Why software robots have a 'lifecycle' too

Software 'bots' form they key enabling functions in Robotic Process Automation (RPA) systems - although machine-based in form and structure, their existence is still subject to a human-like lifecycle that affords them a surprising level of anthropomorphism.

Software has a lifecycle. Not necessarily because it is first created and then ultimately dies, although legacy systems are sometimes retired and killed off in that vein. Software has a lifecycle because we generally talk about the various stages pre-, during and post-deployment that characterise different core intervals in an application or data service's usage.

Depending on which definition of the Software Development LifeCycle (SDLC) you read, the core phases that any piece of software goes through are requirements analysis &planning, design &development, testing, troubleshooting &debugging and then finally onwards to deployment.

This core notion of IT existence has been around for a long time, since the 1960s according to a paper by Elliott &Strachan &Radford in 2004. We could update these cornerstones for the current age of cloud and add in incremental extensions, AI augmentations and interconnection to third party services via Application Programming Interfaces (APIs) if we wanted to.


Into the age of bots

We need to further extend our perception of software lifecycle ‘biology' if we think about the use of so-called bots i.e. the software robots that populate Robotic Process Automation (RPA) systems tasked with shouldering repetitive human tasks inside what are often monolithic IT systems.

Despite their digital DNA and the fact that they are born as robots, software bots have a particular life pattern and not all are born digitally perfect enough to enjoy immortality. As we interface our human working lives with software bots, we impact their life and wellbeing too. CTO of low-code automation company Appian Mike Beckley is vocal on this subject because his company aims to help control bots through its Appian Robotic Workforce Manager product.

"If you design with the concept that bots are perfect, then you are setting yourself up for a fall," said Beckley, speaking at the Appian World virtual user conference earlier this year.

Beckley explains that to build a software bot and understand its place in an enterprise organisation's workflow, the workflow itself must be modelled as a human workflow i.e. to be carried out by a real flesh and blood person. Then, once we start to build the bot, we can dovetail it with the real-world user and the management software layer itself can start to identify the tasks that should be delegated to the software robot.

So in fact, the bot starts out life as a comparative newbie, but it is allowed to assume more mission-critical jobs over time. It appears that bots graduate, grow up, learn more and perform more sophisticated human workflow responsibilities as they take on more tasks for us… and so we can say that they have a lifecycle.


Excessive human-handoffs

Appian CTO Beckley reminds us of the imperfection factor here and the need for bot management. "We might see a bot deployed and work quite effectively. Equally, some bots are deployed and we find that they perform an excessive amount of human-handoffs (where a real person has to step in to direct the work action), so these ones get retired, removed or sometimes just repurposed," he added.

Those dead bots, the handoff abusers, have still been useful because they have served to reinforce the fact that a particular part of a workflow should be left in human hands. But that's okay, because it allows us to engineer onwards and forwards and not try and apply automation where it won't fit.

EVP for products &engineering at RPA specialist UiPath Ted Kummert notes the gestation period for ‘bot birth' is comparatively short and these units of software can be generated quite quickly. Equally then, because (like the real world) an IT environment is constantly changing and effecting those who work with it, bots can ultimately be retired and killed off or at least remodified for new roles.


Anthropomorphised automation

CTO at RPA software company Automation Anywhere Prince Kohli agrees with the human lifecycle parallel that bots exhibit. He says that just like people, bots learn and are able to address more complex environments and processes over time. "The similarities don't end there. As requirements change, bots must evolve as well, bridging the old and the new and maintaining business continuity. In bot lifecycle orchestration, bots are monitored for health, their heartbeats and output measured, given support when needed and new teachings as well," said Kohli.

But notes Kohli, there are times when bots can not evolve and have to be retired gracefully as they hand over to new systems that can do the job more competently. The humanisation factor goes one step further - Kohli says that many Automation Anywhere customers anthropomorphise their bots, naming them and even throwing initiation and retirement parties.

As we get more used to working alongside software bots, we should lose some of the machine-feel factor that can make them seem cold or sometimes dumb and non-intuitive. There are real factors out there that will help decide whether these things live or die.

Some elements of bot automation lifecycle may occasionally ‘break' due to data source corruption, scalability issues, incomplete integration connections or other reasons - and these factors can also be thought of as potential bot killers. So just like people, bots need training, investment, skilling up, tuning and perhaps even the occasional tummy rub.

They really do have a lifecycle. All they don't need is lunch.