Adopting AI into working routines in a reasonable and economical manner will be crucial for small and big businesses. However, a recent study by the OECD shows that large firms are much faster in adopting AI technology than smaller companies. The study also shows that the obstacles to productive AI use differ depending on the company size. In a study for the project “Zentrum Zukunft der Arbeitswelt”, our colleagues from the University of Stuttgart IAT found that even German small and medium enterprises (SMEs) face a wide range of obstacles that hinder faster AI adoption. Thus, there is no quick and easy solution for companies that genuinely want to benefit from the technology. Especially SMEs may need different tools and formats of support at different stages on their AI learning journey.

What comes first: framework or projects?

Most SMEs start their AI journey with an initial flagship project. Often, this helps decrease negative attitudes towards AI among the employees and shows that the company uses the technology in a trustworthy and reliable manner. Specific change processes that influence the general AI readiness of the company are postponed until the flagship project will have been concluded successfully. In most cases, this seems to be a reasonable approach. Other companies, mostly the bigger ones and those that already have concluded their first AI introduction projects, are looking for ways to ensure that they improve the framework conditions that increase the chances of success for future AI introduction projects. However, this does not mean that framework conditions always come second. Some are vital prerequisites even for the first project to be successful, or to be concluded in a reasonable time frame. Eventually, companies need to work on favorable framework conditions, such as an overall AI strategy, competence management, organizational culture, and technical setup parallel to their individual projects. This depends on their status quo regarding key requirements for truly benefiting from AI technology.

Empirically founded tools by the KI-ULTRA project

In the project “KI-ULTRA” (funded by the Ministry of Labor and Social Affairs, BMAS), we have developed two guides that are available free of charge – in German, English, and French. To this purpose, we accompanied 29 participating organizations from Germany (including large companies and SMEs) for approximately 1.5 years while they were working on the introduction of an AI application in their working environment and condensed our findings in the guides.

The project level

A project to implement AI in the workplace should go through four main phases, as described in our guide to the implementation of AI projects:
1) setting goals and thoroughly analyzing the requirements, including stakeholders, key processes, and legal aspects.
2) starting the project and providing resources, defining the team, choosing the technology and – if required – getting the necessary training data.
3) developing the concepts and the technical solutions, such as the AI model, system and data architecture, as well as technical security measures
4) going live and making sure that learning systems are constantly maintained and outcomes – also regarding the employees using it – are evaluated.

It is also wise to define, for each phase, under which circumstances the project should be terminated. In each phase, different aspects may or may not be considered and single steps can be omitted, depending on the situation.
Using this or any other adequate process model can help ensure that important steps are not overlooked and reduces the risk of problems at a later stage.

The strategical level

For the success of such projects, favorable organizational conditions play a crucial role.

1) An AI strategy connects individual projects to the overall objectives, the vision or the mission of the organization. It helps set boundaries and guarantee top level support, which is often crucial for success.
2) Creating a common understanding of how employees of a company want to work with AI is essential for acceptance.
3) Organizational culture shapes the way projects are planned, executed, and accepted. A “data culture” may be nice to have for companies in which AI and data play a major role but without trustful communication and a constructive way to deal with errors, projects in areas of high uncertainty are unlikely to succeed.
4) Competence management means defining today which skills the company will need in the next years and what employees will need to learn to work with the AI tools envisioned or in implementation.
5) Obviously, data and technical infrastructure need to be available and up to date.

The KI-ULTRA guide to strategy and change toward the use of AI gives practical advice on all these aspects and the KI-ULTRA evaluation toolkit (module 1 – “transformation needs”) helps companies identify which aspect to focus on.

Real life examples

Case reports from 13 different organizations that participated in the project demonstrate the diversity of challenges organizations face when introducing AI in the workplace. These reports summarize the specific AI projects covering the starting situation and use case, the technical solution, the stakeholders involved, challenges and lessons learnt. Their purpose is to inspire organizations to create their own use cases.

In the end, it matters less where we start, as long as both the project level and the strategic level are kept in mind. Any activity on one of these sides can create positive momentum on the other.

Leselinks:

Jan-Paul Leuteritz

Autorenprofil - Website



Kategorien: Künstliche Intelligenz
Tags: ,