• Written by Azhar Aliyar, Director

A category manager at a large retailer compares two suppliers using AsscherAi, but the definitions of "supplier" and "category" are not consistent across the organisation, causing confusion and incorrect results. This situation highlights the importance of deciding where the expertise for AI systems like AsscherAi should sit, and who should own the definitions that drive the system.

What each option genuinely commits you to

A central AI team commits you to a top-down approach, where the team defines the terms and rules that govern the system, and ensures consistency across the organisation. This approach can be beneficial in maintaining a unified view of the data, but it can also lead to a disconnect between the central team and the various departments that use the system. On the other hand, embedded AI teams commit you to a more decentralised approach, where each department or team has its own AI expertise and defines its own terms and rules. This approach can be more flexible and responsive to local needs, but it can also lead to inconsistencies and duplication of effort.

The choice between a central AI team and embedded ones is not just about the location of the expertise, but also about the ownership of the definitions that drive the system. In a centralised approach, the central team owns the definitions, while in a decentralised approach, each department or team owns its own definitions. This difference in ownership can have significant implications for the consistency and accuracy of the results produced by the system.

The case where the other one is right

There are situations where a central AI team is not the best choice, and embedded teams are more suitable. For example, in a large organisation with diverse business units, each with its own unique requirements and terminology, a decentralised approach may be more effective in meeting local needs. In such cases, embedded AI teams can provide the necessary flexibility and responsiveness to local requirements, while a central team may struggle to keep up with the diverse needs of the organisation.

On the other hand, there are also situations where a central AI team is the better choice, such as in organisations with a strong need for consistency and standardisation across departments. In such cases, a central team can ensure that the definitions and rules governing the system are consistent and unified, which can be critical for maintaining data integrity and accuracy.

The cost that only shows up later

One of the costs of choosing between a central AI team and embedded ones is the cost of integration and maintenance. In a decentralised approach, each department or team may develop its own AI solutions, which can lead to duplication of effort and inconsistencies in the results. Integrating these solutions can be time-consuming and costly, and may require significant resources to resolve the inconsistencies and ensure data integrity.

In contrast, a centralised approach may require significant upfront investment in developing a unified system, but can reduce the costs of integration and maintenance in the long run. However, this approach can also lead to a rigid system that is difficult to adapt to changing local needs, which can be a significant cost in its own right. To mitigate these costs, it is essential to carefully consider the trade-offs between a central AI team and embedded ones, and to develop a clear plan for integration and maintenance, as described on our website.

A way to decide without a six month evaluation

Deciding between a central AI team and embedded ones does not necessarily require a lengthy evaluation period. One approach is to start with a small pilot project that tests the feasibility of both approaches, and to use the results to inform the decision. This approach can provide valuable insights into the strengths and weaknesses of each approach, and can help to identify the most critical factors that will influence the success of the system.

Another approach is to consider the organisational culture and the existing relationships between departments. If the organisation has a strong culture of collaboration and cooperation, a central AI team may be more effective in leveraging this culture to develop a unified system. On the other hand, if the organisation has a more siloed culture, embedded AI teams may be more suitable for meeting local needs. For more information on how to decide between a central AI team and embedded ones, please contact us.

What this does not do

The decision between a central AI team and embedded ones is not a solution to all the challenges of implementing AI systems like AsscherAi. It is essential to recognise that this decision is just one part of a larger process of developing and implementing an effective AI strategy. Other critical factors, such as data quality, user adoption, and ongoing maintenance, must also be carefully considered to ensure the success of the system.

Furthermore, the choice between a central AI team and embedded ones is not a one-time decision, but rather an ongoing process of evaluation and adjustment. As the organisation and its needs evolve, the decision may need to be revisited and revised to ensure that the AI system remains effective and aligned with the organisation's goals. By recognising the complexity and nuance of this decision, organisations can develop a more effective AI strategy that meets their unique needs and requirements.