• Written by Azhar Aliyar, Director

Can we really get a single, reliable view of our printing and packaging operations, or will we always be stuck with separate systems for estimating, production, and inventory? This question comes up regularly in meetings, and the answer is not as straightforward as it seems, because printing and packaging has some unique challenges that make integration harder.

Challenges in printing and packaging

Printing and packaging is a job-shop environment with heavy setup costs, which means that setup and makeready losses can be significant. This makes it difficult to get accurate quotes, as the cost of setting up a job can vary greatly depending on the specific requirements. Additionally, substrate waste can be a major issue, as different jobs may require different materials and sizes. All of these factors make it harder to integrate systems, as the complexity of the operations can lead to inconsistencies and discrepancies between different systems.

For example, a production manager at a packaging plant may need to check the inventory levels of different substrates, as well as the production schedule, to determine the most efficient way to run a job. However, if the inventory system and the production scheduling system are not integrated, this can lead to errors and inefficiencies. Most people think that this is just a matter of implementing a new software system, but they are wrong - the complexity of the operations and the need for accurate data make it a much more difficult problem to solve.

Read paths, write paths and data consistency

When integrating AI systems with the systems of record in printing and packaging, it is essential to consider the read paths and write paths. The read path refers to how the AI system accesses data from the systems of record, while the write path refers to how the AI system updates the data in the systems of record. The write path needs more care, as it can create inconsistencies and discrepancies if not done correctly. If the AI system is not properly synchronized with the systems of record, it can lead to data inconsistencies and errors.

A shift supervisor at a printing plant may use the AI system to check the production schedule and inventory levels, but if the AI system is not properly updating the data in the systems of record, this can lead to errors and inefficiencies. For instance, if the AI system shows that a certain material is in stock, but the inventory system shows that it is not, this can cause delays and mistakes.

Reconciling definitions

Different systems may have different definitions for the same terms, which can lead to inconsistencies and discrepancies when integrating AI systems. For example, the production scheduling system may define a "job" as a specific set of tasks, while the inventory system may define a "job" as a specific set of materials. Reconciling these definitions is crucial to ensure that the AI system is using the same terminology as the systems of record.

This can be a difficult task, as it requires a deep understanding of the different systems and how they are used. However, it is essential to get it right, as inconsistencies in definitions can lead to errors and inefficiencies. By taking the time to reconcile definitions, organizations can ensure that their AI systems are integrated correctly and providing accurate and reliable data. You can find more information on this topic by visiting our website via this link.

Latency budgets and precomputation

Latency budgets refer to the amount of time that an AI system has to respond to a query or request. In printing and packaging, latency budgets can be tight, as production schedules and inventory levels need to be updated in real-time. To meet these latency budgets, AI systems can use precomputation to prepare data in advance. Precomputation involves calculating and storing data that is likely to be needed, so that it can be quickly retrieved when needed.

For example, a category manager at a packaging plant may need to compare the prices of different suppliers. By precomputing the prices and storing them in a database, the AI system can quickly retrieve the data when needed, rather than having to calculate it in real-time. This can help to meet tight latency budgets and ensure that the AI system is providing accurate and reliable data. If you have any questions about how to implement precomputation in your AI system, you can contact us via this link.

Limitations of AI systems integration

While AI systems integration can provide many benefits, such as improved accuracy and efficiency, it is not a solution for every problem. In some cases, the complexity of the operations and the need for accurate data may make it difficult or impossible to integrate AI systems. Additionally, AI systems integration may not be suitable for organizations with limited resources or expertise.

In these cases, other solutions, such as manual data entry or custom software development, may be more appropriate. It is essential to carefully evaluate the needs and resources of the organization before attempting to integrate AI systems. By doing so, organizations can ensure that they are using the right tools and techniques to meet their specific needs and goals. Some of the factors to consider when evaluating the suitability of AI systems integration include the complexity of the operations, the availability of resources and expertise, and the need for accurate and reliable data.