The successful deployment of artificial intelligence at enterprises – in addition to technological improvements – requires not only the right organisational transformation, but also the development of employee attitudes and competencies, as well as the re-thinking of operational processes. These issues were addressed in a recent article in Computerworld magazine, written by our AI consultant Andrei Damsai and our co-founder Ottó Werschitz. The article is reproduced below without changes.
My art(ificial): intelligence
Change management and the introduction of machine learning
Predictors: the age of AI
In the 1950s, a group of scientists from different disciplines decided to develop the foundations of artificial intelligence (AI). At that time they did not suspect that it would be at least 60 years before the technology was mature and widely deployed. Today, in 2022, AI is no longer just a mystical vision of the future soaring to rocket science heights in laboratories, but a service available to small, medium and micro businesses. This is due, among other things, to the dramatic increase in the computing power of digital devices, the significant reduction in the cost of data storage, and the exponential increase in the amount of data required to train algorithms.
In practice, AI is most commonly understood as a machine learning methodology, whereby an algorithm recognises and learns relationships based on information extracted from data rather than pre-programmed rules.
People, processes, tools
One of the best known frameworks supporting organisational transformation and its management is the PPT (People, Process, Tools) methodology. According to this model, setting up the right professional/operational team, defining processes accurately and selecting the optimal technological tool-set are essential for successful organisational transformation. The introduction of AI can fit quite well into this framework, sometimes called the golden triangle, but it is worth bearing in mind the special characteristics of this field.
In the area of tools, it is relatively easy, as AI has a broad arsenal. There are low-cost IT infrastructures in the cloud (storage and computing capacity, IaaS), turnkey IT platforms (up to the operating system or system environment level, PaaS) and solutions directly related to AI (SaaS). These range from pre-trained models (e.g. Yolo), through modules (e.g. TensorFlow), to services available via APIs (e.g. DeepL).
However, the situation is different when it comes to defining and transforming processes. AI is a highly iterative sport, and the initial phase of most projects will have an uncertain outcome. In addition to business planning, this creates a special situation for change management. Initial specifications can at best only provide directions, decision-making relies heavily on information extracted from data in the process. It is in response to these challenges that methodologies such as CRISP-ML and frameworks supporting AI-specific processes have emerged (see UBER: Michelangelo).
A key factor for the successful implementation and operation of AI is the readiness of the team working directly with the technology and of the whole organisation. Knowledge of the basics of the technology and how it works should not be the prerogative of the technical staff alone. Therefore, it is recommended that effort and time be invested in training the company’s staff. There are several benefits to be gained from developing a core competence in AI within the organisation. Familiarity with the technology increases acceptance, helps to run the project and propagate further improvements (dissolving the feeling of a black-box process), and facilitates communication and collaboration with external partners (suppliers, service providers and consultants).
OK, but how do I become data-driven?
The use of AI does not equal data-driven, but it helps to a great extent achieve it. With the right information and forecasts, managers can make data-driven strategic and operational decisions that support operations. Closely linked to the integration of AI are the identification and review of good practices, the knowledge of internal opportunities (especially along the lines of data, infrastructure and processes), and the choice of appropriate integration points – projects or areas for transformation (a good pilot project can lay the groundwork for further improvements). It is worth considering whether the right expertise, energy and time are available at the organisational level to keep the tasks in-house or whether it is more efficient to bring in an external partner (consultant or integrator). The process is gradual – the organisation can rely fully on the supplier, get involved in the tasks and take the lion’s share of the work, in consultation with the partner. The implementation of AI is not isolated to a single function, but creates value at a system level, along people, processes and tools.
Original source in Hungarian: https://computerworld.hu/cwprint/computerworld-lapozo-h1systems-felho-kozelrol-307915.html