheets on PCs in the 1980s. Programs like VisiCalc and Lotus 1-2-3 sparked a mass transition to electronic spreadsheets that left green-striped ledger paper in the dustbin of history. (Google it.) In the process, accountants and other finance professionals learned a new “language” of spreadsheets. By the 1990s, ERP was also the norm, and every finance professional became conversant in spreadsheets and ERP.
Today, entire departments, including more than a few financial planning and analysis (FP&A) groups, are made up of spreadsheet jockeys. Even the ones that use specialized tools also use spreadsheets to fill the “gray space” between applications: They export data from one application to a spreadsheet, manipulate it in some way, and then import to a second application. We’ve all done it.
But we have to get over it. Spreadsheets are going to be tomorrow’s ledger paper. Finance professionals will need to speak the language of data analytics, robotic process automation (RPA) and AI.
Bilingual in finance and data
The low-hanging fruit is using RPA to automate repetitive tasks, such as transaction processing, and integrating data visualization with analytics. Take our own business. Some of PwC’s tax and audit professionals now use code-based analytics tools and scripting languages to make SQL calls and analyze the data. They then automate that process with RPA tools and deliver visualizations to our clients that, for example, highlight unusual transactions by category over the past 30 days. No spreadsheet required.
They’re not automating just to make a pretty picture. The right visualizations convey stories very quickly, immediately surfacing patterns that might otherwise have gone unnoticed, such as an unusually high number of postings just below authorization limits in a particular month. Of course, “unusual” doesn’t mean something’s wrong. Auditors and tax professionals make their own judgments on the anomalies, once they’re made aware of them. Humans are still in the loop.
Coding and visual design aren’t in the typical skill set of a 2015 accounting graduate. Acquiring analytics-enabled talent will be challenging in the coming years, for all professions.
With more analytics-enabled finance talent, it won’t be long before FP&A departments use data science principles to dive more deeply into data from multiple systems. Today, finance professionals spend half of their time gathering data, rather than analyzing it, according to our Finance Effectiveness Benchmarking Report 2017. In the future, instead of spending time manipulating exported worksheets in pivot tables, they’ll apply critical thinking to the implications of the data and test “what if” scenarios that will improve, say, pricing or inventory decisions.
Finance professionals are numbers-oriented and analytical, but that doesn’t mean they’re skilled in RPA, AI and visualization. They won’t need Ph.Ds. in computer science to learn commercial RPA tools or visualization software. But they will need training.
Those who can, teach
Then, there’s the question of AI. Machines aren’t born smart; they’re born with the capacity to learn. Someone with real expertise has to train them. The judgments that go into making revenue recognition decisions, for example, aren’t strictly rules-based; it requires years of experience to make determinations, considering the timing of payments, knowledge of contracts, legal interpretations in different jurisdictions and other factors.
That’s something that AI can also accomplish—but not overnight. Training machines is an iterative process in which experts give algorithms feedback to help the machines get smarter. The deeper the functional expertise, the more fine-tuned the feedback becomes and the better the machine learns. It took many months for our most experienced lease accountants to train our Data Sieve tool to read commercial leases and check compliance with new lease accounting regulations. Now, we can review documents in as little as a quarter of the time it used to take.
Those kinds of teams—accountants side by side with data scientists—are becoming more common. Finance organizations are adapting to this new staffing and teaming paradigm. And it works: In our experience, we’ve found that CPAs and Ph.D. computer scientists work quite well together. They’re why the future of the finance function comes down to people, not machines.
See our full list of AI predictions here.