An integrated sociotechnical reference model for aligning artificial intelligence implementation in organizations
Using a Design Science Research approach, this study addresses the misalignment between AI implementation and organizational transformation by analyzing existing procedural models to construct the Integrated Sociotechnical AI Reference Model (ISAR-M), which synchronizes technical capabilities with organizational readiness to overcome disciplinary silos and ensure successful AI adoption.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Modern companies are racing to adopt artificial intelligence, hoping these smart systems will revolutionize how they work. Yet, despite the excitement and the strategic importance placed on this technology, most organizations are failing to turn their ambitions into real results. The problem is rarely that the computer code is broken or the algorithms are too weak. Instead, the failure usually stems from a disconnect between the technology and the people who must use it. When a company installs a new system, it is not just a technical upgrade; it is a change to the entire way the organization functions, involving new rules, new roles, and new ways of thinking. If the technology moves faster than the people can adapt, or if the people are not prepared for the new tools, the system stalls. This gap between what a machine can do and what an organization is ready to handle is the central challenge researchers are trying to solve.
A team of researchers at the Munich University of Applied Sciences set out to fix this disconnect. They noticed that while there are many guides on how to build artificial intelligence and many guides on how to manage business strategy, there is no single, unified path that connects the two. Existing guides often treat the technology and the human side as separate problems. Some focus entirely on the engineering steps, like data cleaning and coding, while ignoring how workers will react. Others focus on high-level business goals but offer no practical steps for implementation. This separation creates a dangerous situation where a company might build a perfect technical solution that no one knows how to use or that the company is not ready to support. To address this, the researchers analyzed forty-five different existing models for introducing artificial intelligence. They did not just read them; they broke them down to see exactly what steps they included and what they missed.
The analysis revealed that current approaches suffer from two main blind spots. First, many models are too focused on the inside, looking only at whether the technology works without asking if it fits the market or the broader business needs. Second, they are often too technocratic, meaning they prioritize the machine's performance over the human experience. They might ensure the software is fast and accurate but fail to design the system so that it is easy for a human to understand or trust. The researchers found that while almost all models agree that data quality and business value are important, very few provide concrete, step-by-step instructions on how to manage the human side of the equation, such as training staff or designing user-friendly interfaces.
To solve this, the team built a new framework called the Integrated Sociotechnical AI Reference Model. They did not invent this from scratch; instead, they acted like master builders assembling the best parts of the forty-five models they studied. They selected the two most comprehensive existing guides to serve as a foundation and then filled in the missing pieces with specific, high-quality steps from other models. The result is a single, cohesive roadmap that forces a company to move forward on two tracks at the same time: the technical track and the human track.
This new model divides the process into four main stages: Evaluate, Design, Realize, and Operate. In the first stage, Evaluate, a company must check if it is ready before spending any money. This is not just a technical check; it requires looking at the data, the legal rules, and the people. The model introduces a strict "gate" at this point. If the data is poor or the legal permissions are unclear, the project stops immediately. This prevents companies from wasting resources on ideas that are technically impossible or legally risky. It ensures that the business strategy is actually supported by the reality of what the company has to work with.
Once a project passes the first gate, it moves to the Design stage. Here, the model insists that the people who will use the system must be involved in creating it. Instead of engineers building a tool in a lab and handing it over later, the company must hold workshops where workers help design how the system will look and feel. This ensures the tool fits the actual job. At the same time, the technical team designs the system architecture. The model requires these two streams to happen in parallel. Before moving forward, another gate checks if the design is safe, ethical, and trusted by the people who will use it. If the workers are not ready or do not trust the system, the project cannot proceed, no matter how good the code is.
The third stage, Realize, is where the system is built and tested. The model recognizes that software can be updated very quickly, but people take longer to learn and adapt. To prevent the technology from moving too far ahead of the staff, the company creates safe spaces for experimentation. Workers can try out the system in a low-risk environment before it is used for real work. This helps them get comfortable with the new tools without fear of making mistakes. The model also requires that the system be tested not just for speed, but for fairness and security. A final gate checks if the system is ready to be launched. It asks a simple question: does the value the system provides justify the cost of rolling it out? If the answer is no, the project is paused.
The final stage, Operate, is about keeping the system working over the long term. Artificial intelligence systems can drift over time, meaning they become less accurate as the world changes. The model requires continuous monitoring to catch these changes. But it also requires continuous care for the people. The company must keep communicating with the staff, addressing fears, and updating their skills. The model includes a specific oversight group to ensure the system remains ethical and that human workers are not overwhelmed by the technology. This stage ensures that the relationship between the machine and the organization remains healthy as both evolve.
The researchers tested their new model by running it through several hypothetical scenarios. In one case, they imagined a company trying to launch a customer service bot without checking if their data was ready. The model's first gate stopped the project, saving the company from a costly failure. In another scenario, they imagined a team that built a highly accurate system but skipped the training for the workers. The model's final gate stopped the launch, forcing the team to go back and train the staff first. These tests showed that the model works as a safety mechanism, catching problems before they become disasters.
To see if the model would work in the real world, the researchers interviewed three experts who manage artificial intelligence projects in different types of companies. These experts confirmed that the idea of running the technical and human tracks side-by-side is necessary. They noted that in many current projects, the work of managing people and culture is often invisible and gets cut from budgets. The model makes this work visible and mandatory. However, the experts also pointed out that the model needs to be flexible. In fast-moving startups, the process might need to loop back and forth more quickly than the standard steps suggest. The researchers adjusted the model to include these feedback loops, ensuring it could be used by both large, stable corporations and smaller, agile teams.
The study concludes that the biggest barrier to artificial intelligence is not the technology itself, but the process of managing it. Success depends on a company's ability to synchronize the speed of its engineering with the speed of its people's adaptation. The new model provides a structured way to do this, turning abstract ideas about ethics and culture into concrete steps that can be checked and verified. It suggests that the future of artificial intelligence in business will not be defined by how smart the algorithms are, but by how well organizations can align their technology with their human workforce. By following this integrated path, companies can avoid the common pitfall of building powerful tools that no one is ready to use, ensuring that their investment in artificial intelligence actually delivers value.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.