RPA - A visible struggle

We've seen a thing or two in RPA implementations, good, bad and also what's ugly but that's not what this article is about. This article is not to downplay RPA but rather how RPA leaves a void in a process flow. RPA solves the work issue, provides process insight through logs, even provides BPMN documentation and so on. We feel however it doesn't solve the data overview issue.

RPA?

If you have been around these past few years, you probably noticed that UIPath, Automation Anywhere, Blueprism, NICE mad a big entry into IT. It's to say they empowered the business side to automate processes that IT didn't get time to do or in some cases didn't want or dared to.

But what is RPA? It's a Swiss army knive of tools and tricks to simplify and speedup automation or as wikipedia tells it:"Robotic process automation is a form of business process automation technology based on metaphorical software robots or on artificial intelligence /digital workers. It is sometimes referred to as software robotics".

And it's great, even marvelous but ... how can we see what's going on without needing accesses and tools etc?

The problems

RPA is a fantastic tool it also has things wrong with it, because well it's created and implemented by us, IT/business users. That's making things a lot more complicated all the sudden.

1. Complex processes

We're humans, we like to make things complex. We're trying to automate a whole process in one go or include a lot of exceptions. In some cases we're trying to put so much intelligence in a automation workflow that we don't need a human anymore, unless something goes wrong in the workflow.

The more complex a process is made, the more issues will appear when using the workflow in production. Adding on to the fact that the workflow contains custom developments, code & steps that are making the process harder to debug. Thus the need for an RPA developer that has in dept knowledge. Why? Because a process is complex and a regular business user doesn't have the confidence or expertise to debug.

Should a workflow be so complex that it handles "everything" without simple verification?

2. Process & data control

In a good amount of cases workflows are designed to be triggered by a file based trigger. Something is dumped somewhere on a disk, on a server or SMB and the workflow will start processing it. The workflow uses the file content to execute a workflow and a given output is generated somewhere. From a business side of things we're relying that this just works, invisibly, always, with a minimum of monitoring from business side.

Is it right that business cannot monitor in the processes and content, or being able to re-do/execute some basic elements?

3. No business control

Business isn't the keeper of workflows and processes. They are however impacted 100% if something doesn't go to plan. Many processes are time sensitive or critical to the daily operations of a company, if something goes wrong then there''s no real way of knowing this the moment it happens.

In some cases issues are discovered days after the issue occurred, because other processes failed, because processed didn't run due to bots being blocked/occupied. Starting the work on finding what has gone wrong is cumbersome and complex. it should not be

Is RPA a background tool only? And if so, should business not know or be able to check if everything went well? Are we losing transparency when automating? Should we not be more concerned about & involved with data flowing through a process?

The fixes

1. Diagrams & workflows

Documentation inside and also outside the RPA tools is very important. We often notice that knowledge often only resides with individuals rather then on a confluence page or wiki.

This means training people in documenting, the shape or form of the documentation can be debated upon, however it should be clear, precise and easy to interpret by both the technical and business.

2. Control by design

Processes rely on and generate data, logs, and information. Knowing inputs, outputs, intermediate results, reference data and managing all that is important and having all these things desperate all over isn't going to help.

Having one source of truth and one single tool or database where verification can be made is key to a solid process and monitoring operation.

3. FastAPPS

Automatize offers FastAPPS, data modeling services and consultancy to facilitate faster adoption of your RPA projects and more efficient daily operations.