Automation Is the Last Thing a Startup Needs.

Scaling up production of a new company that involves untested parts is a challenging and risky task. Unlike in the digital world, when working with physical processes production requires more upfront investments. When a company has to commit to significant investments in the early stages with limited experience and knowledge, there is an increased risk of those investments going wrong. How a company can mitigate this risk was discussed in my earlier blog post that you can find here. In the previous text, I suggested what you should do. In this one, I suggest what you should not do: Automation.

Early automation of a production process has many downsides. When production is carried out by human operators instead of machines, they can use all their senses to learn and improve the process. While running the process, the operators do not only learn about the process and get ideas for improvements; they also do quality control. Learning from where the failures are originating from first hand gives the company better position comparing to fully automated production line where the raw materials go in one end and the ready and faulty products out from the other. First-hand knowledge from the production line is essential for process development. In the early stages when the understanding of the production process is low, long hours of human labor are needed to learn what could be automated in the first place. Automating too early in the learning curve will hinder the learning for the presented reasons. Additionally, machines and other types of automation hardware are costly, acquiring them involves long lead-times, machines are not readily adaptable, and they have a high risk of becoming obsolete when build-measure-learn loops are completed.

Instead of automation a cheap and fast way to improve the output of production and to reduce the workload of human operators are Karakuri- methods. Karakuri is Japanese and literal translation of the word to English is “mechanism.” Small makeshift automation solutions in a production line are examples of Karakuri. Let’s say an operator needs to move a heavy machine to her workstation frequently, but it must be returned after using because it would be blocking the operator from executing the next process step. The machine moves on wheels a small distance back and forth, but the high frequency of moving makes it exhausting for the operator. Instead of building larger workstation (expensive!) or automating the movement of the machine by a motor (very expensive!) wire can be attached to the machine on one end and a bucket full of sand to the other. Also, a rail is placed on the floor for the wheels to move on. Now, the operator can pull the machine to her and lock it in place while in use. After using the weight of the heavy sand, the bucket will pull the machine pack automatically (and cheaply!). A more straightforward example of Karakuri- solutions are shelves that are tilted to a small angle making the products on the shelf to fall to the edge closer to the operator. In general, shelf-made Karakuri solutions are cheaper, easy to maintain and easy to improve compared to purchased automation solutions.

One of the only places where automation can be recommended early on is data collection. Data collection and analyzing tools such as data loggers and analyzing software are relatively cheap, but they are also a lot more reliable than human operators. Data is in the core of the learning process, unlike improving the speed or quality of the production where automation could be considered as well.



Nothing is more important than AI

Most of us are familiar with the concept that it is better to fix the cause than the effect. This idea is at the core of the Lean Manufacturing concept, created by Taiichi Ohno and Toyota.  Lean Manufacturing was later developed into the Lean startup- movement inspired by Eric Ries that aims maximize the speed of learning.

Even though we’re aware of this principle, we very often end up doing the opposite. It is natural to just fix the effect, as it is faster and the results are easier to see. Yet in the long run, you lose time and money with this method, because, with the root cause unfixed, the same effect will occur.

The future of technology holds amazing sci-fi -like tools. Nanotechnology, bioengineering, genetics, robotics, space-travel and so on. But the only technology that can help us fight the root cause of poor productivity and resource wasting is the development of Artificial Intelligence, AI. All the listed technological advantages aim to increase the efficiency of human activities, but the difference is that AI is the only one that can help us understand which activities are worth doing in the first place, and how to do them in the most efficient way. For example, while robotics help us do things in high efficiency, maybe that activity was not needed at all in the first place. In other words, while AI can give us the GPS coordinates to our destination and map to follow, robotics can only make us move incredibly fast, and us humans decide on the direction in which we want to move. My example of robotics comes from physical production, but the same idea applies also to product design, marketing, HR and so on.

Our world is not lacking food, physical resources or talent. Yet we are wasting all of these in staggering amounts daily, especially human talent. We need improvement in the use of the world’s resources, and for this AI is the most important technology.

The significance of AI has been recognized widely, and all major technology companies are involved in developing ways to improve the technology and offer multiple ways to implement it into our lives. Even though the benefits of AI management tools are clear, strong incentives exist in the modern global market to fight against the “AI manager”.

AI management tools would steer companies toward long-term, sustainable growth, but the contemporary stock exchange and marketplace encourages to seek short-term profits. Since investors want to see short-term value increase in the companies they invest in, the management they name have incentives to sacrifice long-term plans over short-term ones. This is obviously nothing new, but every time AI takes a step forward, the decision to act against facts will have a higher opportunity cost. Eventually, this cost will be so high that it cannot be ignored, but the sooner we can bring AI into decision-making, the better. And since we are fighting issues like global warming and other ecological disasters, there really is no time to waste.

Fighting against better solutions sounds counter-intuitive, but the world is full of similar examples where existing structures prevent obvious improvements from taking over. Why do they still have separate taps for hot and cold water in the UK? Why do Phillips-head screws still exist when we have Torx screws? Why don’t all cars come with Run Flat tyres? The Run Flat tyres, the PAX system, is especially interesting. The PAX solution offers better safety and flexibility for the customer, the technology was shared with multiple tyre and car manufacturers, and market studies show that the demand is there, but still, we are not using it. The answer lies in co-adaptation and co-innovation risks. When there is some external entity in the supply chain that is not benefiting of the improvement, they will not be active in taking it into use, or, even worse, they might fight against the change. In this case, a co-adaptation risk occurred. Car service companies did not have high enough incentive to invest in PAX- machinery, and cars with these tyres could not find convenient service network. This theory and the PAX system example is presented in Ron Adner’s book The Wide Lens.

Technology is advancing faster day by day. AI management tools will be the most significant of all upcoming advances because they will make all other advancements faster. Improving the technology and creating applications for it is not enough, as there are obstacles to solve before AI is widely taken into use. Finding ways to smooth the path for AI to enter our world is the second most important task we face now, right after AI itself. And these two tasks will be the last ones us humans will have to do on our own.

There is nothing quite so useless as doing with great efficiency something that should not be done at all.” – Peter Drucker

Further reading:

Lean Startup:

The Wide Lens:

The PAX- system:

Deep Knowledge Ventures, the first company to appoint AI to its board of directors: