Digitizing Procurement: A time and place?
The advance of digital procurement is set to continue in 2019.
We ask Productive Procurement founder Richard Bell what procurement managers can do to benefit from new technology, and is there a time and place for RPA vs manual process?
Why are there so many manual activities involved in processes such as procurement, at present?
“I should begin by pointing out that office automation is not new. ERP systems (Enterprise Resource Planning system such as SAP or Oracle) have been with us for decades, as have other tools such as spreadsheets, which were introduced in the 80’s and 90’s. Offices since the 1990’s look very different to how they would have looked 50 years earlier. However, many procurement professionals spend too much time performing manual tasks which add little value to their work. There are several reasons for this.
Firstly, there are a number of companies that do not use an ERP system. The reason for this is a combination of direct cost and disruption. They can be extremely expensive to install, particularly if you opt for one of the big software vendors. More worrying for a lot of companies is the disruption involved in the move from using a range of different systems and spreadsheets, to trying to condense it all into one ERP.
Secondly, there have been a lot of cases of ERP systems being installed and being less than effective. For the companies that do install an ERP system, it is rare that an out-of-the-box format will perfectly fit the processes the company is using.
Successfully installing and using an ERP can be done in a number of ways. You can reengineer your processes, which will involve your users and business people having to change their daily processes; we all know having to change the way you work every day is not easy.
You can create a tailored change to the ERP, but this can be expensive. The expense is derived from the cost of understanding and coding the change and, more importantly, maintenance of the non-standard adaptation. Unless the adapted routine benefits the work of many users, the cost may well outweigh the benefit. Therefore, the manual routine remains in place.
Thirdly, over the last 5 - 10 years there has been a growth in online applications which are not interfaced to ERP systems, resulting in more manual work. If you’re looking up, for example, exchange rates online, you would potentially re-type the data into your spreadsheet, manipulate it, then finally enter it into the ERP.
Finally, there are some tasks that require judgement that, in the past, has been too complex for computers. Imagine that a customer sends in a complaint that says: ‘I’ve been home all day waiting for a technician to come, five hours later, still no sign of anyone, I’ll never recommend your company. Thanks a lot!’.
You may have written a program to assume that if a message contains ‘thanks’ it’s positive but a human would understand the context of the message and recognize that the customer is not happy. It’s extremely complex to write computer code which can handle all of the permutations of natural language.”
What has changed? Why is process automation such a hot topic now?
“There are two factors. The first is the advent of Process Automation Applications which requires little coding from vendors such as UiPath, Blue Prism, Automated Anywhere and Microsoft (Flow). This means business users can automate their own work.
The second factor is the advent of so computerized cognitive activities using a technique called machine learning. Due to the massive volumes of data coming from all our mobile devices, such activities have become very effective in the last few years. Machine learning uses statistical modelling to look for patterns in data. If, for example, you give a supervised training to the computer by feeding it 5,000 comments, classified according to customer sentiment, the computer learns which word patterns correspond to positive or negative sentiment. You then feed in new, unclassified comments and the computer can interpret the sentiment of these new comments, based on the supervised learning exercise.
A large amount of venture capital has been invested in RPA vendors, enabling the development of increasingly sophisticated functions. There are now built in, or native links to the cognitive activities; you can subscribe to text-clustering services from companies such as IBM and Microsoft and the RPA packages have native links into those modules.
Beyond this, programs written in opensource languages such as Python and R can be incorporated into RPA sequences. It’s more complex than just buying services from Google or Microsoft, but with the appropriate level of programming skills, companies can develop RPA sequences better suited to their own requirements. The more you become involved in RPA, the more the business user becomes more like a programmer because it’s a relatively easy way to learn sophisticated techniques.
If you combine the use of RPA software and the increasing sophistication of machine learning, business users can start to automate their own roles so that they are doing less drudge work. They can focus on more value-enhancing activities such as, within a procurement space, developing specifications and ensuring new solutions are socialized with their colleagues and effectively implemented.”
Why would companies leave people to self-automate their roles? Is it not better to engage professional programmers?
“Manual tasks which are performed frequently and in the same way by many people in an enterprise, can best be automated by engaging professional programmers to develop an application. Such applications are better suited to traditional programing methods, which, because they do not rely on the user interface, will always be more reliable than RPA.
But many finance and procurement activities that are time consuming for an individual are particular to that individual’s work. The cost of having a programmer, internal or external, to understand, create, test and maintain the program is likely to be very high, and the ROI very low. In this situation it makes sense to embark on an RPA program.”
What are the challenges companies are facing in digitizing their work?
“The first challenge arises when a department manager does not have a background in, or enthusiasm for digitization, and is unaware of the degree to which their teams’ work can be automated. Such leaders don’t know what they don’t know and, as a result, process automation opportunities are not considered.
For companies which have started working with RPA, a common issue is conflict between IT departments and untrained RPA developers. Such developers are busy setting up pilot projects, but the IT team may be frustrated as they feel they should be applying the discipline that they themselves would apply to developing business applications. For example, professional developers use techniques such as object-orientated development, where applications make use of code libraries. This is something that comes naturally to IT professionals, but being unaware of such best practices, untrained RPA developers develop their own code rather than using a more reliable library.
Another example is security. IT people in their professional training learn how to keep data secure and that might be a secondary factor for someone who is just getting started with RPA, which can obviously create some friction.
Another challenge is that automated sequences can be quite inefficient or prone to failure because the untrained RPA developer’s technique lacks sophistication. For example, there are ways in RPA packages to automate spreadsheet activities. But in reality, a much more efficient way is to use data tables that sit within the RPA package, although this is not instinctive to anyone who hasn’t had RPA training.
There are also some basic coding techniques that can add reliability and speed to the way the RPA works and that’s something that needs to be learned. These are code snippets, which are not difficult to learn. It’s like going on holiday and learning enough to order a beer; you don’t know the language in full, but you know enough to get what you want.
The last challenge is that pilot projects are failing to scale up to have a meaningful impact on the enterprise. There are many reasons for this. In addition to the points above regarding the reliability of automated processes, an example of why automation fails to scale is that pilot projects are selected based on the knowledge and enthusiasm of the RPA developer rather than selecting projects which can be cascaded to other developers and departments.”
How can these challenges be overcome?
“A good RPA training course will cover all of the points we mentioned. These are tailored for people without an IT background, allowing new RPA developers to overcome many of those challenges. For example, they would learn how to use data tables instead of using spreadsheets, they would learn some essential coding skills to help them do things that, even with a spreadsheet, they’d never think possible. It improves the functionality that is open to them even with just a little bit of coding.
I recently conducted a four-day training course, on which there was a lady who had worked at the company for years and during the opening discussion, admitted to being quite daunted by RPA. She didn’t say much on the first morning, but by the afternoon she was understanding and playing with code, learning how to adapt it for her own automation project. It’s by this process of guided playing that you see people develop their skills quite fast. The course is very applied; with templates and model sequences for each of the exercises.
As well as coding skills, it’s important to understand security risks and techniques, and this really helps build the bridge to IT. I always emphasize that RPA developers must include the IT community in what they’re doing.
One of the nice things about IT development is that it’s an incredibly sharing community. There are websites, for example, UiPath’s user forum and Stack Overflow, where people have shared vast amounts of knowledge. Thousands of developers have put forward solutions to problems, for anyone to access and use.
The other set of people who benefit from the training are Automation Project Managers who are accountable for turning the pilot projects into full programs. There are a lot of soft skills that are important to ensure the program scales up.
Firstly, understanding the selection of pilot projects; how to set yourself up for success so that once you have the pilots working, you can strategically think through them so that they cascade into different areas.
Secondly, RPA pilot developers need to have a positive mindset. If RPA developers are negative, then the pilot programs are not going to cascade as well as with someone who is a real evangelist for the work. Within the technical community these soft skills can be really under-appreciated. Scaling Up Excellence by Robert I. Sutton and Huggy Rao of Stanford University follows the experiences of many companies which have successfully scaled initiatives and there is a range of best practice and scaling examples that are worked into the course based on this book.
Setting realistic timelines is also hugely important. We’ve all experienced a senior manager wanting to come in with aggressive time targets. In the course we talk about managing a battle from the ground rather than in the air because when you only manage it from the air, you miss the detail that can be fundamental to success. If the set timescales are unrealistic you need to push back, or you’re setting yourself up for failure.”
What advice would you give to a procurement leader who is about to recruit new team members?
“The first thing is to realise the opportunity and risks which are afforded by process automation and machine learning. You need to create an awareness for yourself of what’s out there and what’s possible. Again, training courses can really help with that.
Machine learning based text categorisation enables a much more sophisticated approach to category management, particularly indirect spend data. In the past, if a company has an ERP the procurement team are often keen that a cataloguing system is used. However, the requesters get frustrated because they can’t find what they need in the catalogue; they don’t know their way around it or there are too many items.
Machine learning will have a major impact on indirect spend management. The need for people to manage spend taxonomies and catalogues will disappear, so that makes a really big difference to the way your team might be set up. If you weren’t aware that this capability exists, you might put a job out into the market for a data analyst, which would be irrelevant if you were to introduce this technology.
Before posting any job, I would recommend reviewing the role with an automation mindset to ensure it is future proof. Select candidates who have the skills and mindset to introduce and use new technologies as they become available.”
Richard Bell is founder of Productive Procurement, dedicated to using the latest technology to enable procurement professionals to focus on uniquely human activities they excel in.