AI is not a threat - it is an option. As with all new technologies, AI enables automation of processes and functions which have been manual until now.
This, of course, means that some job functions will disappear, but more importantly that new ones will be created.
I think we should think of AI as ‘Intelligent Assistants’ who can help us navigate the complexity that the human brain is not designed to handle.
For example, take an Excel spreadsheet. We might all be able to review 10 columns and 100,000 rows. It might be more difficult to handle 1,000 columns and 1,000 rows. To a machine it is completely the same.
A more concrete example is the work we are doing for a German car manufacturer.Their cars have more than 100 on-board sensors which, in the case of "incidents," collect detailed data on all aspects of the vehicle.
For the car manufacturer it is interesting to know whether an ‘incident’ is due to the driver or whether it is a manufacturing fault. In the latter case, it may be necessary to revoke vehicles.
With the help of Machine Learning, we have helped the car manufacturer to reduce the time required to process this information from one month to one day. This is equivalent to a business case of more than 70 million EUR per year.
Easy access to data is crucial
In several sectors, there is a principle that ‘developers’ must not have access to production data.
This principle works well for general software development and partly for business intelligence, but it does not work for data science, where you learn the ‘solution’ from data.
Here’s an example: a person works in a major technology company where they are responsible for the development of data science products. For the purpose of solving a task, they ask the ‘data department’ for a data set.
In addition, it turns out that the analysis they do generates very strange results. As the person looks more closely at the raw data, they discover what they’re reviewing is actually a test data set.
Another example is that companies, in addition to development, must be able to specify requirements and test AI solutions, but more than just describe the user interface and functionality, such as: "you must be able to send a message and receive a response ..." , "Messages appear as speech bubbles".
With AI solutions, we must relate to more technical requirements such as: “the solution must choose right in more than 90 per cent of cases”, “it must be possible to explain why the solution makes that choice”, “a small variation in an input parameter must not fundamentally change the result”.
POC, MVP and then what
The daily press gives the impression that the companies are farther ahead than they are.
Our impression is that many have made proofs-of-concepts, fewer launched real solutions and very few (if any) have redeemed the full transformation potential.
In particular, it seems difficult to scale from point solutions to a real strategic capability. Where ‘bottom-up’ works well tactically, there is a different need to set the framework and control ‘top-down’ when scaling.
Not surprisingly, it is difficult, as it has consequences for organisation, employees, data as well as technology.
Data is the raw material from which future products will be built
If the data is the raw material that the products of the future are built of, then it is important to have enough raw material when you need it. So please, if nothing else, make sure you get started recording all the data you can.
My view is that we are too cautious – not least because of GDPR – and that the ultimate consequence is that we are falling behind in international competition.
It is through the interconnection of data that it becomes possible to optimise holistically (end-to-end) as opposed to the point optimisation that is widespread today.
Here we really have an opportunity in Denmark and the health service is an obvious example – think about how much more health we could get if the systems talked together and could be optimised across private practitioners, hospitals and municipalities; and think about how our unique datasets in that area can form the basis of a number of new health-related services with associated export adventures.