• Written by Muhammed Abdulla NC, CEO

In theory, applying AI to production should streamline processes and improve efficiency, but in practice, it often falls short due to recurring failure patterns. A common mistake is automating a process without first mapping it out, which can lead to poorly designed systems that fail to deliver the expected benefits.

Automating a process nobody mapped

When implementing AI in production, it's crucial to understand the existing processes and workflows. However, many organisations rush into automation without taking the time to map out their processes, resulting in systems that are not tailored to their specific needs. This can lead to inefficiencies and errors that might have been avoided with proper planning. For instance, a shift supervisor checks the maintenance log, but if the log is not accurately updated, the AI system will not be able to provide reliable insights.

A thorough process mapping exercise can help identify areas where AI can add the most value, ensuring that the system is designed to meet the organisation's specific requirements. By taking the time to understand the existing workflows, organisations can create a more effective and efficient AI-powered production system.

Treating output against plan as complete when it is not

Most people get wrong the idea that AI output against a plan is a complete solution, when in fact it's just the starting point. Treating AI output as the final answer can lead to oversights and mistakes, as the system may not have considered all the relevant factors. A category manager compares two suppliers, but if the AI system only considers price and not quality or reliability, the comparison will be incomplete.

It's essential to review and refine the AI output to ensure it's accurate and comprehensive. This may involve checking the data used to train the AI model, verifying the assumptions made by the system, and testing the output against real-world scenarios. By doing so, organisations can ensure that the AI system provides reliable and actionable insights that support informed decision-making.

Giving the capability only to the analyst who already had it

Giving the AI capability only to the analyst who already had it is a mistake that limits the potential benefits of the system. This approach can create a bottleneck, where only one person has access to the insights and expertise provided by the AI system. In a large organisation, this can hinder collaboration and decision-making, as others may not have the same level of understanding or access to the information.

By making the AI capability available to a broader range of users, organisations can promote collaboration and ensure that everyone has access to the insights and expertise they need to make informed decisions. This can be achieved by providing training and support to users, as well as ensuring that the AI system is user-friendly and accessible. For more information on implementing AI in production, visit our production page.

Declaring success against a metric with no baseline

Declaring success against a metric with no baseline is a common mistake that can lead to misleading conclusions. Without a baseline, it's impossible to determine whether the AI system has had a positive impact or not. For instance, if an organisation implements an AI-powered quality control system, but doesn't track the defect rate before and after implementation, it won't be able to determine whether the system has improved quality.

To avoid this mistake, organisations should establish clear baselines and metrics before implementing an AI system. This will enable them to measure the impact of the system and make data-driven decisions. By tracking key performance indicators and comparing them to the baseline, organisations can determine whether the AI system is having the desired effect and make adjustments as needed.

Some common metrics used to evaluate AI systems include productivity, quality, and customer satisfaction. By tracking these metrics and comparing them to the baseline, organisations can get a clear picture of the system's performance and make informed decisions about its continued use.

When AI is the wrong choice

AI is not always the right solution for every problem, and organisations should carefully consider whether it's the best approach before implementing a system. In some cases, alternative solutions such as process improvements or training may be more effective and efficient. For example, if an organisation is struggling with data quality issues, it may be more effective to address the root cause of the problem rather than trying to use AI to work around it.

Organisations should consider factors such as the complexity of the problem, the availability of data, and the potential return on investment when deciding whether to implement an AI system. By taking a thoughtful and nuanced approach, organisations can ensure that they're using AI in a way that adds value and supports their goals. If you're considering implementing an AI system and want to discuss your options, visit our contact page to get in touch with our team.