A readiness checklist for operations AI
Operations teams can use AI to investigate signals preceding unplanned downtime by asking what deviated this shift, on which line, and how it compares with normal, then act while the deviation is still small, reducing downtime and increasing overall efficiency.
Every week, operations teams sit through meetings where they are forced to wait on answers to basic questions, such as the current stock levels or the status of maintenance requests. This is a clear sign that the organisation is ready to start using AI to answer questions about its own data. However, before starting, it is crucial to have certain elements in place to ensure a smooth and effective implementation.
Connected data
The first and most critical element is to have all relevant data connected and accessible. This includes data from various systems, such as enterprise resource planning (ERP), customer relationship management (CRM), and maintenance management systems. Without connected data, the AI system will not be able to provide accurate and comprehensive answers, which will lead to frustration and mistrust. For instance, a shift supervisor checks the maintenance log to see if a particular machine is available for use, but the log is not updated in real-time, causing delays and inefficiencies.
The consequences of not having connected data are severe, and can break the entire operations process. For example, if the inventory levels are not accurately reflected in the system, the organisation may end up with stockouts or overstocking, leading to wasted resources and lost sales.
Agreed definitions
Another essential element is to have clear and agreed-upon definitions of key terms and concepts. This is often overlooked, but it is crucial to ensure that everyone is on the same page. Most people get wrong the idea that definitions are obvious and do not need to be explicitly stated, but this is wrong. Without agreed definitions, the AI system may provide answers that are misleading or inaccurate, leading to confusion and mistakes.
For example, the term "inventory level" may have different meanings to different people, and it is essential to define it clearly and consistently. This can be done by creating a glossary of terms and ensuring that all stakeholders agree on the definitions. The category manager compares two suppliers, but if the definition of "lead time" is not clear, the comparison will be meaningless.
Permissions and access
Permissions and access control are also critical elements to consider before implementing AI. It is essential to decide who has access to what data and what level of permission they have. This will prevent unauthorised access to sensitive information and ensure that the AI system is used responsibly. Permissions should be decided before anyone asks, to avoid delays and misunderstandings.
For instance, the maintenance team may need access to the maintenance log, but not to the financial data. By setting clear permissions and access control, the organisation can ensure that the AI system is used effectively and securely. This is particularly important in operations, where data-driven decision making is critical.
A first question worth answering
A first question worth answering is one that operations already waits on, such as the current stock levels or the status of maintenance requests. This will help to build trust and demonstrate the value of the AI system. For example, the supply chain manager may ask the AI system to provide the current inventory levels of a particular product, and the system should be able to provide an accurate and up-to-date answer.
This will help to identify the potential benefits and challenges of using AI in operations and will provide a clear direction for further implementation. By starting with a question that is already relevant to operations, the organisation can ensure that the AI system is used to address real-world problems and improve efficiency.
Limitations and wrong choices
This AI checklist for operations does not provide a comprehensive solution for all operational challenges, and it is not a replacement for human judgement and expertise. There are times when it is the wrong choice, such as when the data is incomplete, inaccurate, or biased. In such cases, the AI system may provide misleading or inaccurate answers, leading to mistakes and inefficiencies.
Additionally, the AI system should not be used as a substitute for critical thinking and problem-solving. Operations teams should still use their expertise and judgement to interpret the answers provided by the AI system and make informed decisions. If you are unsure about how to implement AI in your operations, you can contact us for guidance and support.
Frequently asked questions
What are the key elements to consider before implementing AI in operations?
Connected data, agreed definitions, and permissions are crucial for a smooth implementation.
How can AI help operations teams reduce unplanned downtime?
AI can identify signals preceding downtime, allowing teams to act quickly and reduce stoppages.
Why are agreed definitions important for AI implementation in operations?
Agreed definitions ensure everyone is on the same page, preventing misleading or inaccurate answers.
What is the first question worth answering when implementing AI in operations?
A question that operations already waits on, such as current stock levels or maintenance requests.