Between those two works of fiction, and years before Newsweek magazine’s 1965 feature on “The Challenge of Automation,” economist Doris E. Pullman accurately predicted the benefits, costs, and inevitability of automation, including today’s hottest version—robotic process automation (RPA). “Probably the most important effect of automation will be the employment changes,” she wrote for the August 1, 1958, issue of the Journal of the American Association of Industrial Nurses. “Automation will reduce the dangerous, monotonous, heavy, fatigue-producing production jobs.”

Automating the “danger” out of today’s financial processes may be the only note that sounds a bit anachronistic when applied to RPA, unless you consider the danger to finance professionals of catastrophically bad business decisions based on faulty data handled incorrectly by humans. But certainly the financial processes—90% of which are suitable for automation—can be mind-numbingly monotonous, fatigue-producing, and, as a result, slow and error-prone.

Click to enlarge.

RPA, according to the technology consultancy Gartner, is a set of advanced technologies that can be programmed to perform a series of tasks that previously required human intervention. McKinsey has described it as “taking the computer out of the human.”

RPA can be used to automate virtually any software applications. This includes financial planning and analysis (FP&A) or corporate performance management (CPM) software. For example, RPA bots could be programmed to log in to your CPM solution, run a complex calculation, consolidate results, and then export this new data. The same bot could then be programmed to log in to a data warehouse and import this new data to be accessed by your business intelligence (BI) software application.

While some FP&A software incorporates elements of automation, RPA and FP&A aren’t at all the same. It’s like the difference between self-driving technology and the car itself. The self-driving technology automates the processes a driver would have to control (speed, braking, direction, etc.), while the car is the technology that’s being automated. Some people enjoy driving, but for many others, driving is a boring but necessary evil.

The clear benefit of RPA is in taking dreary, repetitive, error-prone manual tasks out of the hands of sometimes bored and inattentive humans and giving them over to computers, or bots, which operate much faster, never tire or get bored, and don’t make simple math errors if programmed correctly.

“The [bots] should add to the economic and personal status of individuals who will take on the new highly paid skilled jobs of engineering, controlling, maintaining and repairing (them),” Pullman wrote in 1958, except she called them “new machines” instead of bots. In other words, automation will free people from drudgery, enabling them to enjoy a better, more lucrative work life performing more valuable work.


When RPA, sometimes called smart automation or intelligent automation, is used to automate financial processes, it solves three fundamental problems humans face when doing the same tasks. It breaks through bottlenecks in financial departments; it takes over processes that have a high incidence of human error; and it scripts processes that finance professionals find tedious, time-consuming, and of low value.

Because bots relieve highly trained finance professionals of some of their most tedious and time-consuming tasks, those professionals have more time to engage in meaningful analysis. They can understand the stories the data is telling, rather than simply reporting the numbers.

RPA comprises two distinct components or levels: routines and licensed bots. Routines are specific scripts—like computer code—that you write to perform certain tasks, such as logging in to a system, collecting specific data from an identified source, or running a process on that data.

In finance departments that still rely on manually assembled spreadsheet-based systems, a finance professional uses a giant spreadsheet workbook made up of dozens or hundreds of individually linked spreadsheets. These highly trained professionals create macros—a set of stored functions that automate and make some processes repeatable (i.e., take the sum from cell E17, add it to the number in cell B24, and enter the result in cell C88). RPA routines automate this system of macros. That’s part one.

Part two involves the licensed bots—sophisticated software tools that can execute the scripts you’ve written for them without human intervention. For example, they might replace a manual process in which Mary, a finance professional, comes into work Monday morning, logs in to the finance system, and initiates a process to pull specific data from, for example, the enterprise resource planning (ERP) system into a reporting tool. Then she waits for that process to be completed.

Once it’s completed, Mary has to spot-check the results to validate the data, looking at a handful of data points to make a determination of whether the data is correct. She then initiates another process, perhaps consolidating values; waits for that to happen; and spot-checks those results. Then she initiates the report generation process and samples those. Finally, she distributes those reports to the executives who need them.

In this scenario, Mary has spent the first half of her Monday tediously generating and delivering a report that will help executives make business decisions. It’s likely she doesn’t feel that her work so far justifies the time, expense, and effort of her years of training.

In an organization with an RPA bot (or multiple bots), that process can be programmed to happen automatically at any time—like midnight Sunday night. The bot initiates, logs into the finance system, fetches the data from the ERP system, and loads it into the reporting tool.

When data is loaded, it will run tests instantaneously on many thousands of data points to identify and flag any errors and anomalies. Mary might have caught some of them, but many would have gone unflagged. The bot runs and validates the reports and tells Mary about the anomalies. She is then able to investigate and resolve them before approving distribution of the final report, all within minutes of arriving at work on Monday morning.

Simultaneously, the bot might be overseeing or directing other apps to load data into the reporting tool. For example, it could extract currency exchange rates from bank sites to load into the reporting tool, initiating multiple currency conversions for the report.


All robots, from Gort to Rosie to RPA bots, do what they’re programmed to do. They execute scripts. If the Jetsons want their domestic helper Rosie to wash the dishes, there’s a scripted process the bot must run.

  1. Access an application.

In this example, Rosie is programmed to run her dishwashing script at a specific time, likely after meals. When her internal clock hits the appointed time, Rosie accesses the dishwashing application, which tells her the steps she must follow to achieve the task.

She opens the dishwasher and runs an analysis to determine whether it’s empty or full and whether any dishes inside are clean or dirty.

If the Jetsons also want Rosie to do the laundry, a laundry script would have to be written. Because a single bot can do only one thing at time, a second bot would be required if the family wants laundry done at the same time dishes are being washed. If the tasks can be done sequentially, a single bot can be programmed to proceed from one to the next. Because RPA bots are licensed tools, maximizing their efficient use is essential to maximizing return on investment.

  1. Apply if-then logic.

Many of the steps bots follow use if-then logic. If the dishwasher is empty, load dirty dishes. If the dishwasher contains dirty dishes, load additional dirty dishes. If the dishwasher contains clean dishes, put the clean dishes in their appropriate cupboards, then load dirty dishes into the dishwasher.

When dirty dishes are loaded, analyze soap dispenser. If empty, go to cabinet where dishwasher soap is stored. Retrieve soap. Add dishwashing soap to soap compartment.

In a finance RPA situation, if-then logic might be scripted to check if a data cell is empty and then fetch data from another source. Or if there is an anomaly, send an email reporting the anomaly to a human for adjudication.

  1. Navigate with an app.

The RPA bot can navigate within an application, such as to pull data from other sources and act on it within the application. For example, check the sensor on the soap dispenser.

  1. Open files and attachments.

The RPA bot can open files or attachments sent to it from other applications and act on that data. In typical scenarios, a bot might be getting five or six data feeds from the general ledger, the ERP system, an HR system, accounting information systems, a customer relationship management system, and perhaps a production system on the shop floor.

  1. Copy and paste data.

The bot can copy and paste data, moving it from one app to another, such as from data files or attachments received from other applications. When importing data from other sources or copying and pasting between data sources, some anomalies are likely to be introduced. RPA bots can read structured data and highlight anomalies. A bot can perform thousands of checks on the data and flag anomalies that need to be addressed.

  1. Restructure data.

A bot can restructure data to make it consistent. For example, it can retrieve a currency exchange rate from a bank website, provided that data is on the web in a structured location. The bot then can apply the exchange rate as directed. It can pull in the monthly rate, period-end rate, and more.

  1. Prepare and send email.

Once the data has been assembled, manipulated, and scrubbed for anomalies, the RPA bot can then prepare and email the final report.


Most FP&A software will streamline some of the repetitive tasks that the finance department must process. Unlike machine learning and AI, which organizations are also using in part to automate workloads, RPA doesn’t strive to learn or to solve problems. It’s governed by set business logic and structured inputs and doesn’t deviate from its rules.

While an RPA bot can import data from an external system, it can’t do the external system’s job. Just as Rosie the Robot could initiate a laundry process, she would have to rely on the washer and dryer to wash the clothes to the completion of that cycle and to determine if the clothes are actually dry.

When another application communicates to the RPA bot that the application has finished the task it was assigned and has reported the result, the bot takes that result and carries on with its scripted performance.


The main benefit of RPA is in freeing up highly trained staff to do work that adds more value rather than grinding through the same tedious numbers gathering they’ve been doing week after week for years. When you have humans doing a job they don’t hate, you have fewer errors and a much faster process. Bots don’t have feelings. They’ll do the same job quickly and accurately 10,000 times without complaint.

Here are three examples where the finance team will benefit from RPA.

  1. Automate the month-end close: Most CPM software vendors have some workflow built into their application, but to really add to the value of the automation, the software needs to incorporate other tasks, such as data loading and validation, to automate the month-end close process. Within the workflow, you may need to automate the use of exchange rates, import general ledger data from multiple sources, and include intercompany transactions. Once this data is in, the process to convert currency, prepare intercompany elimination journals, and calculate critical key performance indicators (KPIs) can be automatically initiated as a scheduled overnight task. This enables the finance analyst to shift focus from data gathering to data analysis.
  2. Automate report preparation and distribution: Preparing reports for meetings and/or distribution is another repetitive, low-value-add task that should be automated. Reports should automatically incorporate user-access restrictions when they’re generated. For example, a simple income statement should be usable across multiple departments or entities and, when viewed, should show a user only the departments or entities they’re allowed to access. Piecing together multiple reports in a binder that can be sent to individual users or groups is an essential requirement that RPA bots can take over, ensuring financial analysts no longer waste time collating and distributing hard copies of reports.
  3. Automate the pre-population of budget data: RPA bots can also be used to pre-populate upcoming budget documents with values based on overall corporate guidelines. This gives budget contributors a head start on data entry and cuts the time needed to complete a budget proposal. With a user-friendly interface, budget contributors can populate the new budget with a combination of prior actual data and forecasted values applying, for example, a 5% increase to revenue, decreased labor costs for specific products, or a generic 2% decrease to all operating expenses. These automated calculations will generate data as the starting point for the budget. Users save time and are able to generate “what-if” scenarios more easily.


The most common concern about the implementation of RPA is the same fear people have always had about automation and the adoption of robots for business processes: They fear they will lose their jobs to robots. In fact, some jobs will change, and some tasks will require fewer people to perform.

The choice is to continue doing those dull jobs inefficiently and imperfectly or move to bots. The choice should be an easy one. The fear of bots can be mitigated with a transition plan for staff who will have a reduced workload because of bots taking over some of the manual tasks they perform. For them, the focus must be on the new value they can contribute once freed from grunt work.

A big part of any RPA adoption plan is educating people. Get them involved early. Have them on the team coding the bots, or if they aren’t able to be coders, have them participate in identifying processes that can be scripted. Making them part of the build keeps them involved and relevant. They identify processes for coders, coders do scripting, and noncoders help with the testing. Everyone is part of the team. Without that involvement, some people may feel threatened and may even sabotage the work in an effort to save their role.

Part of that education and involvement is identifying for staff the career track opportunities that RPA offers. Gartner has identified several different categories of professional expertise needed in an RPA environment.

First, you need a person or team to oversee the whole automation program—people with a holistic view of the organization and its systems, workflows, and reporting.

Bot maintenance is also required. Bots are not 100% set-and-forget tools because businesses are dynamic. A bot may be programmed to pull data from the general ledger (GL), but if the GL managers add 27 new accounts, then someone has to tell the bot to add those 27 new accounts. There’s always going to be maintenance, and, in many cases, especially at the beginning of the process, it’s useful to have people with expertise in the legacy systems and processes.

Also essential for cost-effective use of RPA is expertise in bot scheduling because organizations are paying for each bot license. Timing of the scripts matters because if you’re scheduling them for maximum productivity, you may be wasting a resource. If Bot Two is waiting 30 minutes for Bot One to finish a process, maybe Bot Two can do something else with those 30 minutes. RPA teams will constantly tweak their tools so they don’t have to pay for underutilized bots.

Even if you’re starting down the RPA path without immediate plans to use bots, you’ll want to anticipate that possibility in the future. Maintaining scripts and scheduling the bots running those scripts for maximum utilization is a fairly technical role that traditional finance professionals may not want to do. As an organization starts to move to the RPA path, it should start looking for people who are interested in automation and start including RPA oversight in job descriptions.


With RPA, a vendor might perform a comprehensive analysis and report that the organization needs 37 bots to maximize the benefits of automation. Buy one. Start small. Automate one thing. Tweak that until it’s nearly perfect. Once you have that mastered, go to a second process. When that is fine-tuned, add another. Get some quick wins, and maximize your utilization of that first bot until you see and understand a clear path that justifies a second.

It’s also important to remember that you don’t have to reinvent the internet for every process. If you have written a script that says, “log in to ERP,” that’s something you’ll do many times. Don’t write a new log-in script for every system. Reuse and adapt what you can where you can.

Finally, it comes down to having the right people doing the right things. There are different approaches to RPA adoption that have worked in the market. In some cases, full responsibility for the process automation resides in finance; in others, it’s spread across teams in various departments.

It’s entirely possible that some people won’t have the skills—or the ability to acquire the skills—to enable them to thrive in an RPA environment. The economist Doris E. Pullman recognized that in 1958. Ultimately, bots will replace some people. But many more, especially in finance, will find new opportunities to provide value far beyond crunching numbers and emailing reports.

About the Authors