No matter which line of business you work in, chances are you’ve heard a lot over the past couple of years — if not longer — about how automation is going to change the way you work. It’s been considered the holy grail of enterprise productivity for decades, and with the advent of technologies like Robotic Process Automation, it seemed like the age of real, effective automation might finally have arrived.
For Accounts Payable (A/P) teams, automation promises to eliminate manual, boring tasks and free us all to focus on the things we really enjoy doing — the thorny problems that actually require your attention, rather than the repetitive, low-value grunt work. That’s a hugely exciting prospect.
But with so many ways of applying automation and so many unique use cases to explore, it’s not always easy to know where to start — or even to understand what’s practically involved in automating A/P processes, and in fact whether automation is even the right answer to the challenges facing the function.
In this blog, we’ll take you through the basics of automation in A/P — answering some commonly asked questions, exploring the technologies available, and helping you define an automation strategy that’s right for your organization.
Let’s get started.
IBM describes automation as an application of technology, programs, robotics or processes to achieve outcomes with minimum human input. Put simply, it helps you replace tedious and human error-prone tasks with automated processes.
There are several good reasons why businesses choose to deploy automation in general – and a number of different ways for them to do it. Generally, they want to:
Increase productivity by reducing the time it takes to do something.
Reduce manual work to focus employees on more high-value tasks.
Reduce defects and human error, which increases quality and avoids rework.
Reduce variance to better ensure consistent standards.
Cut costs by requiring less time and resources to get things done.
In A/P, automation is most commonly used to simplify and streamline time-consuming tasks and processes like invoice processing, payment approvals, and all kinds of data entry. It’s not replacing your people — just giving them more time to focus on the tasks they’re best suited to handle.
Practically however, automation is very much an umbrella term that encompasses several different types of automation — and they’re not all equal. Here’s a quick overview of the main types of automation used in A/P, and the advantages and drawbacks associated with each of them.
Most of us forget that excel macros are as much a type of automation as real-life robots. The different types of automation we cover here are primarily defined by the level of complexity of the problems they can solve and the scope of the area they cover, from automating processes within existing software, to end-to-end business workflows and beyond.
When we think of automation, most of us will naturally start to imagine sophisticated AI and machine learning technology. But in practice, A/P automation can be as simple as a spreadsheet formula.
Spreadsheets are still widely used to collate, store, and process vast amounts of accounting data. Used to speed up the process of manually processing spreadsheet data, formulae provide a simple way of improving what are otherwise very labor-intensive — and frankly, unengaging — tasks.
In A/P, spreadsheet automation is widely used for reporting — helping teams quickly calculate metrics like average days to pay, or paid-on-time rates. Figures are automatically shared between sheets, and complex calculations are processed instantly, right in the spreadsheet.
Anyone can go in, change some figures, tweak some formulas, and generate the insights they need in seconds. And therein lies its downfall. Spreadsheet automation is completely dependent on manual human input, and incredibly vulnerable to human error.
Carefully crafted formulae can be broken in an instant — sending entire reporting structures into disarray. And when data is automatically shared across sheets, one incorrect entry quickly becomes 10 incorrect entries, throwing off every metric they’re used to calculate.
It’s the house of cards of the automation world. Straightforward and great when they work, but prone to collapse when you accidentally lean on the backspace key.
Optical Character Recognition is the automated conversion of physical documents — whether handwritten, typed or printed — into digital formats.
In A/P, it’s widely used for invoice processing, converting paper-based documents into digital formats for more convenient processing and analysis. If you’ve ever been tasked with manually entering physical invoices into digital systems, you’ll understand why this is so valuable for teams that haven’t fully digitized their procure-to-pay processes.
It’s great at automating loathsome manual data entry processes, which gives A/P teams more time to focus on the things they’re really good at. But it’s not without its limitations. Unfamiliar document formats, poor handwriting, and print quality can all impact a system’s ability to convert physical documents into digital data. Studies have found that even the most sophisticated OCR solutions are prone to error — and those errors end up having to be fixed manually, adding to the rework.
When data is misinterpreted from a physical document and flowed directly into your systems, you’ve got a problem. Human error is one thing, but if you’ve got an automated process introducing error into the data your team depends on, things get exponentially worse.
Those challenges can of course be ironed out by ensuring physical invoices are produced in a consistent format, and to a consistent standard. But, as more and more A/P departments move towards fully digitized procure-to-pay processes, the more pressing question is whether it’s truly worth investing in standalone OCR technology today. The days of paper invoices certainly seem to be numbered, throwing the long-term value of OCR into question.
Often used as a cost-effective way to improve efficiency, and a step up in technology from spreadsheets, RPA is an unsupervised type of automation — a hard-coded ‘robot’ which needs rewriting if the end user’s requirements change.
By emulating a click path in an application’s user interface, RPA automates repetitive tasks that would otherwise be completed manually by a human. Traditionally, RPA has focused on simple tasks — what the Forrester Wave Report on RPA calls ‘the rule of five for task automations’. These are tasks that require fewer than five decisions, five accessed apps, or 500 clicks. In A/P departments, it’s commonly used to replicate data between systems and sheets as it’s captured, process it quickly, and streamline routine invoice processing work.
While strong in its ability to automate individual tasks, much like spreadsheet automation, RPA is also extremely fragile. Because it replicates click paths, as soon as any one of your A/P systems changes — on the front or back end — the automation breaks, and must be rebuilt from scratch. And let’s face it, sometimes those changes go unnoticed for days before someone realises something’s broken, and the manual rework accumulates.
This makes it great when automating processes involving systems that are unlikely to change, but extremely limiting in scenarios that involve evolving systems, or regular interface updates.
At its best, automation is flexible, scales with the organization’s growth, and can tackle even complex cases that otherwise would require human judgement.
Take, for example, having to update the price on an invoice. Any number of A/P systems could flag the issue for a human to come and fix. The system knows there’s a problem, but it’s up to an A/P professional to step in, check the PO, recheck the terms of the contract, send an email to sourcing to confirm the change, and manually update the invoice once that’s all done. The issue detection may have been automated, but resolution is a highly manual process, every single time.
With intelligent workflow automation however, workflows themselves can be set up to know which source documents to check when an issue is detected. Confidence thresholds can be defined to help systems determine the likeliest correct price without human input. And all relevant updates can be made automatically, without ever dragging that A/P professional away from whatever else they were focused on before the issue was flagged.
This is the holy grail of automation. It’s the vision of an automated A/P department that we’ve been promised for years — a system that detects, flags, and resolves exceptions entirely autonomously, only ever bothering people for the highly sensitive stuff according to complex business logic.
The Celonis Execution Management System enables intelligent workflow automation as part of execution management. It is a new class of technology that acts as an intelligent execution layer, sitting above existing systems, constantly measuring execution capacity and taking corrective action in real time where necessary — by alerting the right people, or automating intelligently directly in your systems.
It’s unique in that it enables intelligent supervised automation — AI-driven automation with a human-in-the-loop for oversight and exception management as needed — and the user sets the terms of engagement.
The fact that it automates the back end makes it more resilient and scalable than RPA, giving it a key advantage when it comes to identifying and removing execution gaps. It isn’t hard-coded, and therefore doesn’t break the way RPA does with an interface change either.
EMS brings together process mining, machine learning and automation in an intelligent execution layer designed to optimize performance across the enterprise. It is used across the organization to free up working capital, reduce operating costs, and increase productivity.
Although often considered a back-office function, A/P has an undeniably critical impact on a business’s bottom line. This was never more obvious than during the COVID-19 pandemic, when finance leaders rushed to free up working capital to ensure the survival of their business — and that of their suppliers’ — using the fastest levers at their disposal: Accounts Payable and Accounts Receivable.
But because each invoice contains several points of data that need to be manually recorded from the point when an invoice is received, through to when it’s paid, A/P processes are hugely labor-intensive. And that’s before you even consider the risk of human error which can mitigate any cost savings you’re trying to make, and have a major negative impact on working capital.
It’s why A/P needs automation.
By eliminating routine, manual tasks — such as invoice approvals and matching, invoice data entry, and payments — automation can free your people to focus on higher value, strategic work that requires human judgment.
Popular examples of A/P automation include:
Processing and categorizing received invoices — funneling the complexity of invoice formats into a single categorized workflow, eliminating the need for each invoice to be scanned and its data captured.
Routing and approval — clearing process roadblocks by automatically sending invoices to the right person for approval, supported by single-click validation to improve efficiency.
Data capture and validation — accelerating data entry and eliminating the risk of human error.
Invoice matching and verification — automatically pairing invoices with purchase orders and supporting documents, eliminating human labor which can cause bottlenecks.
General ledger coding — ensuring every figure is entered quickly and accurately across the organization.
There’s a fallacy that a greater focus on automation will lead to people losing their jobs. While it’s true that automation can eliminate routine, manual tasks, it doesn’t mean that it will put people out of a job — far from it, in fact.
In reality, the time that would have otherwise been spent on basic, time-consuming tasks can be used instead to focus on more challenging, value-adding tasks that require human skills and expertise.
That is, if automation is applied correctly in the first place.
Without a comprehensive understanding of how your existing processes work before applying automation techniques, you likely won’t see the full benefits and could even risk overcomplicating your processes. Automate a broken or dysfunctional process, and you’re just replicating inefficiency at speed.
It’s why a well-crafted automation strategy is needed before you take the first step.
Choosing the right Accounts Payable automation strategy doesn’t need to be complicated. In fact, the more straightforward you can be with your requirements, the easier it is to pick the most effective starting point.
For example, one of the key questions you need to ask is: “are you using one or more standard ERP systems and do they have an API or microservice layer?”. You might not know the answer off the top of your head — but it’s worth connecting with IT to find out. If the answer is yes, then this suggests that you could benefit from an EMS that’s configurable, scalable, and can support complex business logic.
If not — and if your processes to be automated aren’t business critical — a supervised RPA or custom build may be more appropriate. RPA is best suited to front end use cases, due to the need for a clear click path that replicates human activity. But you should also be sure that the processes, actions, and underlying systems you’re applying it to aren’t going to change, as that will throw out the entire click path and break your automation. And unchanging processes are rarer than one might expect, so watch out.
It’s a good idea at this point to decide what you need to automate, and why. As mentioned earlier, having a nuanced and tailored approach to automation will not only help you whittle down your options, but it’ll also produce far better results.
With a well-considered strategy that targets your main sources of inefficiency, you’ll be well on your way to measuring your execution capacity, understanding the right course of action to eliminate execution gaps, and improving outcomes for your entire business.
A/P automation can be a daunting prospect if you don’t quite know where to start. The trick is to start simple, focusing on your most routine and time-consuming processes — especially if you’re already aware of some problematic bottlenecks within them.
If you’re not sure where your execution gaps are — or even if you are — it’s always good practice to leverage a technology like process mining to validate where your biggest inefficiencies lie. Process mining extracts the event log data in your transactional systems to build a breathing, moving picture of your processes as they actually run, making it very simple to identify which execution gaps are getting in the way of your execution.
Having said that, we’ve previously identified the six most common challenges across the A/P process — any of which might be a good starting point for automation.
A good place to begin for example is with electronic purchase orders and invoices. Thanks to ongoing digital transformation efforts, the rise of remote working, and people being more comfortable with digital documents, we predict that paper invoicing has finally had its day.
Not only can automation speed up paper invoice processing (while it still exists), it also opens the door to touchless data capturing and reconciliation — supported by AI and machine learning — accelerating workflows and eliminating the risk of human error.
It’s also worth considering automated approval workflows. One of the biggest barriers to invoices not being paid for on time is the number of approvals they have to go through. By using tools like three-way matching — designed to match invoices to purchase orders and goods receipts without the need for human input — you can save time and money across all lines of business.
These are very simple, basic use cases to start with. Once you have a handle on those, you can move on to more complex intelligent workflows, such as the ones mentioned earlier, and the world is your oyster. Whatever process you decide to automate first, it’s a good idea to choose a vendor that has the experience and data-driven platform to help turn your strategy into tangible results.
According to EY, as many as half of all initial RPA projects fail. Without the right strategy and direction in place, any automation project (no matter how simple) can easily fail to deliver the results your team is looking for.
The beauty of an EMS is how it combines multiple advanced technologies in one complete software-as-a-service (SaaS) platform that’s resilient and offers real-time, actionable insights — and then takes intelligent automated action for you.
With all the capabilities of a business intelligence platform — augmented by process mining — an EMS doesn’t just automate your most critical manual A/P processes, it can help you understand the best, most valuable places to apply automation across A/P.
By delivering that insight and intelligence, an EMS provides self-guided automation – targeting the processes and tasks most in need of automation, and ensuring that you never waste your time pursuing automation purely for automation’s sake.
Find out more about leveraging the EMS for Accounts Payable automation — by discovering the Accounts Payable Execution App, powered by the Celonis EMS.
Insights to inbox - Monthly newsletter
Where Process Mining Enthusiasts Become Business Execution Leaders