While Accounts Payable is a mature standardized process, automating it is not as simple as it sounds. The challenges of Accounts Payable automation are numerous in fact.

Pushing paper around, key-in in data, cross-checking data in several applications (swivel chair tasks, as they are commonly called) lead to considerable pains whose costs are difficult to quantify. They are only revealed in hindsight, once you implement automation.

Manual work leads to high paper volumes, lengthy approval cycles, late payments, weak supplier and buyer relationships, uncaptured discounts and limited visibility and therefore control over spend. These are just the direct pains. We also registered, in the organizations we worked with, over-staffing, low employee morale, people assigned temporarily from other departments to AP just to handle the large invoice volume. This prevents CFOs, CPOs and CEOs from actually focusing on their real work – innovation and growth. Instead they do damage control.

The PayStream Advisors 2017 Invoice Workflow Automation Report lists the challenges that AP professionals report at different revenue segments.

Challenges of AP automation

The challenges of Accounts Payable automation

Grouping and identifying root causes, we boiled down these 8 types of challenges to only three. We see manual tasks as a consequence of the three causes. If all data would be standardized, accurate and structured, manual operations would soon be a thing of the past.

  • Non-standard documents – Invoices arrive in a lot of shapes and sizes: pdf attached to email, paper, xls or even fax. Sorting and getting them through an approval process is time-consuming. Talking to suppliers to change their processes is pointless, even if they are willing to do so, they may lack the finances or the technical skill to upgrade their software tools.
  • Errors, exceptions, approvals and workflows At least 30-45% invoices contain errors. Correcting them involves phone calls, tens of emails and arguing with suppliers. Many don’t have an associated PO – such as general expense invoices. General expense invoices also need to go through an approval process, that often is manual – as in employees taking the paper invoice to their supervisor’s desk with a post it note attached.
  • Unstructured data – For their business needs, companies require all sorts of data on their inbound invoices in order to be posted in their ERP. Many suppliers are unwilling or unable to supply this information (such as PO no or receipt advice no). Chasing data and manually adding it to the ERP is tedious.

Historically, there have been efforts to automate some tasks of the AP process, but results were inconclusive because no single solution tackled all the challenges identified above.


For example OCR technology – hailed to solve all the data capture issues from paper documents. In real life examples OCR-processed invoices achieve a small rate of straight-through processing, around 30%, from scanning to ERP posting, without human intervention. Reasons are multiple – missing information on invoices, scanning failures, PO matching errors, paper print quality etc. What’s more, OCR however sophisticated, cannot support many validation rules that companies require. As an example, for one of our retail customer, we implemented over 130 rules and validations on a single invoice. So OCR, by itself, it’s not a solution.


Hailed as the new go-to technology in automation, RPA is just a tool in the automation toolkit. By itself it has drawbacks such as:

  • difficulty to scale at enterprise level without significant resources (servers, system architects, programmers, business analysts etc).
  • “standard” processes often turn out to be not so standard, requiring tweaks and adjustment that burden the company with an additional level of IT and information governance
  • due to these hurdles, efficiencies take much longer than anticipated

In fact, a McKinsey article states that “several robotics programs have been put on hold, or CIOs have flatly refused to install new bots—even those vendors have worked on for months—till solutions have been defined to scale the program effectively.”


Pure eInvoicing is another effort bringing some level of automation. But in practice, einvoicing means just EDI – electronic data interchange, a simple transfer of documents from suppliers to buyers. So you eliminate paper and save on shipping costs and possibly on data input in your ERP. But validation and control and matching are still done by hand. Another drawback of einvoicing solutions is that it’s difficult to persuade your suppliers to enroll in a network and issue einvoices, many preferring the status quo. According to Billentis, the typical rate of einvoicing adoption in 5 years is about 30% of suppliers, which is too low to register any significant benefits.


ERPs as well as BPM tools proved limited for true automation – there are so many versions, configurations and variants of these softwares that they just don’t communicate well without significant investments in customizations. Add to that specific country reporting financial rules, and things just start to fail.

It’s clear that none of the solutions on the market offer straight-through processing to ERP as well as serving secondary processes such as controls, workflows, payment reconciliation and compliance.

So, where does that leave Accounts Payable automation?

A true RPA solution for AP automation has to deal will all the challenges listed above and span the full process of AP automation, not just bits and pieces.

Tackling AP automation challenges

A solution that covers all formats and all channels

Like it or not, supplies vary in capabilities. Especially if you are not a powerful buyer able to dictate the terms to your suppliers, you need a solution that is able to receive whatever format the suppliers is able to send. This is the reason why most purchase-to-pay solutions fail – they cannot accommodate all formats and channels, leaving the buyer to negotiate with each supplier a standard document format, usually an einvoice format. In practice this is unfeasible: Billentis shows that after 10 years of mandating a einvoicing solution, less than 50% of the suppliers embraced the solution.

For our RPA-based platform it makes no difference how suppliers send the documents. DocXchange can handle hundreds of formats and input channels, including einvoices and even paper and pdf files through our smart OCR tool. Robots capture the metadata, transform the document into whatever format the buyer is able to receive and push it to their ERP.

Of course, the most cost-effective option for all parties is for suppliers to send einvoices – setting the stage also for their Accounts Receivables automation. We have a skilled in house supplier onboarding team which is tasked with supporting suppliers to migrate to einvoicing – this is why our average onboarding rate is over 90% in 2-3 years, much higher than average.

Errors, exceptions, approvals and workflows : A powerful rule engine ensuring matching and content validation

Errors are a fact of life – about 35% of invoices contain at least one error. Aberdeen Group estimates that 82% of invoices have an exception.The way to tackle these errors is to find their root causes, which, in AP departments fall into one of several buckets:

  • human errors – these errors are easily identified by robots and corrected, either by a human or through an automated workflow, also managed by a robot. For example – misspelled information, fields that do not contain the proper data or the data in the required format (for example the order number) etc.
  • data errors – master data frequently presents discrepancies – wrong product codes, quantities etc, often lead to wrong deliveries, differences between POs and goods receipt etc. Replacing a human with a robot and having it execute processes on bad data, only makes things worse and sooner. For these types of errors, robots needs a master data repository against which to perform matching.

In a properly automated Account Payable process, no matter how data was introduced, robots are able to identify format errors and content errors and perform validations against numerous business rules. For example, for one of our retail customers, we taught the robots to perform over 140 checks on a simple supplier invoice.  From VAT totals to discount checks, our robots encompass every imaginable business rule for inbound documents.This scale of automation goes way beyond what simple OCR tools can do.

Robots also do the matching between various documents (PO, invoices etc.) and even against the master data file, so that everything is in perfect compliance and traceability. This saves hours in audit effort.

More, when identifying errors, robots can handle the exception management by themselves – notifying the parties involved and suggesting corrections based on previous data. Of course, some may still prefer to handle exception management manually, but having a daily report with all the documents processed and the specific errors identified is still a huge time-saver.

Together, these validation rules and the master data harmonization tool (DxCatalog) are able to reduce the error rate from 35% to virtually 0%.

Unstructured data: data enrichment, master data harmonization

Finally, simply having robots extract data an input it in an ERP is not enough. While data may be 100% accurate and traceable, and people free from significant amounts of manual labor, companies, for reporting purposes, need extra data to accompany their documents. Our robots are trained to learn what data is missing, search for it and insert it at the right place.

To sum up, neither one technology is a panacea for AP automation. You need to think your AP process step by step, end to end, find out if any of the steps can be eliminated entirely and only then look for a solution that takes care of all your needs. Ideally, a solution that blends RPA with OCR, with einvoicing and easy ERP integrations, all in the cloud – in order to avoid laying more strain on an already busy IT department.