Robotic process automation technology, an estimated $5 billion market by 2024, is generating a lot of excitement in the enterprise. Designed to automate rules-based, repetitive tasks performed by humans, RPA's software robots offer enterprises the potential to cut labor costs, increase accuracy and improve compliance with relatively modest investment compared with other major software deployments. But as IT executive Gerson Benker makes clear in this interview, deploying digital labor, as he calls it, takes work to get right and a thick skin.
Benker is vice president of global IT operations at Carestream Health Inc., a $3 billion provider of medical imaging systems and cloud-based IT solutions. His IT environment spans 100 countries, serving 7,000 users worldwide, including people in field service and in sales. The business, headquartered in Rochester, N.Y., was formerly Eastman Kodak Company's Health Group and now is owned by Onex Corp of Canada.
An outsourcing expert, Benker served as executive program lead in strategic outsourcing at IBM prior to joining Carestream in 2014. At Carestream, he recently has been immersed in a major transformation of the company's IT operations, turning to managed services provider IPsoft for help. He is using IPsoft's 1Desk to automate many IT management tasks.
Here, he talks about some of the technical and human challenges that an IT transformation of this scale presents to CIOs.
Editor's note: This interview has been condensed and edited for clarity.
How do you define robotic process automation?
Gerson Benker: I really define it as digital labor. That's what it is all about for me -- replacing human labor with digital labor, or as IPsoft calls it, with virtual engineers.
And why do we do it? You can increase your volume of support without increasing your cost and your staff -- actually, it can decrease the staff and it is more cost efficient.
IT has been automating processes from its beginnings. How is robotic software automation different?
Benker: I have been in the data center business for 15 years. And yes, when we talk about data center automation, there are tools -- I can push a button and deploy the virtual machine. That is true, but at the end of the day, there has to be somebody sitting at a console and telling the tool what to do. When we talk about [RPA], this is really digital labor in the truest sense, meaning, once you implement it, the virtual engineer will take care of it end to end.
And that is the difference: The end-to-end work is 100% completed versus, 'Here I am, a human, in front of a console and I have an electric hammer instead of a normal hammer. I have tools that help me to automate to do stuff faster.' That's the kind of automation everybody else has out there.
[Robotic process automation] depends on being able to describe the processes and being able to mimic what the human being was doing with machines. When you implement other tools, they have pretty rigid ways, processes and capabilities that they can execute on: They can have great automation built in to execute certain tasks, but that's it. They are rigid. They are what they are; there is no variation possible, or you break the tool. With IPsoft, especially when you get your own instance of IPcenter, you actually can totally customize it and adjust it to your environment.
And that's what we have done in the last six months. We just sat there and 100% programmed it toward the automation that we need here at Carestream and my environment to fully automate certain tasks, whether it is password resets, request management, escalation management, etc.
Teaching digital labor to work
How has the digital labor disrupted your IT organization?
Benker: The disruption is obviously that people's jobs change. Just within the last year, I have removed several people who were made redundant due to automation. We have tried to reskill people and redeploy them, but that is not always possible.
The other disruption is for IPsoft to adjust its virtual engineers and workflows and automation processes to our network. And there are some hiccups. Not every network is the same; it's not the same environment in every company. So, there's some hiccups in the beginning -- I don't want to call them quality issues; it is really a learning curve.
Gerson BenkerVP of global IT operations, Carestream Health Inc.
How big a learning curve?
Benker: It takes several months, and sometimes even longer, to get it all right 100% -- to get the virtual engineers to understand the workflows and who needs to be notified when. Sometimes notifications go into a black hole for a while, because they just don't grasp our processes correctly.
So, it is a little bit painful, but once the process is adjusted and once everybody has learned, the end result is a healthy environment. The error rate is so much lower -- it is almost zero -- once it is implemented correctly. But there are things the virtual engineers have to learn.
Let's take an example we had recently. Servers are being patched, which is an automated process executed by IPsoft through their virtual engineers, and a server doesn't reboot correctly. What do you do? Do you try it again with different configuration possibilities or do you execute a process where you escalate to a human? This is something you have to teach a virtual engineer: Do you want me to try to reboot this server 11 times, with different ways of bringing it up, or is that not the right way to do it, because it is a very old server and if something isn't right with it, we'd rather have a human look at it?
Here's another example we had recently. Servers are coming up correctly, but the virtual engineer -- the monitoring tools -- finds out that certain processes didn't start afterwards. The virtual engineer can try restarting those in a different service order; the virtual engineer can try to execute some other commands to bring that service up and running … or again, the virtual engineer can say, 'I am not going to do this because it is just too delicate and it may break other things; I will involve Tier 3 and have somebody look at it from a human perspective.'
Approval processes are another area. Someone on the service desk wants something -- a fully automated, software package: Who does the workflow go to [to] approve that package, or does the digital engineer say, 'OK, I am going to give it to this person based on these criteria and install it?'
These are the kinds of things that digital labor has to learn that takes weeks or months, right? And the better your processes are documented -- the better your processes already are -- the easier it will be for IPsoft to work with you to implement this correctly upfront.
If you don't have a lot of documentation, then it is a learning experience together.
I imagine companies that don't have good documentation of workflows have to use people whose jobs are going to be replaced to do that documentation. How do you deal with a situation like that?
Benker: Yeah, I have experienced that in the last six months to a year; these people basically work with [the vendor] to eliminate their jobs, and you deal with it several ways.
Hopefully you are in a position to say, 'Listen, guys, I am trying to get you out of the mundane day-to-day jobs. I have bigger and better things for you to do.' Or sometimes you can incent people, saying, 'Please do this for me, stick around, and I will give you a retention bonus of so much if you hang in there and do it.' And there are people who are perfectly happy to do that. They take the bonus and go on vacation for six months.
It really depends on your organizational makeup, what kind of people you have, how close are they to retirement, can you reskill them and so on. There's no real straight answer to it, but it is a difficult situation, I have to say, in the least.
As someone who's been in outsourcing for a long time, you're probably familiar with this kind of disruption under other guises.
Benker: That is why this is a little easier for me, because I have done this for 20 years. I don't get very emotional about this, but I realize on the other hand it is a transformation. Nowadays when we talk about IT transformation -- and this is what a lot of CIOs and others don't really recognize -- with the technology transformation, there needs to come an organizational transformation. One and the other must be connected, and if they are disconnected, usually it is a train wreck at the end.