This episode was written by Flt Lt James Kuht & edited by Lt Col Henry Willi. The thoughts are the authors own and do not represent the Ministry of Defence. The episode can be found above, or on Spotify & iTunes. You can find out more by following us on twitter, @ReDefPod.
In this episode we explore automation – comparing an amazon order to going to military stores, exploring Elon Musk’s attempt to make his Tesla factory a fully automated ”unstoppable alien dreadnought”, and finally how and why we might use automation in Defence to punch above our weight, reduce bureaucracy, and “take the robot out of humans”.
What is it?
Going to a military clothing stores can be an ordeal. Limited opening hours, manual logging of what clothes you take, one-size fits no-one clothing, and often leaving without that crucial strap which holds your webbing together.
Let’s compare it to Amazon, which is essentially a high-performing stores department. Except unlike stores, you can order, anytime, anywhere; one click, an automatic confirmation email arrives and a parcel follows the next day with barely any human intervention. Don’t like the product? Fill in the returns proforma, a returns label is automatically created and you’re automatically refunded within hours. During all of this, your details are automatically collected, and the store automatically notes your preferences and targets you with adverts of products you may like, the next time you visit.
The process is automated to a tee, and has been a critical component Amazon’s rise to becoming a stores department worth $1trillion. An almost fully automated stores department which we’ve never entered or even spoken to a member of staff from.
So simply put, automation is getting your computer to do the repetitive tasks in a process that humans would otherwise do, only more quickly, more accurately, and tirelessly. One professor is famed for describing it as “taking the robot out of humans” (Professor Leslie Wilcox ref: AIMM, 2020).
Now automation has existed for decades, but it’s use is becoming more mainstream due to technologies such as Robotic Process Automation, or RPA for short. RPA differs from traditional automation for two main reasons. Firstly, you don’t have to write code to program RPA to do what you want – you just show the RPA software, otherwise known as a “bot”, what to do, and it does it (AIMM, 2020). This makes RPA highly accessible to workers, and infinitely customizable to the organisation. Secondly, because an RPA bots works just like a human, and can switch between different applications by just the click of a button like us, it solves significant integration problems. Many large organisations suffer from data siloing in different applications, RPA is not fussy – just show it what data you want copying across from one application to another and it simply will mimic your actions to achieve that, requiring no integration.
So let’s delve into a few examples of automation.
A large bank received 1.5 million claim forms per year, each needing to be saved, processed, added to a database, approved or denied, and replies sent. Dull work fit for a robot… so they deployed 85 bots to run the process. They estimate that the human equivalent to add this capacity would have taken 200 full time employees, at over three times the price (Boulton, 2018).
Apart from automating a simple, pre-existing process like claims, HR process or invoices, RPA can add new value in other ways not previously practical or possible. The UK retailer group “Shop direct” created a bot to identify customers who are affected by floods, and automatically remove late payment charges from their accounts. (Schatsky et al. 2016)
Beyond these two simple process automation tasks, automation can be combined with Machine Learning, the topic of our next episode, to create “Cognitive automation”, Automation that is able to be smart, and make decisions in a situation-dependent fashion not entirely hard-wired in rules.
One example of this was achieved by Virgin Trains – they created a cognitive RPA for dealing with emails from customers requesting refunds for delayed trains. The bot could use natural language processing, a machine learning technique, to read and understand the sentiments of the customers email, categorise them, and issue the refund to the customer, or escalate their complaint to customer service. They estimated that the solution reduced the daily processing time and manual labour involved with these emails by 85% (Schatsky et al. 2016).
Who’s been nailed by it – what went wrong?
Now let’s discuss when automation has gone wrong, and what we can learn from it. Now, of course, not many companies want to share their stories of failure – but one story is in the public domain – the legendary entrepreneur Elon Musk’s Tesla Factory. Elon is said to have recounted a dream to his coworkers – in which he said he’d seen a factory of the future – a fully automated manufacturing plant. He described it as an “Unstoppable alien dreadnought” (Duhigg, 2018).
Despite Musk’s best efforts trying to make the dream a reality, a year later he tweeted “Excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated” (Musk, 2018).
What did he mean by this? Well though automation is really effective at rules-based processes, for example: “if X do Y”, it’s not as good as humans yet at complex decisions, where perhaps X & Y are less clear, or complicated by situational variables. In the factory example, imagine the process of screwing on a car bumper. If a screw’s thread is slightly misaligned, a human would quickly correct for that. The automated solution may not even notice, and would likely force it, perhaps destroying the thread.
5 years from now
If we grab it by the horns
So, we’ve had a whistlestop tour through automation, and some examples of its use. Now let’s look to the future – where could the greatest gains for the British military be over the next 5 years?
In the near term, the low-hanging fruit for automation is freeing our personnel’s brainpower from the bureaucracy and form filling, so we can employ them to their maximal capacity.
Perhaps ordering kit from stores will be automated not dissimilarly to amazon – you’ll be measured using a 3D body scanner from your phone’s camera, request what kit you need and, from your email landing in stores to receiving your kit delivered to your quarter to any returns you need to make, it will all be automated. This will enable large efficiency savings, and start to build up a clean and structured database to better inform our clothing procurement decisions, or even automate them.
Perhaps when you move to a different base, automation will mean that you don’t have spend a day formally exiting the base, then onboarding in your new job, getting access privileges, register at the medical centre, dental centre, stores, the MT section, the mess. Relocation leave is currently 5 days, of which plenty is menial form-filling which is simply a waste of time and leads to significant job dissatisfaction. How many times have we provided the military with our details, can it not just be automatically transposed to our new roles or locations?
If we zoom out further to the more ambitious projects, perhaps we will be using more cognitive automation solutions.
Imagine calling the MOD operator and having a bot answer the phone. She’s available at any time with no delay; understands your request by utilising natural language processing, validates your credentials, makes relevant suggestions, and then connects you to the right number. A form of this already exists on MoDNET, “Ask Sally”, though very few people truly use it as their first port of call, preferring to pick up the phone to a human. To be used, we’ll have to make sure these technologies are accessible and unarguably superior to encourage mainstream use.
And finally, imagine you’re a busy engineer of a drone squadron. Perhaps automation will mean you no longer have to spend hours a day writing reports, because any errors logged by your personnel are automatically transposed into a concise report using natural language processing (an AI technique we’ll explore in episode 5), and logged into a central database in an efficient and consistent manner. This structured and consistent dataset then allows automation of spare parts ordering, early flagging of repeated faults, and paves the way for predictive maintenance in the future, as the data is clean and properly organised.
If we miss the opportunity or screw it up
So it’s a no brainer right – we automate the boring stuff and us humans get all the exciting work – what could go wrong?
Well, it seems that the biggest thing that could go wrong is by rolling automation out in a haphazard fashion, with a lack of governance, development to common standards, and ownership. Everyone knows a story of an excel guru in their workplace who created a clever macro that automated everything until one day it broke, he’d moved on, and no-one knew how it worked or how to fix it.
The answer probably lies in a hybrid approach. As bottom-up schemes like the jHub Coding scheme equip our staff with the skills to automate many aspects of their work, the top-down Automation Centres of Expertise need to empower them with access to the tools to build great products, whilst ensuring that digital chaos doesn’t result.
Success would be contingent on three central functions – firstly, defining automation best practices and incorporating them into our upskilling efforts, secondly, building safe and secure infrastructure to test and deploy these products and thirdly, providing permissive-but-structured governance. The governance needs to be implemented not to discourage individuals to build products that improve the MoD, but to scale successful ones rapidly, merge duplicates, and kill off unsuccessful ones. Only through this mechanism can we change the culture of our workforce to one empowered to innovate, without causing digital chaos.
This approach – centralised development of best practices and governance, coupled with decentralised digital empowerment of the workforce – is one that will not just help us automate our processes, but implement all of the exponential technologies explained in this podcast series – some refer to it as the “base-layer approach”.
We hope this podcast has shed some light on automation, and how use of bots to automate processes may be seen as a threat to some jobs (Frey & Osborne, 2017) but may free us from huge amounts of bureaucracy.
At the moment we are at a stage where automation has proved effective for rules-based process automation, but we’re accelerating into an age of cognitive automation, where bots will be able to automate more and more complex tasks. One key bottleneck preventing us from reaching truly “smart” automation is clean, structured and accessible data to train the intelligent bots, as discussed in our data episode.
What does this mean for leaders? Soon you may be managing as many bots as bodies. Identifying opportunities for where automation can add value in your workplace will be of paramount importance, as will allocating the unique skills of your personnel as the repetitive jobs they were performing beforehand are taken by automation. Automation tools such as Microsoft flows are starting to become available to MoD employees, why not consider making a process map of a part of your work, and exploring whether it may be amenable to automation?
As UK Defence is expected to deliver more with the same personnel and resources, we’ll need to fight above our weight – filling in forms and spreadsheets is not an effective use of our time, so let’s use automation to “take the robot out of our humans”.
AIMM, 2020. What Is Robotic Process Automation?. [online] Available at: https://www.aiim.org/What-is-Robotic-Process-Automation [Accessed 3 June 2020].
Boulton, C., 2018. What Is RPA? A Revolution In Business Process Automation. [online] CIO. Available at: https://www.cio.com/article/3236451/what-is-rpa-robotic-process-automation-explained.html [Accessed 3 June 2020].
Duhigg, C., 2018. Dr. Elon & Mr. Musk: Life Inside Tesla’s Production Hell. [online] Wired. Available at: https://www.wired.com/story/elon-musk-tesla-life-inside-gigafactory/ [Accessed 3 June 2020].
Elon Musk (@elonmusk), 2018/04/13. Available at: https://twitter.com/elonmusk/status/984882630947753984?lang=en [Accessed Monday 6 January 2020].
Frey, C.B. and Osborne, M.A., 2017. The future of employment: How susceptible are jobs to computerisation? Technological forecasting and social change, 114, pp.254-280.
Schatsky, D., Muraskin, C. and Iyengar, K., 2016. Robotic Process Automation. [online] Deloitte Insights. Available at: https://www2.deloitte.com/us/en/insights/focus/signals-for-strategists/cognitive-enterprise-robotic-process-automation.html#endnote-sup-2 [Accessed 3 June 2020].