G-Agent in Manufacturing
How the G-Agent works in the Manufacturing industry: Given that supply chain as well as shop floor operations in Manufacturing are complex, a pure delegation of authority to AI is most likely not good enough. The example thus shows how AI based design can be matched with the G-Agent based governance:
Shopfloor operations in Manufacturing have been extensively addressed with standard operations procedures (SOP) and parameters to secure Good Manufacturing Practices (GMP). Process analytics have tried to understand the stochastic part of it and increase transparency where typically judgment of the operator and thus heuristic decision making comes into play. Many consultants have spent decades to make such heuristics explicit. Software companies have tried to translate such heuristics into proper coding, with mixed outcomes dependent on the complexity of operations.
Typically, the approach to develop and define operations procedures has been rule-based in the past. The supply chain managers and shopfloor operators have referred to rules of lot size optimization, sequencing, network and material balancing and so forth. A weak point of such rule-based approach, however, is its limited potential to cope with the upcoming complexity in Manufacturing and its individualization. There is a seeing-knowing gap to recognize existing complexity as well as a knowing-doing gap to manage complexity in the right way. You might remember Ashby`s law of “requisite variety” suggesting that only complexity “eats” complexity. This is exactly where the two types of gaps are coming from.
Now, the technology to recognize complexity (and thus close the seeing-knowing gap) with means of better diagnostics tools, digital twins and sensors has developed rapidly. E.g., tracking & tracing tools help to understand the structure of the supply chain network.
For managing higher complexity (i.e., closing the knowing-doing gap), requirements have always been clearly defined in terms of self-organization, however, there were little tools and solutions, which practically helped to implement. In shop floor operations for example, first attempts refer to Wildemann’s Modular Factory, Warnecke’s Fractals, then Ptak`s DDMRP and the begin of Machine Learning. Proper technology to manage higher complexity in a self-organized fashion and at scale has been pre-mature for long.
This is where artificial intelligence (AI) come into play. Manufacturing becomes self-trained and self-organized rather than rule based in a conventional sense. Instead of teaching “golden standard” based on rules, AI will imitate proven practices. Faced with a diagnosis, the neural network chooses a path based on what operators have done in thousands of similar situations before, successfully. Therefore, Manufacturing players will need to fuel their systems with a huge number of quality proven examples of effective operations. The pure mass of data facilitating self-learning will not only allow to imitate practices, however, there is a good chance that it will create new practices to manage the Manufacturing supply chain and operations even more effectively in a “lot size 1” world.
That said, data governance must evolve towards being AI ready. The Chief Operating Officers in Manufacturing need to be able to determine the degree of resilience of their supply chain and be comfortable with GMP compliance. Of course, there are procedures to change procedures but, as of now, they are not necessarily AI-compatible. The G-Agent, therefore, is asking for a real change process – not limited to a “technical” governance topic – where AI and engineering cultures come together, appreciate each other and create a new, innovation-driven Digital Manufacturing mindset.
Let’s take the example of deploying the G-Agent for any kind of Operations AI related to a Pharmaceuticals Product production. This can be complex, but here is a simplified scenario to illustrate the subject:
- Data source: PPS best practices: Databases with PPS schemes, supply chain planning and shop floor histories.
- G-Agent Functions (including GAuth+):
– Data validation: The G-Agent receives data from validated operations and supply chain databases.
– Condition check: The G-Agent checks whether the data meets certain criteria (e.g. lot size, product characteristics, order type, etc.).
– PPS validation: The G-Agent provides a PPS validation, which is being compared with the output of the PPS AI supporting the operations. In case of deviations, an alert is being created. - G-Agent process
– Expert review: The operations manager reviews the the G-Agent alerts, compares it with the outcome of the Operations AI and either confirms or rejects it. In the first case, the G-Agent will be completed
– Automated notification: The G-Agent automatically sends notifications to the operations governance team based on the results, like COO and CHRO. If new best practice based on actual experience is being created, reviews and approvals will be initiated to fuel both the Operations AI as well as G-Agent afterwards.
This is a simplified example, but in practice, additional functions and security measures are being considered to create a fully functional and secure G-Agent based governance. If you would like to divei deeper into certain aspects or have any questions, we are happy to help!