The time of experimenting is over, and companies are beginning to demand real results from AI.
Three fundamental questions are:
Many companies currently have data and AI on their agenda. Their focus is very much on how to manage the data, what kind of data platform to create or what kind of organization to form for governing the data. Since there is an increasing pressure to achieve something, lots of experiments around the data are conducted, often labelled as Proof of Concept.
Pressure, however, is mounting to start seeing real-world AI applications. The time of experimenting is over, and companies are beginning to demand real results from AI. Companies can spend a lot of time setting up their data governance and platforms while business is eagerly waiting for the time when they can start projects making good use of the data. That, hopefully, puts pressure on IT and data organizations.
It would be wise to begin from the desired outcomes, to find real business challenges that can be resolved with the help of data and AI. When there is a clear need to resolve a problem and get real business benefits, then the work with data also becomes better focused.
That takes us to our second subject. As a “use case” for data and AI, companies often mention analytics. AI is considered yet another tool to analyse data. Technology companies are providing analytics tools on their platforms, so the focus is very much on the data itself and technical tools.
Another viewpoint is to make AI part of operations and process control. This puts everything in another perspective. When AI is embedded into existing operational systems at the mill, or AI algorithms are actually allowed to run the process, its value increases. Numerous customer cases already prove that.
This in turn forces you to think where AI should be running. Should it run far away in the cloud, close to analytics tools, or should it run close to operations, processes, and data? It is obvious that AI will become as business-critical a component as process and quality control systems currently are.
The jet/wire ratio is a good example of how cloud of cloud services jump into the AI discussion. Why is everybody interested in the jet/wire ratio, and why is this a problem?
Jet/wire ratio is an interesting process value because it affects so many things at the mill. Your chemical supplier needs it in their wet end chemistry analysing service, provided from their cloud. Your paper machine supplier may want it in their cloud service to control web strength. Another partner is offering you a web break analytics tool, and the jet/wire ratio is obviously a key data tag to follow. Your predictive maintenance and condition monitoring partners might also be interested in it and want it included in their cloud-based service.
This means there is a cloud of cloud services that all need your process data.
You might say you replicate the jet/wire ratio to your cloud and all your partners are welcome to use it for their particular needs from there. Big respect to companies who have advanced this far. The reality, though, is different in many companies. Unresolved issues are whether those cloud services are actually able to control the process, or how the customer who is buying those cloud services can utilise the results of AI algorithms in their own systems.
To conclude, data & AI strategy, technical platform, data governance, and experiments are highly relevant activities in companies right now – but the most important target should be to provide real value-adding solutions to create real and tangible business benefits.