Intentions and visions are great when Danish companies announce that they want to be "data-driven". At the same time, there are plenty of technically skilled women and men who can program algorithms and understand how to work with AI and Machine Learning.
Unfortunately, there are few successes, and one should ask the question: Why doesn't it work? The answer can be found, among other things, in the fact that the focus is almost exclusively on technology and technical skills, but the purpose is forgotten!
Technologies and methods within Data Science, AI and Machine Learning have enormous potential for use in companies and organizations, where many have long since recognized the business potential and gained an understanding that AI and related concepts are not just hype.
Success stories are shared widely, and we see both willingness and visions of being more "data-driven" and leveraging data optimally in both smart and innovative ways. However, the good intentions to stay ahead with intelligent AI solutions are often met with the same challenge: How do you actually create value with AI?
With the right approach, AI can make a huge difference in a wide range of business areas, including automation and streamlining of routines and processes, as well as processing and classification of text and images. But let it be said right away: It is not easy to succeed with AI, and the challenges can be many and perhaps even difficult to anticipate.
Many companies are eager to quickly get started with the use of AI to ensure their future competitiveness, but projects often come to a halt because there is not sufficient focus on how value is actually created. The right skills in a Data Science team are, of course, a prerequisite, but far from a guarantee of the company's success with AI. In reality, effort and commitment are required in several areas. This involves collaboration between model developers and the rest of the business, a collaboration that should be much closer than one might initially think.
It starts with a business problem
Before embarking on an AI project, it is necessary to involve the company's stakeholders and end-users from the beginning and identify their needs for the solution. It is ultimately about solving their problem and equipping them to use the solution. Therefore, it is of no use developing strong models in the engine room if no one intends to use or knows how to use them. Stakeholders and end-users should instead be involved so that together, a specific problem can be defined that needs to be solved, and clear success criteria can be formulated to determine when the final solution can be put into use.
With a clear and defined problem in hand, it is important to investigate whether it is possible to solve the given problem with the available data. It is important to identify and ensure access to relevant data that needs to be included in the solution. Additionally, it is important to ensure high data quality so that the business users can ultimately trust that the solution provides an output based on correct data.
At the same time, in many business areas, there may be significant ethical considerations to take into account before, for example, developing an AI model that is used to make decisions affecting people.
A successful AI project must, of course, be able to create value in the future. Therefore, it is necessary to have considered how models are put into production and monitored over time. The world is constantly changing, and so is data. Therefore, it is also necessary to have control over, among other things, processes for retraining models, which in turn requires a good MLOps setup that simplifies, monitors, and stabilizes the process.
Value creation with AI therefore requires a culture and mindset that inspires data and AI within the organization and motivates stakeholders and end-users to want to solve problems with sophisticated models. It is not enough to invest in technically skilled people or software, but the organizational setup and supporting processes must be in place to succeed with AI.
This can be achieved in different ways. Among other things, it is important that as a Data Scientist or model developer, you have a clear understanding of business problems and that you are able to communicate insights and results along the way to a non-technical audience. Equally important is for the business to open up and share their work processes, challenges, and needs. These are the subject of the solutions, and without them, no value is created. It is largely about trust and knowledge. To achieve this, the relevant parts of the business must be on board and their challenges must be in focus from day one.