The accountant's intro to AI, RPA and intelligent automation

RPA benefits

Enterprise resource planning, treasury, finance and accounting systems for the back office haven’t changed significantly in years. While consumers are reaping the benefits of advanced technology in the form Amazon shopping recommendations, instant re-routing through traffic via Google or automated personal investment guidance — finance & accounting (F&A) staff are stuck with manual processes, Excel spreadsheets or user non-friendly interfaces in an enterprise software system. Yes, we have ERP’s in the cloud which is helpful, but these are really utilizing the same underlying approach — rigid business rules — that don’t fundamentally change the intelligence driving the software.

Thankfully, advancements in intelligent automation, increased computing options in cloud software and a realization that F&A staff can use some help too are driving companies to create solutions that fundamentally change how F&A work gets done.

This article describes the core components of intelligent automation with specific examples of use cases that can help F&A staff work smarter, close the books more efficiently and spend more time doing higher value work (or maybe just not working on the weekends).

Robotic process automation (RPA)

RPA is a technology that essentially records what you do - kind of like an Excel macro on steroids - and replays that set of specific tasks in the specific order recorded. Platforms like Blue Prism and UI Path have made the creation of the “bots” easier by giving users an interface to help “program” the manual tasks that you want to have a bot do. In the accounting world, RPA is useful for tasks such as logging into a system, automatically sending an invoice to a customer or automating some manual processes in AR.

The benefit to F&A staff of RPA is that it can help increase the speed of the processes you already have as long as they are repetitive, discreet and don’t require human assessment. RPA literally replicates what you do, but does it faster and by a robot. This can be helpful for certain F&A repetitive tasks. RPA can also be cost effective to implement for a sophisticated user.

The downside is that the fundamental process hasn’t changed — it’s not different, just faster. There is no intelligence to help you make decisions and by implementing RPA you are also locking yourself into what Gartner calls “technical debt”. This means that you’ve institutionalized a process even further without improving the underlying approach, which makes changes to process or upgrading to new technology that much more difficult and expensive in the future. Also, companies with numerous “bots” often require hiring additional IT staff to manage their “bot farms.”
VC investment artificial intelligence chart

Artificial intelligence

AI is a term used to encapsulate many different technology approaches to emulating human intelligence. Natural language processing, machine learning and deep learning are all specific applications under the umbrella of artificial intelligence. Most of the AI applications we see in our daily lives (thank you Netflix Recommendations!) are either machine learning or natural language processing.
Sage's Pegg chat bot mobile phone interface

Natural language processing (NLP)

Pictured: Pegg, a chat bot developed by Sage
NLP is a branch of AI that helps the interface between humans and computers using natural language. NLP helps understand and interpret words to be used in different processes. If you’ve ever visited a website and there is an automated chat bot that responds to your questions — that is an example of NLP at work.

In the accounting world, companies like AODocs are extending their NLP and ML capabilities to take some of the pain out of invoice management by automatically capturing information from invoices and triggering the appropriate workflow. These types of solutions can greatly reduce or eliminate manual data entry, increase accuracy, and match invoice to purchase order.

The challenge with NLP is that understanding the nuances of human communication is hard — especially without tone, body language or facial expressions. This is one of the reasons NLP is a very helpful technology for legal, accounting and other non-emotive documents.

Machine learning (ML)

ML is the current work horse of artificial intelligence. It is a branch of AI that use known data - i.e. prior data with known answers, also called “ground truth” data - to learn, identify patterns and then make decisions based on that learning with minimal human intervention. ML platforms can learn from additional inputs to “tune the model” to be more accurate with predictions and outputs. The ML model is an algorithm that is adjusted to “fit” specific use cases.

ML combined with use-specific software is starting to be applied to F&A problems like account reconciliation. Companies such as Sigma IQ are using ML to build next generation matching reconciliation engines that use a known set of matches from a previous period to build the relationships and the algorithm to predict matches from the same data sources in future periods.

Intelligent process automation (IPA)

Much like artificial intelligence, IPA encapsulates a series of technologies and workflow processes that use automation to replace manual, repetitive human activities with machines. McKinsey outlines five core technologies that make up IPA: smart workflow, machine learning/advance analytics, RPA, NLP and cognitive agents.

Smart workflows are process management software that sits on top of multiple process to integrate tasks across humans and machines. An example would be a process to automate account reconciliation by combining RPA and ML: using RPA to pull data, a machine learning platform for the matching reconciliation and additional automation to prioritize and post correcting entries.

Cognitive agents are AI based technologies that combine ML and NLP to build a virtual “workforce” or agents that is capable of executing tasks, communicating, learning from data sets and even making decisions based on “emotion detection”. Many of the chatbots that you experience on websites from companies like Drift fall into this category.