For some months I worked in Robotic Process Automation business and from growing customer interest I believe the business will continue to grow in northern Europe. Big companies with thousands of people (like in banking) have done it for years but I think it is the time for medium-sized companies.
Here are some things I learned during the months.
1. A lot can be done with just rule-based process automation
I got questions like “is it just rule-based RPA or real AI”? There has been so much hype about AI and it seems there’s an expectation of everything should be “intelligent”. The truth is a lot can be done with just rule-based RPA. I would recommend starting with rule-based RPA and adding AI (actually machine learning) for specific decisions in the process when it makes business sense.
2. For small and medium-sized companies the main benefit is well-being of employees
It requires so much work to describe what the software robot should do, develop the robot task and continuously maintain the robot that good business cases are rare when you’re automating tasks for just one or two employees. That said, I believe there can be big benefits in creating employee well-being by automating repetitive stuff.
Trending AI Articles:
1. Deep Learning Book Notes, Chapter 1
2. Deep Learning Book Notes, Chapter 2
3. Machines Demonstrate Self-Awareness
4. Visual Music & Machine Learning Workshop for Kids
3. The deeper you get into RPA the more rigor of professional software development practices is needed
When there are several software robots the need arises for centralized logging, error reporting, visibility for management, tools for debugging, configuration management, version history, etc. Unless you go with tools and tasks which can be automated without much or any programming logic things will start to look like a software development process.
And it may be challenging to keep software professionals (with in demand skills) interested in creation of software robots.
4. Software robots are fragile and require a lot of maintenance
A software robot can stop functioning because of anything: an unexpected application dialog appears, something in input data has changed, network lag, software or operating system got updated or an application window is in a different location. Expect a situation to arise with each automated task at minimum once a month but possibly every week and plan who is going to deal with it, how, in what time frame and how much it will cost.
Also, start with robots as simple as possible and add complexity by adding new simple and separated (in IT the term is loosely coupled) software robots working in unison to finish the bigger task.
5. Customers don’t want to pay a lot for RPA because they think implementation is simple
Advertising of easy process automating is working against the business. Customer’s don’t want to pay for the costs when they have been led to believe a task can be automated in minutes or hours.
6. RPA is not stealing jobs
I did not see any signs of RPA taking jobs in the near future. RPA is definitely not a silver bullet that will fix all efficiency and cost issues.
Predicting the future is difficult but I will try just for the fun of it.
I think several platforms for RPA will pop up in the near future. Later the said platforms will pivot the focus from managing software robots to managing software agents which use APIs to perform the same or similar tasks, collect data and integrate business processes. Before that happens, computer vision will be applied to speed-up development of software robots, and machine learning will help the support people to manage problem situations, e.g. by bringing errors immediately to attention in an understandable way and suggesting fixes very much like Google Mail offers suggests reply template to answer emails.