We’re drowning in a deluge of PR telling us that everything will soon be automated and run by AI. But as Justin Warren explains, automation isn’t always the right answer.

It is immensely gratifying to this author that the IT industry has discovered the Toyota Production System, and that automation is a good thing. Humans are being freed from the drudgery of soul-destroying menial tasks.

But if everything is automated, humans won’t have jobs any more. That’s either excellent or terrifying, depending on which start-up you’re talking to.

Should we automate everything? Just some things? Which ones, and how? Answering these questions for the firm has substantial strategic implications, and that means boards need to have a view on how automation will be used.

When Everyone Is Special, No One Is

Automation works best on repetitive tasks, where each task is essentially identical to the previous one. We therefore need to identify tasks that occur repeatedly, and which create enough value in aggregate to make the cost of the automation process worthwhile, whether that be buying a robot, or writing some software.

And here we come to a conundrum of automation. If we’re buying a robot  from a third party, then the market for these robots must be big enough for the external firm to sell them. If we’ve established that the robot can automate a standard process, and if the vendor is selling these robots to other customers, then all these customers must have the same set of tasks to automate. The same applies to automation software.

If multiple competitors are all automating the same set of tasks, then these tasks are not a source of competitive advantage. Any firm can perform them if they buy the same robot we did. Similarly, any firm could perform the same automation tasks if they buy the same software as us, or outsource to the same firm.

The key in using the automation technologies, therefore, is to support the way in which your organisation creates its unique value for customers, rather than slavishly copying everyone else.

Automation Provides Leverage

Automation magnifies the effects — both good and bad — of the tasks being automated. Like financial leverage, it magnifies the good and lucky choices, but also the bad and unlucky ones.

Knowing when and where to deploy automation is therefore a strategic management decision, and should be treated as such.

What To Automate?

The best candidates for automation fall broadly into two categories:

Firstly, tasks that do not add substantial value to the customer experience, but do add cost. What is the competitive advantage of having email? Of outsourcing email services to Microsoft’s Office365 service? These functions are hygiene factors, because not having essentially the same system as everyone else places a firm at a competitive disadvantage.

A note of caution, however. Over the past decade or so, many firms outsourced major sections of their IT functions to third-parties to focus on their so-called core business. These firms now find themselves in a world where IT provides enormous value to customers, but they no longer have any in-house capability. Will you be giving up future competitive advantage to save a few dollars today?

Secondly, tasks that add value to the customer and that are not reliant on humans to generate that value are good candidates for automation. Replacing the human waiters in a high-end restaurant with robots would fundamentally change the experience, possibly destroying it altogether. However, mass-produced frozen pizza doesn’t need humans to be laying the same amount of cheese on each base by hand. Automation will work well here.

The challenge is accurately determining in advance what customers actually value, compared to what you imagine they value. Would customers value “artisinal hand-crafted frozen pizza” enough to pay for the cost of the extra human labour? This is the core of what running a business is about, so hopefully you have at least some ideas here, but it pays to check your intuition with research and data before you destroy what your customers love about your products with ill-conceived automation.

What Not To Automate

Areas that should not be automated tend to fall into the following categories.

Firstly, tasks that should not be performed at all. A common mistake is to apply technology to reduce the cost of business processes that should not exist. Before spending a lot of time and effort automating a process, first ensure that it is necessary to perform it in the first place.

Secondly, if the setup cost is greater than the long-run aggregate value of the tasks being automated, then it is a poor candidate. This sounds obvious, but determining the value created by a given task is often challenging, and the time horizon for “long-run” is arbitrary. In the long-run, we are all dead. Similarly, the cost of a project is often underestimated, particularly for technology related projects, so what looks economical in the initial business case may not pan out if the project overruns its budget as most typical technology projects do. Are you confident you’re better than average at these projects?

Thirdly, automation is a poor choice if the cost of changing the automation system is too high. The ability to change and adapt to future circumstances becomes more important the closer the automation system comes to the customer. Imagine a fashion house spending millions of dollars on a machine to build last year’s pants that cannot be changed to make this year’s pants without spending more than the entire season’s profits.

Finally, any processes that are important parts of the value your organisation provides to its customers, but you’re not sure what effect automating the process will have. Automating here risks damaging the customer experience while failing to derive any of the benefits of automation.

Choose Wisely

In most organisations, plenty of business activities are obvious candidates for automation. The more arguable cases can be safely left until later. Spending limited management and board oversight time on marginal prospects is a poor use of resources when plenty of juicer fruit hang much lower on the tree.

Have the discipline to identify the lowest risk and highest payoff candidates, and address them first, before moving on to more challenging, marginal cases.

It is important to clearly articulate why the chosen cases are the most attractive. Staff who are keen to automate are extremely valuable champions. If they become discouraged because they feel their advice on what to automate is falling on deaf ears, they are likely to decamp for greener pastures, placing your chosen automation projects at risk.

Organisational change is difficult enough without alienating your biggest internal supporters.

It’s also important to understand the reasoning, because business conditions may change. Actions by competitors might mean that excellent choices from six or twelve months ago may no longer be. Or innovation by vendors might make a once unattractive option very attractive indeed.

It is perfectly acceptable to change your decisions based on new information when the reasoning is clear.

Next Steps

  • Perform an audit of the use of automation in your organisation. Where is it being used, and where is it absent?
  • Look at the overall business strategy and identify which tasks add value or competitive advantage, and why. Are some tasks hygiene factors that should be automated or outsourced? Are you sure they will stay that way?
  • For tasks that add competitive advantage, are they sufficiently static and numerous to be worth automating? What is the cost of changing your automation system if the underlying tasks need to change, such as to respond to changing customer desires or actions by competitors? Will automating these tasks constrain your organisation unduly?
  • Are there redundant business processes you can simply eliminate instead of automating?

Further Reading

  • On the concept of financial leverage.
  • Hertzberg 2-Factor Theory, which is where hygiene factor comes from. That’s from the book:
    Herzberg, Frederick, Bernard Mausner, and Barbara B. Snyderman. The Motivation to Work. 2. ed. New York: Wiley, 1959.
  • The concept of “Core Business” stems from people misunderstanding this HBR article: Prahalad, C. K., and Gary Hamel. ‘The Core Competence of the Corporation’. Harvard Business Review 68, no. 3 (May 1990): 79–91.