Robots and AI in Accounts Payable
Since earliest times people have been inventing machines which can replace human labour to perform tasks faster, more cost-effectively and often with greater accuracy.
The evolution of technology continues at an ever-increasing pace, but now the emphasis is different, especially in the world of software systems.
Robots and Ai are the tools of today. They help us tame the complex systems we have built.
The robots in this white paper do not have a physical manifestation though. They are software, not metal or stone, but they do work nonetheless. Automating tedious, repetitive manual tasks which, in human hands, are time-consuming and as a result often error-prone.
Read on to find out which automation tools to pick from the ‘toolkit’ and how they will help you find your AP automation solution.
Short on time?
Or why not listen to our podcast on robotic process automation instead.
Since the earliest times, people have been inventing machines which can replace human labour to perform tasks faster, more cost-effectively and often with greater accuracy.
The First Industrial Revolution was founded on this principle, with extraordinary gains in productivity often powered by steam or water.
According to Wikipedia, the power loom increased the output of a worker by a factor of 50 and mechanised cotton spinning increased output by a factor of around 500.
While the invention of machine tools such as the screw-cutting lathe and milling machine made the mass production of precision metal parts possible for the first time.
In the Digital, or 3rd Industrial Revolution computers have continued this transition away from manual effort to automated process.
Even with simple tools such as spreadsheets replacing hours of labouring over calculators and more complex software running almost every process in an organisation.
The evolution of technology continues at an ever-increasing pace as we move through the 4th Industrial Revolution, but this time the emphasis is different, especially in the world of software systems.
So what is that difference and how can you capitalise on it?
The use of information technology has made modern business operations possible but in many cases, teams have had to adapt to the increasing complexity and rapid changes in software we have built.
This is nowhere more evident than in finance departments.
As the number and complexity of systems has grown, so have the gaps between processes.
These are often filled by manual tasks such as updating spreadsheets, downloading emails or keying the same data into multiple systems to keep them all in sync.
Of all the technologies in the 4th age, robots and Ai are particularly powerful tools to helps us tame the complex systems we have built.
The robots in this guide do not have a physical manifestation though.
They are software, not metal or stone, but they do work nonetheless. Usually automating tedious, repetitive manual tasks which, in human hands, are time-consuming and as a result often error-prone.
Humans use software everywhere in a finance department, but if a software-related activity is frequently performed and you can write some rules around it, it’s a candidate for robotic automation
Software robots alone are great at automating simple repetitive tasks, like posting simple invoice data into a finance system.
But what about exceptions like coding an invoice with split tax codes, or matching a multi-line invoice to the correct receipts and orders?
For that, you need to introduce intelligence – artificial intelligence, Mr Bond.
Artificial intelligence and software robots were made for each other.
Ai has many facets, but some of those which are most relevant to finance teams are:
- Computer vision (the ability to recognise different document types, or to locate the correct fields on a data entry screen)
- NLP (the ability to understand the sentiment and meaning of some information extracted from a document – such as recognising that the tone of a letter indicates a complaint)
- Machine learning (software robots learning from observing the actions of humans to increase their accuracy and effectiveness)
- Data. Alongside software robots and Ai, the third element we need is Data.
Whereas for humans creating audit trails and recording progress and tasks accomplished is simply more manual work, robots are quite brilliant at collecting data which opens the door for us to analyse everything including:
- Current status
- Process efficiency
- Financial analysis
- And last but not least. This is the force that drives machine learning, allowing software robots, data and Ai to combine to give us cognitive automation.
One exciting example of how robotics, data and machine learning can work together in accounts payable automation is in 3-way invoice matching.
Given a set of rules in the form of Ai, software robots will successfully match invoices to receipts and purchase orders, freeing up a great deal of AP team resource.
Some invoice formats though will remain unmatched and these exceptions will inevitably be dealt with by a human hand.
Using a combination of Ai and data gathered by the robots, machine learning can be applied to improve the accuracy of the software robots by observing and understanding which invoices the humans eventually match.
Over time, this machine learning will increase the productivity of the robots – with a corresponding decrease in work for the AP team.
Likewise, the robots can learn about invoice rejection and queries raised to business users to incrementally improve the automation of the whole matching process.
But there is a caveat.
While the software could learn and self-improve by changing its own processing rules, allowing it to do so in an area such as supplier payment, where a mistake could prove very costly, is risky.
No doubt this is why in our survey this summer, when asked if they would allow automated processes to self-improve by machine learning, 30% of the respondents said they would allow robots to modify key financial processed with minimal supervision and 70% said only with total supervision.
For the AP team, the perfect path for an invoice is ‘Straight-through processing’.
This is where an invoice is received, registered via data entry into the finance system, matched with no queries, posted and paid all with no hitches.
Before automation, this perfect path is still a manual straight-through process.
But in a transformed process, with software robots in place, ‘straight through’, means untouched by a human hand – with no manual effort required at all.
And the added bonus is that email invoices can be processed automatically without printing.
eInvoices can also be received directly into the system – which will be increasingly important as Governments throughout the world increasingly mandate the use of eInvoicing.
Whilst automation of the perfect path is completely achievable, there will always be exceptions which cannot follow that path and need to be processed differently.
Exceptions will include invoices with invalid data, invoices with no receipt, non-PO invoices which require coding and authorisation, back-to-back invoices, CIS and all manner of other anomalies.
But the truly transformed process will provide a robotic solution for many of those activities as well as the Perfect Path, assisting to get coding correct first time, issuing chasers and reminders to tardy business users, answering supplier queries, etc.
So despite the caveats, combining robots, data and Ai to make intelligent automation
gives us huge power for process transformation.
Until recently, AP automation has been based on Robotic Process Transformation or as it is also known, Business Process Management (BPM) technology.
Robotic Process Automation (RPA) has now also become a potent tool, so let’s look at the differences between the two approaches.
Robotic Process Transformation
In an AP context, RPT typically includes:
- Document capture (via email, paper scanning, or EDI)
- Data capture (often using OCR)
- Comprehensive but easy to use forms to replace complex finance system data entry screens.
- Content management to store documents and audit trails
- Workflow to create a cohesive, joined-up process.
- Direct interfaces to the finance system to transfer invoices ready to post
Software robots are to be found throughout, replacing manual tasks with fast, accurate
Robotic Process Transformation tends to replace the manual processes which existed pre-automation with a far more efficient, streamlined process which embodies best practices evolved through many years of experience.
Even if only applied to part of the entire AP operation, it is primarily a means of process transformation, replacing the existing process with streamlined best practice.
Robotic Process Automation
RPA, on the other hand, tends to be more on the surface, using software robots to mimic human activity such as opening a spreadsheet, logging in to SAP and transposing data from one to the other, or opening emails, using OCR to read the contents and keying the data into a user screen.
- Opening emails & attachments
- Logging into applications
- Filling out forms and spreadsheets
- Copying data between screens
- Extracting data from multiple sources
- Logic-based decisions
Where Robotic Process Transformation usually transforms to a new process, Robotic Process Automation is more often used at a more granular level to automate elements of the existing process and can, therefore, be thought of as a means of task automation.
The comparison above is very much an oversimplification to give a flavour of the differences.
The real answer is that these types of automation overlap in many areas and undoubtedly complement each other.
While both approaches can bring great benefit, each also has some potential drawbacks.
With its focus on best practice, RPT may not be geared towards tackling intricacies of a bespoke, complex custom-built process while an RPA only approach to automating a full end to end process may lead to multiple disconnected robots and a fragmented system.
Each approach has strengths and weaknesses but these two technologies can work seamlessly together to produce a solution which capitalises on the strengths of both.
Choosing is like having a tool bag.
It depends on the job you are doing; sometimes you need a spanner and sometimes you need a screwdriver and very often you need both to work together!
Which takes us on to the big question.
Robots promise a fantastic future but where to start and which automations are needed?
Before embarking on an automation journey or going anywhere near choosing the automation tools, a thorough understanding of the following is a must:
- Why? Clarify the drivers for automation.
- Where? Look at the organisational landscape.
- What and when? Identify current status and priorities
- Who? List the stakeholders and partners, consider capabilities and resources
- How? Decide on the approach
The answers to all of these may at first seem obvious, but standing back and giving considered thought is essential because understanding even the small factors can make a big difference between success and failure.
Why? Clarify the drivers for automation
There are many reasons for applying automation the obvious one is:
- Cost. Saving cost is often the first immediate result of automation
But other factors may be important too including these:
- Throughput. Will volumes change? When? Is the organisation growing rapidly, planning a merger, changing or adding products or services?
- Supplier relations. Supply chains can often be disrupted by broken AP processes which result in accounts on stop or loss of early payment discounts.
- Fraud. Duplicate invoices or fraud may be costing serious money.
- Human resources. In an economy with very low unemployment, it is getting harder to find the right people to work in the team – or perhaps you need to re-deploy people to work in other aspects of finance.
Where? Organisational Landscape
It is crucial to understand the real drivers behind the automation project, but it is also necessary to consider the organisational landscape.
Look at the organisation from a distance, what are the key features that determine the current and future shape of the process?
There could be requirements for:
- Multiple financial entities with different process requirements
- Trading in multiple countries
- A variety of business units with different rules
- Currency and language support
- Multiple ERP’s and other systems involved in the process
In addition to these factors which make up the organisational background, an honest appreciation of your supplier relationships is crucial.
AP automation is likely to need buy-in from your suppliers, for example, to switch to eInvoicing or to implement a no PO no-pay policy.
Who holds the power in the relationship, you or your suppliers?
Maybe you can persuade your suppliers to help your vision dreaming of a touchless straight through process – but be realistic – it’s no good pursuing that dream if your suppliers are never going to comply.
What and When? Identify current status and priorities
What is the situation in AP now and what are your pain points and what are your priorities?
It may be that much of the AP process is running smoothly and a solution is only required in very specific areas such as matching or non-PO approvals.
An AP process scenario with pain points shown in red
Or it may be that the process has many pain points and requires a thorough overhaul and transformation to allow AP to meet business goals….
A more complex AP process scenario with pain points shown in red
…and in that case, prioritisation will be required, ideally focussing on quick wins to release team members to assist with subsequent phases of automation or taking short cut task automation solutions initially in order to buy time for a larger transformation project.
An AP scenario with pain points prioritised into 3 levels
Who? The team
An AP automation project will involve multiple stakeholders from different disciplines from both inside and outside the organisation, typically including:
- AP Team
- Sponsors, Board
The ability to gain buy-in from all of the stakeholders right from the start of the project is a good indicator of future project success – and is therefore worth an investment in time, particularly around communications.
Communications are absolutely key so it is vital to think about who will manage the message to the stakeholders. Ensure that sufficient resource is allocated for communications.
As well as stakeholders, a successful team will also certainly include Partners.
You are inevitably going to work with partners – using additional software, working with a change consultant perhaps or a software provider.
A good partner will invest their time in understanding your drivers, landscape and team properly.
When you are setting out to change processes a customer/supplier relationship with your software provider is no good. You need a partnership.
Whoever your team is made up, there is one more magic ingredient that can spell success for your project.
Evangelists are advocates in your stakeholder groups who really get it and want to help you achieve a resounding success, finding the Evangelists in your AP team, IT and in the wider Business and then keeping them involved should be a high priority mission!
How? The Approach
Taking a ‘Big bang’ approach can be spectacularly successful, the route to fastest ROI but beware – things can go badly wrong because of a misplaced desire to conquer the world in one go!
If the big bang fails, there will be a subsequent loss of confidence of business, sponsors, board and other stakeholders and it is very hard to come back from that to retrieve a failed project.
A better approach is often to phase the project.
Phasing a project can mean many things and usually requires going back to the ‘landscape’ that was discussed earlier, examining the key features that determine the shape of your transformation process and breaking the project into phases based on priorities.
- Phasing a project can bring benefits including:
- Building stakeholder confidence: success breeds success and unlocks greater buy-in
- Gaining experience: experience on one phase will help you succeed in subsequent phases
- Time to experiment: phasing can allow time to experiment with areas of the process without fully committing money or resource – and experimentation leads to the right route for the long term
- Freeing up resources: Changing a process whilst continuing to keep the day-to-day operation running can be very full on and risks overworking the team with potential loss of key people which sets the whole project back.
- Staging a project gives time for the business, and the AP Team in particular, time to breathe between phases, settle in and perfect new processes
Assessing the pain points and priorities for automation will provide a good base for identifying potential software choices, be they RPT or RPA based, or both but when considering the tools to use for AP automation, think also about the resources that are available within IT and AP.
Who has the time, the skills and experience?
Who will analyse the current process and requirements?
Who will design, build, test, train and subsequently support your transformed AP process?
Some task automation can be accomplished by IT, AP teams or business users but beware that automation can end up getting more and more complex as the exceptions begin to be to get added into the process.
Lots of separate automations can be very difficult to orchestrate.
The more complex the requirement the closer automation gets to being software engineering which, when done properly, requires a specialist skillset with accompanying professional working practices.
Building a solution in-house can work well, but using partners, although appearing more expensive short term, may give you the benefit of a wealth of best practice and experience in AP as well as providing long term resilience to change or loss of key personnel.
For the solution to be robust it needs to be controlled, changed and fixed long after the person who designed it has moved on.
A good software platform can be extended to solve the same type of problems in other areas of finance. It should be the gift that keeps on giving!
With the right software platform in place, AP automation can lead on to automation of other areas including:
- Statement reconciliations
- POD processing
- Credit control
- Sales order processing
- Sales invoices
And in addition to finance, automation in other departments such as:
- Customer Services
Implemented well, robots can have a dramatic effect on the health of the business’s financial processes.
Looking at the bigger picture of the effect of automation, not just in finance but across all professions and types of work, researchers believe that there will be a shift for humans towards areas which are harder to automate and one of those is creativity.
Combining robots, data and Ai to make intelligent automation gives us huge power for process transformation and in turn, frees up our teams to work in more productive and enjoyable areas.
Deciding on where to start your AP automation journey is just the first step!
If you would like to discuss any of these topics in more detail or to talk through where and how to start your AP automation journey, please get in touch, we will be happy to help.
Have any questions?
Call now to find out how we can helpGet in touch