Economics of Automation: Man vs. Machine 

Determining the value of automation is essentially one big exercise in calculating comparative advantage. To illustrate this, take two examples: Manufacturing via a completely manual process versus manufacturing via a completely automated process. These, of course, are extremes, and while manufacturing continues to integrate more and more automation, there will always be manual processes and people involved in the manufacturing process in some way. As Elon Musk put it, “Humans are underrated.” 

However, when determining the value of automation, one should be careful to ensure that the humans in the calculation are not undervalued. The completely manual process still does offer a good way to baseline any automation. Take a simple automation task as an example: Boiling water for tea. In both cases, there will be fixed costs such as the heating apparatus (stove) and the vessel to hold the water (kettle), and variable costs such as the water itself and the heating fuel. In the completely manual process, the filled kettle is placed on the stove and the operator is told to boil the water, at which point the operator turns the stove on. Once the operator sees that the kettle is boiling, they turn the heating source off, and notify the next process step that the water has been boiled. 

In the completely automated process, the water filled kettle is placed on the stove, a command is sent to the heating unit to boil the water, at which point the stove is turned on. A temperature probe in the water monitors the temperature increase, and once the water reaches 100°C, the stove is turned off by the control unit, and a notification is sent to the next step that the water has been boiled. 

Again, these two examples are the extremes; there are many examples where a level of automation is used in between these two endpoints. For instance, a thermometer may be used, or there might be a temperature switch, or even something as simple as a whistle on a kettle could all be seen to be automation or instrumentation additions to the completely manual process. 

Human vs. Machine: The Cost of Labor 

To determine the cost of the completely manual process, we would look at the variable costs involved, which in this case would be the labor, the energy, and the raw material, water. The labor would be the cost incurred in having a person physically present to monitor the heating of the water. The energy would be the fuel required to heat the water from room temperature to boiling. Finally, the water itself, which in this case will be assumed to be regular tap water. While the water and the fuel costs should be easy to calculate, the labor can be less so. Yes, there is the standard $/hour, but it would also be good to have someone that could act as a backup in case that person is unavailable. Also, since this application involves heat and heated liquids, there is a good number of hazards involved, which will require personal protection equipment (PPE), training, and insurance. 

Now that the cost of the completely manual process can be calculated, what about the automated process? How can this initial calculation be used to determine whether the automated system is worth the investment? Starting with the simple and moving to the complex, the savings in both the water and fuel would be seen in terms of efficiencies. If the automated system can determine when the water has boiled more accurately than the manual process, then there will be savings in two areas. The first is quality: If the operator stops the process before the water has reached the target temperature, then the water will be cooler than requested, resulting in rework (i.e., putting back on heat until it is boiled). The second would be the unneeded use of utilities in the case that the operator stops the process after the set point has been exceeded. In this case, there will be fuel that was used when it wasn't needed, and raw material, water, lost to evaporation. 

Last, but not least, the question of labor. In many cases regarding automation, this is usually seen as the greatest savings. This is most often the case involving tasks that are dirty, dull, or dangerous. Why these three? Dirty applications typically require paying a premium, regardless of the skill involved in the actual task, which provides more potential capital for automation. Dull applications are typically repetitive, which would allow for initial investments in automation to be amortized over longer periods. Finally, dangerous applications allow for people to be removed from these applications, reducing potential harm. 

The easiest labor savings that can be associated with this would be the cost of having the person monitoring the heating process for the duration. This would either be a direct cost, having to hire someone to specifically perform this duty, or an opportunity cost, having to pull someone from a different job to attend to the process. The less identifiable costs would be those associated with the labor. To start, the cost of having someone train the person monitoring the process and the cost of lower-than-expected efficiencies during the startup phase. Past the startup phase, there would need to be a second person trained that could step in as required to cover, the potential costs of an injury, ongoing PPE maintenance costs, and other ancillary costs to having an employee such as human resources, payroll, and management. 

Finally, and this is where even lower levels of automation can have a significant impact, there is the opportunity cost associated with the application. Opportunity costs are indirect costs and can be tricky to calculate, but essentially ask the question: What positive impact would this employee be having elsewhere if they weren't watching a pot of water boil? Having to keep the employee there means that they are not doing something else that could be adding value to the company, so when calculating costs, not only do the direct costs of having the employee there need to be calculated, but so do the indirect costs of them not being able to perform another task. This can be a significant reason to automate, even at lower levels of automation. If the person can leave the process unattended for a certain amount of time without impacting the quality or efficiency of the process, what additional value could they add?

*Remember, opportunity cost is the value the person could add to the organization were they not standing around watching water boil (i.e., maintaining machinery, calibrating sensors, writing reports, etc.). This will vary depending on what task the person will do with their new spare time.

With the costs of the manual process identified, how would the completely automated process compare? Would there be an advantage? Starting again with the savings to the water and fuel, the exact savings would be dependent on the skill of the operator, but seeing as even the least expensive temperature sensors have an extremely high level of accuracy, most are able to provide a measurement that is within a fraction of a degree to the actual temperature; there would be some savings here with less fuel used, less water evaporated, and less rework required. This could be significant in larger industrial applications, though in a smaller application, such as a pot on a stove, probably negligible. 

In the smaller application, the more significant savings would be regarding the labor component. The completely automated system would eliminate the cost of the operator and any ancillary costs, which could result in significant savings. In this case, however, equipment would need to be purchased, and a system engineered, programmed, and installed, all of which could lead to significant costs. This cost would then need to be amortized over the life of the equipment, which would then provide a $/hour cost figure. This would provide the fixed cost of the equipment, but there are also variable costs to consider. 

Besides the upfront initial investment costs, automation requires two additional major cost components: Maintenance and calibration. While there are recommended intervals to follow in many cases, though some have regulated intervals, the actual intervals that are followed can be determined by a simple cost/benefit equation. 

The cost of performing scheduled maintenance is offset by the benefit of a reduction in the potential of unscheduled maintenance. Scheduled maintenance can be scheduled at a convenient time, typically involves a known maintenance routine, and overall has a lower cost associated with it. Unscheduled maintenance schedules itself, will require triage to determine the maintenance routine that will need to be performed, and has a higher, in some cases much higher, cost associated with it. It can be likened to gambling, but fortunately in this case, using an unfair advantage like artificial intelligence (AI) to bet on which interval to use is not considered cheating. 

Calibration is a little different than maintenance. While it should be, and is often regulated to be, done at regular intervals, it will rarely stop a process. What will happen over time is increasing uncertainty will be introduced into the process. This could result in increased material usage. It could also result in decreased material usage, though in many cases there is some regulated measurement point that would ensure that a certain minimum amount of material is being provided in required applications. For instance, in packaging cereal, if the box indicates a required weight, the manufacturer will be required to meet that weight as a minimum. Any measurement under that weight will be considered a defect, and will not be allowed to be sold, while anything over that will be a viable product, though one that may have a lower margin on it due to the higher product cost. 

With that in mind, increasing uncertainty could have two separate costs: Inefficiency and increased scrap costs. Increasing the time between calibrations has the potential to increase either one or both costs. Notice the word “potential.” It will not be a given that the process will become worse over time, but there is a good likelihood of it happening. So, to determine the cost benefit of performing a calibration, on one side would be the direct and indirect costs of calibration, the service, and the loss of production, respectively, and on the other side would be the potential costs of not calibrating the system. Again, in some cases, there will be a maximum calibration interval that is forced on manufacturers, but in others, it will be a recommended interval, typically from the manufacturer. 

Who Wins? 

Overall, does the cost of having someone watch water boil cost more or less than having an automated system? If the automated system costs more, what about augmenting the manual process? This is sort of a grey zone between the black and white of a fully automated and a fully manual processes, but essentially it could be seen as a manual process with some automation used to assist in the process. 

In this simple application, a bit of automation can be added: A kettle whistle. This is probably one of the simplest bits of automation that can be added, but in this case, a significant one. The kettle whistle has one function, and that is to provide a whistle when the water boils. It will not be as accurate as a temperature sensor, but it could be more accurate than the eyesight of the operator. It will not automatically turn the heating element off, but it will allow the operator to leave the process unattended (provided they’re within earshot). This would have a significant impact regarding opportunity cost, since the operator could leave the system to run while working on another task, adding value in another area while still being able to ensure the quality and efficiency of the boiling water application. 

Using the completely manual process as a benchmark, this simple bit of automation would provide a decrease in opportunity costs, since the operator could attend to other duties, provided they were in hearing distance of the process. It would also reduce the amount of time that the operator would need to spend close to the heating element and boiling water, which would decrease the potential for an incident. It might not have a major impact on the amount of fuel used and water lost, but provided the savings are less than the cost of the kettle whistle, it would be easy to justify implementing it. In many cases, this type of compromise is an extremely viable solution. It allows for the automation of low-level, repetitive tasks, while retaining the flexibility that an operator can bring. Furthermore, much of this compromise can be done without engineering work, which eliminates much of the upfront costs involved in automation. 

When it comes down to determining the “winner,” it really comes down to the process and the application. Those that fall under the dirty, dull, and dangerous category are going to have a much higher payback than those that do not. However, with the continuing increases in capabilities and decreases in costs and complexity of automation, the picture is shifting to allow for more automation to be brought in in an economical manner at a lower threshold. 

One thing to keep in mind is that straight up comparisons between the two extremes can work but remember to consider the saying “time is money.” Much like the opportunity cost of the operator having to watch the kettle boil, locking up cash in capital equipment means it cannot be used elsewhere. So, the automation system, with its higher upfront costs, is going to commit more cash at the beginning than the less automated system with lower upfront costs. This cash that is committed could theoretically be used for investing, hence the opportunity cost. Since investing typically pays out over a longer period, larger investments with longer paybacks need to be costed differently than those with shorter paybacks. 

How is this done? Typically, by discounting future payouts, so the operator salary in year 2 will have a lower present value (to compare it against the upfront cost of the capital equipment) than the year 1 salary. There are entire textbooks dedicated to how to properly do this calculation, but as a ballpark (and this will depend on the market), a 7% discount isn't a bad one. This would mean the operator salary would be 100% in year 1, 93% (100% - 7%) in year 2, 86.5% (93% x 93%) in year 3, and so on. Typically, if the payback is less than a year, this can be ignored, but for larger and more expensive projects, this discounting can be extremely important in properly comparing the two options.