• Written by Azhar Aliyar, Director

In practice, enterprise AI systems like AsscherAi, which enable organisations to query their own data in natural language, often struggle to deliver in the printing and packaging sector, where job-shop economics and heavy setup costs prevail, making it harder to balance the need for security and privacy with the need for efficient procurement and deployment.

Challenges in printing and packaging

The printing and packaging industry poses unique challenges for enterprise AI systems, particularly in terms of setup and makeready losses, quoting accuracy, and substrate waste. For instance, a production manager at a packaging plant may need to query the system about the availability of a specific substrate, only to find that the system's response is based on outdated information, leading to unnecessary waste and delays. The system's inability to account for the complexities of job-shop economics, where setup costs are high and production runs are often short, can result in inaccurate quotes and inefficient production planning.

A shift supervisor at a printing plant may check the maintenance log to determine the best time to perform routine maintenance, only to find that the system's recommendations are based on a simplistic understanding of the production process, failing to take into account the intricacies of setup and makeready losses. This can lead to unnecessary downtime and reduced productivity.

Data processing and storage

When deploying an enterprise AI system like AsscherAi in the printing and packaging sector, it is essential to consider where data is processed and stored, and whether it trains anything. The system's ability to learn from data and improve its responses over time is critical, but it also raises concerns about data privacy and security. For example, if the system is trained on data from multiple sources, including supplier databases and customer records, there is a risk that sensitive information could be compromised.

The category manager at a packaging company may compare two suppliers based on their pricing and delivery schedules, using data from the enterprise AI system, but if the system is not properly secured, this data could be accessed by unauthorised parties, potentially leading to fraud or other malicious activities. To mitigate this risk, it is essential to implement robust security measures, such as encryption and access controls, to protect sensitive data.

Prompt injection as a data-boundary problem

Prompt injection is a significant concern in the printing and packaging sector, where the enterprise AI system may be vulnerable to malicious input that could compromise the security of the system or steal sensitive data. This is a data-boundary problem, where the system's ability to understand natural language queries can be exploited by attackers to inject malicious prompts. For instance, an attacker may craft a prompt that appears legitimate but is actually designed to extract sensitive information from the system.

To address this risk, it is essential to implement robust security measures, such as input validation and sanitisation, to prevent malicious prompts from being injected into the system. The system's developers can also use techniques such as data masking to protect sensitive information and prevent it from being accessed by unauthorised parties.

Least privilege for an assistant

When deploying an enterprise AI system like AsscherAi in the printing and packaging sector, it is essential to ensure that the system is granted the least privilege necessary to perform its functions. This means that the system should only have access to the data and systems that it needs to perform its tasks, and no more. For example, if the system is used to query the production schedule, it should only have access to the production schedule and not to other sensitive data, such as customer records or financial information.

Granting the system excessive privileges can increase the risk of a security breach, where an attacker could use the system to access sensitive data or disrupt production. To mitigate this risk, it is essential to implement robust access controls, such as role-based access control, to ensure that the system only has access to the data and systems that it needs to perform its functions. Organisations can also contact our support team for guidance on implementing least privilege access controls.

Limitations of enterprise AI systems

While enterprise AI systems like AsscherAi can be powerful tools for organisations in the printing and packaging sector, they are not a panacea for all security and privacy concerns. There are limitations to what these systems can do, and they may not be the right choice for every organisation. For instance, if an organisation has highly sensitive data that requires extreme security measures, an enterprise AI system may not be the best choice, as it may not be able to provide the necessary level of security and privacy.

In some cases, the costs and complexity of deploying an enterprise AI system may outweigh the benefits, particularly for small or medium-sized organisations. In these cases, alternative solutions, such as manual data analysis or simpler automation tools, may be more appropriate. Ultimately, the decision to deploy an enterprise AI system should be based on a careful consideration of the organisation's specific needs and requirements.