Recipe for smart services: data, brains and technology
Smart services have also become the key theme in the dialogue on citizen services. The data and possibilities already exist, but the beef is in how smart services are built and how we can make them deliver business value in an agile way. High-quality, rich data can provide valuable information to support business processes.
Data can be turned into information using advanced solutions, such as machine learning or artificial intelligence (AI), but it all starts with data. The quality of the data is directly correlated with the quality of the results. AI methods can be used to effectively refine data when that data is of high quality, there is enough of it and it is in a format that machines can understand.
The use of data begins with data refining. If data is in a natural language, it is first converted into a format that can be processed by a computer (natural language processing). For example, Google’s TensorFlow machine learning library has been taught to understand what is seen in a photograph. Meanings expressed in news, on social media, in patient records and pre-trial investigation records – which until now have only been available in a natural language, i.e. in text format – are converted into a computer language, showing how the meanings are related to each other. By contrast, numerical values – obtained from, for example, car speedometers – can be collected directly in the right format, knowing that the quality is excellent. We cannot truly utilize the data unless it has been collected and turned into the right format.
Understanding what we are doing
Before collected and refined data can be used as information, we need to understand what we are doing. It is essential to have a thorough understanding of the customer’s business logic, process or other object to be able to use the data to improve the quality or cost-effectiveness of the business process. Moreover, the right algorithms and AI methods are needed to transform the data into an usable format.
It is important to note that data research know-how is required in addition to technology. That is why Tieto works in close co-operation with universities and other companies with special expertise. When we have sufficient technology, sufficient expertise, relevant data and an understanding of the process to be developed, we can really get to work.
Lean Startup for fast development
Lean Startup is a method that can turn smart services into products very fast. The first phase is feasibility testing. A small amount of data is modelled to see what can be done with it. As successes pile up, data sources are added and analysis models developed, until a minimum viable product (MVP) is created, enabling the quick testing of the developed analysis models in a real-life environment. It typically takes one to three cycles, each lasting for three to five weeks, to get to the MVP phase. If the system to be developed is simple, the cycles can be shorter.
The speed at the project is greatly affected by the quality of the datasets – i.e. whether high-quality data already exists or whether it needs to be collected and structured first. All in all, the quality of the existing data is a major factor. Data structuring and cleaning can easily take up most of the time – or, if it fails, can destroy the whole project.
The Lean Startup method and getting data under the control are the key starting components. Algorithms and models are naturally important as well, but the availability and quality of data are paramount. What is needed after that, in order to create an analysis model that will deliver the desired benefit, is simply some brain capacity. It is important to remember that data and brains dictate whether a smart service will succeed. There is no magical artificial intelligence that will automatically lead to the desired outcome.
Read more about the opportunities of the future data-driven world: