• Written by Muhammed Abdulla NC, CEO

This week, the production manager at a large manufacturing plant must decide where to allocate the budget for quality control, and currently, this decision is based on intuition and past experiences, rather than a thorough analysis of the costs associated with each step of the AI workflow.

The difference between unit cost and total cost of an AI workflow

The unit cost of an AI workflow is often the focus of cost optimisation efforts, but it is the total cost that really matters, as it takes into account the volume of data being processed and the complexity of the workflow. A reduction in unit cost may not necessarily lead to a reduction in total cost, especially if the volume of data increases or the workflow becomes more complex.

For instance, a shift supervisor may notice that the cost per unit of a particular inspection process has decreased, but the total cost of the process has actually increased due to an increase in the number of units being inspected. This highlights the need to consider the total cost of the AI workflow, rather than just the unit cost, when making decisions about cost optimisation.

Caching, batching and model routing as the three real levers

Caching, batching, and model routing are the three main levers that can be used to optimise the cost of an AI workflow. Caching involves storing the results of expensive computations so that they can be reused, rather than recalculated, which can significantly reduce the total cost of the workflow. Batching involves processing multiple units of data at once, which can reduce the overhead costs associated with each unit. Model routing involves directing data to the most appropriate model, based on the specific requirements of the data, which can reduce the computational costs associated with processing the data.

By using these three levers, organisations can reduce the total cost of their AI workflows, while also improving the quality of the output. For example, a category manager may use caching to store the results of a complex calculation, so that it can be reused for multiple units of data, rather than recalculated each time.

The quality data this depends on

The ability to optimise the cost of an AI workflow for quality depends on having access to high-quality data, including inspection records, non-conformance reports, batch genealogy, and customer complaints. This data provides the necessary insights into the workflow, allowing organisations to identify areas where costs can be reduced, without compromising quality.

For instance, inspection records can provide information on the frequency and type of defects, which can be used to identify areas where the workflow can be optimised. Non-conformance reports can provide information on the causes of defects, which can be used to target specific areas of the workflow for improvement. Batch genealogy can provide information on the origin and history of the data, which can be used to identify patterns and trends that can inform cost optimisation efforts.

What it changes about defects and CAPA records

The use of AI cost optimisation for quality can significantly change the way that defects are traced and CAPA records are managed. By using data and analytics to identify the root causes of defects, organisations can reduce the time and cost associated with tracing defects back to a particular batch or supplier. Additionally, CAPA records can be used to identify areas where the workflow can be improved, reducing the likelihood of future defects.

For example, a quality manager may use AI cost optimisation to identify a particular supplier that is associated with a high rate of defects, and use this information to target specific improvements in the workflow. This can help to reduce the number of defects, and also reduce the time and cost associated with managing CAPA records.

Measured with key performance indicators

The effectiveness of AI cost optimisation for quality can be measured using key performance indicators such as first pass yield, defects per unit, and time to root cause. These metrics provide a clear indication of the quality of the output, and can be used to evaluate the success of cost optimisation efforts.

For instance, a production manager may use first pass yield to evaluate the effectiveness of a particular cost optimisation effort, and adjust the workflow accordingly. Defects per unit can be used to identify areas where the workflow can be improved, and time to root cause can be used to evaluate the speed and effectiveness of the defect tracing process. To learn more about how to implement AI cost optimisation in your organisation, visit our production page.

What this does not do

AI cost optimisation for quality is not a replacement for human judgement and expertise, and it is not a guarantee of perfect quality. It is a tool that can be used to support and inform decision-making, but it is not a substitute for careful planning and execution. Additionally, AI cost optimisation for quality may not be the right choice for every organisation, and it is important to carefully evaluate the potential benefits and limitations before implementing it. If you have questions about whether AI cost optimisation is right for your organisation, contact us to learn more.