Traditional organisations are struggling to implement artificial intelligence (AI) as part of their analytics portfolio. Furthermore, the use of AI in organisations is still relatively new, and research shows that early adopters should expect and manage technical challenges in order to minimise disappointments. The Artificial Intelligence Adoption Framework (AIAF) provides managers that are interested in the adoption of AI in organisations with some insights into the enabling factors to support them in their AI adoption initiatives. Firstly, the framework provides a high-level overview of what a data-driven organisation is. Then it gives information on how to facilitate the organisational AI adoption decision stages.

The AIAF consists of:

1. Introduction to AI and Data-Driven Organisations

A data-driven organisation is an organisation that uses analytical tools and abilities, that creates a culture to integrate and fosters analytical expertise and acts on observed data to achieve benefits. However, the quest for organisations to improve their data-drivenness is much more comprehensive than simply supporting organisational decision making by means of reports and dashboards. The pursuit for data-drivenness should also include the enabling of forward-looking analysis, where organisations not only use data to report on the past, but utilise models to predict the future Organisations that have successfully adopted AI technologies as part of their analytics portfolio have reached an advanced state of data-drivenness. AI can be classified based on intelligence (artificial narrow intelligence, artificial general intelligence, and artificial superintelligence), based on technology (for example, machine learning, deep learning, and natural language processing) or based on function (conversational, biometric, algorithmic, and robotic). AI is centred around the idea that mental processes can be simulated or replicated in computers. But, the idea is not that AI will replace humans, but rather that AI allows for augmented analytics within a human machine-partnership. Additonally, AI has the ability to learn from data and act autonomously. These autonomous actions might impact humans and will lead to important legal and ethical questions whose answers affect both producers and consumers of AI technology. Therefore, it is important to acknowledge that AI adoption should be done in a responsible manner, and it is the duty of AI researchers to ensure that the future impact of AI is beneficial.

2. Facilitating the AI adoption decision process

The adoption decision process are the phases that potential adopters of AI will go through when deciding to adopt AI as part of their analytics portfolio. To facilitate the AI adoption decision process five phases are identified, namely: 'increase AI knowledge', 'create AI adoption buy-in', 'support the decision process', 'provide implementation support' and 'measure and confirm outcome'.

In the AIAF, each phase contains the enabling factors related to the phase and can be used by organisations to support their AI adoption initiatives. Furthermore, AI adoption should be seen as a continuum. Organisations can always have more data, AI is ever evolving, therefore keeping up with the technology advancements is a continues process.

Increase AI knowledge
Increasing the knowledge of AI in organisations is the first stage of the innovation-decision process and occurs by exposing individuals and the organisation to AI. The focus in on increasing the awareness of the innovation.

Create AI adoption buy-in
Many organisations may have knowledge about new innovations but have not adopted them yet, therefore the goal of this stage is to develop a favorable attitude towards the innovation.

Support the decision process
During the decision stage of the innovation-decision process, the individual weighs the advantages and disadvantages of adopting AI and forms an intent to adopt or reject AI adoption. Furthermore, this is not only an adoption decision, but a financial investment.

Ensure implementation success
There is a difference between the decision to adopt an innovation and to implement the innovation in practice. During the implementation stage of the innovation-decision process, the individual puts the innovation into use, either implementing successfully or unsuccessfully.

Measure and confirm outcome
The last stage of the innovation-decision process deals with the confirmation and continuation regarding the innovation. This stage includes the recognition of the benefits of using AI, integrating the AI into an organisation's routine and promoting it to others.

Drive sustained adoption momentum
As AI is a moving target and at the frontier of computational advancements, AI adoption should be seen as a continuum. As a result it is essential to conserve AI adoption momentum, by implementing a continuous improvement mindset. This can be supported by an innovative company culture, by ensuring that the value of adopting AI is known and constantly removing barriers that might hamper the adoption process.

3. Technical enabling factors to support AI adoption

Only realising the benefits of utilising AI and having the required skills is not enough to lead to AI adoption success. The required technical aspects should also be put in place, for example a data and AI platform.

From the analysis of the data, 14 topics emerged as enabling factors to support the technical aspect of organisational AI adoption. ‘Automate informed decisions’, ‘competitive advantage’, ‘efficient decision making’ relates to the importance of having a business case for implementing and adopting AI (Chui 2017). ‘Invest in compatibility’, ‘implement standards’ and ‘architecture strategy’ are all related to IT governance and operational processes, which can be supported by MLOPS (Liu et al. 2020) and governance bodies (Ienca 2019). ‘Process and technology knowledge’, ‘democratise AI’, ‘demonstrate ease’, ‘pilot projects’, ‘test systems’, ‘showcase benefits’, ‘awareness methods’, ‘highlight positive outcome’ could all be grouped under the objective of achieving democratisation of AI in organisations. Furthermore, the study highlighted the importance of an enterprise data platform to support analytics and AI, which provides organisational wide data asset capability and increase data reliably.

Establish a data and AI platform
Organisations need to establish not only the AI platforms, but also ensure the data is available which AI experts need.

Increase AI democratization
Access to data and AI platforms should not be limited to a select few.

Implement AI governance
Just like in other technologies, best practices and IT governance is also applicable in the case of AI.

Ensure positive business case
Organisations should implement AI, not only for the sake of technology, but because there is a positive business case.

4. Critical success factors for AI adoption

The AI adoption critical success factors relate to the technological, environmental and organisational elements that organisations should consider when adopting AI as part of their analytics portfolio.

The AI TOE considerations are the technological, environmental and organisational elements that organisations should consider and relate to the critical success factors when adopting AI. From a technological point of view, organisations should ensure that the needed IT infrastructure in place. This involves the setting up the required data ecosystem and buying or building the appropriate AI tools, this should be done in such a way that it can lead to a relative advantage for the organisation. Furthermore, the characteristics of the technology should allow for observability which enables transparency and explainability. AI solution development should be done in a manner that renders the models more understandable to stakeholders and addresses AI interpretability needs. Top management support, having access to the required skills, competencies, and resources are some of the organisational success factors in adopting AI. Additionally, in the context of an organisation's subjective norms, ensuring fairness in AI is another organisational consideration. Additionally, considerations such as slack, absorptive capacity and culture play an important role in adoption. A competitive environment is one of the main factors that influence organisations to adopt AI. Aspects such as governmental regulations, customer readiness and industry pressure are other examples of critical environmental considerations for organisations when striving to adopt AI. Additionally, aspects such as a regulatory environment insists that accountability in the organisation is set in place. Lastly, AI also impacts its environment, the energy consumption of running large scale AI deep learning models, should not be underestimated and the environmental impact thereof cannot be ignored.

5. Differences between AI and traditional data-driven technologies

Understanding the similarities and differences between adopting more ‘traditional’ data-driven technologies and AI can be very helpful for managers within organisations as this information will allow them to use their experience from adopting other traditional data-driven technologies and assist them to understand the important differences. Most TOE considerations related to traditional data-driven technologies and AI are the same, however, some fundamental and impactful differences exist.

Traditional data-driven technologies are easier to understand than AI, this leads to the challenge of building AI knowledge and to democatise AI. Furthermore, traditional data-driven technologies are more human-centered than AI. Therefore, the human aspects for AI should take special care, for example ensuring ethical AI and preserving human control over AI. Lastly AI has the ability to learn and act autonomously, this give AI the ability to potentially lead to a lot of efficiencies, however, the impact of AI and automation must be considered.

Conclusion

The framework has the objective to assist organisations with a high level overview of enabling factors related organisational AI adoption, specifically for the use of AI to as part of an organisation's analytics portfolio. Please feel free to provide feedback on the framework via the feedback functionality via Ailea.

---