Date: 12 September 2022
While automation has much to offer, it also has its limitations. After multiple years working in this environment – from a test automation point of view and addressing general business processes – we’ve identified the five most common pitfalls or problems that beset many projects. Common to the five is that automation should be based on an organisational strategic vision and foundation rather than tackled as a tactical measure to solving immediate issues. Check out the list, so your automation project doesn’t get stuck or come short of expectations.
1. Automate everything!
No, don’t automate everything. Automation looks impressive on the drawing board; no question about it. It often seems so incredible that you and the other folks in the room have a significant ‘Eureka’ moment, seeing the fantastic potential automation has for everywhere in the business.
‘Incredible’ is precisely the right word because, quite often, it is, in fact, not credible. What looks great on a drawing board doesn’t always translate into something that works great on the factory floor (or in the business processes). Nevertheless, automation has its place, and the first place is addressing relatively simple repetitive actions or tasks. The first steps towards automation should be taken carefully, with truly ‘easy’ targets. Run these first, and you’ll quickly sense the potential and the equally fundamental limitations. Automation is a tool; it generally eliminates tasks but will make them faster or easier. End-to-end automation, too, is rare, and no amount of wishful thinking will change that.
2. Automation is set and forget!
Absolutely not. Automations are a bit like cars. Great when new, but without regular attention, they quickly deteriorate. Building automation is only the first step in a long relationship where maintenance, management, support, adjustment, and near-constant tinkering are required. You cannot build automation and expect it to take care of itself. Keeping a close eye on it means getting the value out of it – automation has a cost of ownership, and the trick is making that cost of ownership lower than the cost of doing things manually.
This is also part of why automation doesn’t replace people, despite the enduring economic fallacy of displacing workers (this fallacy goes back to Arkwright’s spinning jenny and the angry workers who broke the machines – the Luddites). Instead, automation can reduce unpleasant, dull, and menial workloads so people can do more exciting things. Like, for example, building and supporting automation.
3. Automate out of the box!
This is a perennial favourite among the marketers of automation (and a great many other) software solutions. Unfortunately, despite decades of these promises, we haven’t caught on quite yet that nearly all ‘out of the box’ functionality is anything but. Remember ‘Plug and play’ peripherals? There was a reason this was amended to ‘Plug and pray’, as religious intervention is often your only hope of it ever happening.
With automation, your unique processes, software, infrastructure, and so on all have a role to play in getting successful automation up and running. Business analysis, customisation, integration, ingenuity, hitting it with a hammer – some or all of these tasks may be necessary before anything works ‘out of the box’. ‘Low code-no code’ isn’t quite as easy as it sounds. But it is a credit to the marketing folks!
4. Automate with existing software!
This is more of a system integrator than vendor promise, and it is enticing because who doesn’t want to use their existing software for a more significant advantage? This is not to say that it can’t happen because we don’t know what’s in your current stack. It could be the case that you have every last API, RPA, and BPO available to accelerate your ERP. But in our experience, this is most often not the case. Automation is at stake because the build takes time and effort, usually involving multidisciplinary teams. Shoehorning existing software – forcing it to fit – is a bad idea. If you have the right tools, go for it…but don’t sell yourself short and poison the automation well when things don’t work out as planned. This is another pitfall: your automation efforts should be strategic, not tactical (that is, taking place within the broader context of organisational goals). When it’s strategic, you’re not going to cut corners. Are you?
5. Automation in Isolation!
As mentioned in the introduction, automating in isolation is a bad idea that often causes problems in the long run. Building quality automation depends on cultural collaboration in your delivery teams, so the results are highly valued, highly used, and enduring. Unfortunately, many test engineers focused on test automation do not communicate proactively with businesses and other stakeholders. The result? Processes are automated, sure. But they aren’t the ones the business needs help with as the engineers (who probably feel clever and under-appreciated) aren’t aligned with the critical business and project needs. Before engineering happens, discovery is essential, as is transforming test cases into user journeys.
Finally, it doesn’t matter where your automation is – whether on tests, business processes, data integration, or elsewhere. These five shortcomings are common to all of them. Automation has much to offer, but to plagiarise the Gartner Hype Cycle a little, unrealistic expectations will lead to troughs of disappointment. Approaching automation with a realistic and cautious mindset means assuring your path to value and a solid, recurring return.
A footnote. Automation is not the same thing as Artificial Intelligence (AI). Some automation can and does include AI, but most aren’t. Despite that, people do conflate and confuse the two. AI is a simulation of human intelligence, where computers ‘learn’ and adapt their behaviour to changing inputs or circumstances.
On the other hand, automation is an extension of the human controlling it (a tool) and is as good as the people standing behind it. So yes, you can build some AI into automation as and where necessary or possible. But that doesn’t make AI automation or automation AI.