AI in the printing industry – may remain an abstract desire for many
Artificial intelligence (AI) has a media presence in our industry comparable to the paper price situation in the past year 2022. Few opportunities are missed to present AI as the grandiose problem solution of the future. This stirs up hope and creates desire – good for all those who have an “AI solution” in their portfolio of offerings. It is undisputed that the rapid and increasing speed at which AI models and their capabilities are currently developing is impressive.
Nevertheless, Apenberg & Partner perceives a distorted picture in the market. Not because the capabilities of artificial intelligences are exaggerated, but because far too little is said about the necessary prerequisites.
Because: AI is not a magic wand that you can unpack today, wave over your own company tomorrow, and then hold incredible insights in your hand. Rather, AI is a tool and like every tool has concrete use cases with concrete prerequisites. As long as the prerequisites are not met, the tool will not work. For example, trying to sink a nail into a cabinet wall with a cordless screwdriver will probably drive you to despair. Without the necessary slots in the head of the nail or rather screw (prerequisite), the tool (cordless screwdriver) does not have the required point of attack.
The same applies to AI. In addition to the operating requirements of the tool (power, computing power, etc.), there are also requirements for the material that is to be machined or processed with the tool. In the case of artificial intelligence, the material is data. These must exist in the first place, just as you first need a screw in the above example. But that is not enough. Depending on which form of artificial intelligence is to be used in concrete terms, the data will also have to have concrete properties. If these properties are missing – such as the slots in the screw head – the material cannot be used in conjunction with the tool.
Now, the requirements for the material (data) differ depending on the tool (AI variant). Nevertheless, some basic, universal requirements exist. For example, you will need the data in a certain quantity and up to date, which is why automatic generation of the data becomes a mandatory requirement in many cases. Likewise, the data that your company generates automatically should have a sufficient relation to reality, i.e. be valid, so that the findings of the AI remain correspondingly applicable in the real world (see “shit in, shit out“). Last but not least, the data should meaningfully represent a delimited sub-process or the overall process. For example, you will not be able to implement sustainable sales management without comparing the actual performance (post-calculation) against the sales values assumed in sales (pre-calculation).
From our point of view, these requirements are discussed too rarely. One notable exception is the entertaining talk by Tobias Kaase and Dominik Haacke, which I was able to enjoy at this year’s OPS. Here, the management of media print solutions from Paderborn describes their rocky but worthwhile path to the use of AI – and this much should be said in advance: they talk about four phases (data strategy, data technology, business intelligence & analytics, and then AI) and just as many years of costly transformation of their own processes and the software solutions used.
Compare this reality with the finding that 76 percent of respondents to a representative survey by Apenberg & Partner say they do not trust their own data, and you have to conclude that AI is further away for many companies in the printing industry than they themselves might think. How do you see it? What experiences have you had on your way towards digitization? Let’s talk about it!