PDF, AI, and the Downstream Dreams
The first rule of the Matrix is that you are not even aware of its existence. The second rule of the Matrix is that whatever you discover after breaking out of it is another Matrix. Ultimately, the only choice we often have is to select the Matrix that offers a stronger economy, a closer-knit community, or, ideally, both.
Every engineering and manufacturing corporation is a Matrix in disguise, navigating a perpetual subconscious struggle between efficiency and innovation. Humans are simultaneously the source of innovation and its greatest enemy, because while innovation promises significant benefits to one group of employees or shareholders tomorrow, it occasionally forces another group to skip their lunch today. This is especially true for those on the shop floor, who often bear the brunt of whatever their white-collar colleagues in engineering and information technology teams have optimistically devised for them.
The educational and motivational divides are real, and mistakes are expensive. Unless the shop floor is entirely populated by robots and all engineering instructions are delivered in a fully machine-readable format, technical manuals will continue to act as a precious phone line in the Matrix, connecting different worlds. Moreover, unless engineering organizations find a much cheaper and safer way to transition from their disparate, current methods of generating technical publications to interactive formats like S1000D, PDF will remain the standard medium. This also implies that the breakdown in the digital thread will persist as a glaring chasm of inefficiency.
Inefficiency bites when you are building a detailed work plan from a 1,000-page PDF manual. It stings when you are trying to update the same plan with information more relevant than what is provided in the official documentation. This is especially true when you want to preserve precise links to the original source of truth, which includes textual descriptions, tables, diagrams, and drawings. It stabs when you attempt to quickly identify the differences introduced in a new version of the same manual. It pinches when that manual originally produced in English is misinterpreted in Malaysia, Indonesia, or Africa.
Here lies the opportunity we at Senticore are pursuing: to reconnect the digital thread between engineering, manufacturing, and MRO by enabling the seamless transformation of PDF manuals into a generic JSON format, allowing effortless manipulation and integration with downstream use cases.
Speaking of downstream, the benefits are undeniable whether these manuals seamlessly integrate with cutting-edge Manufacturing Execution Systems like iBase-t or simply become interactive for those companies still performing manufacturing, maintenance, repair, and inspections the traditional way.
Incidentally, while we initially focused on complex discrete manufacturing sectors like aerospace and defense, we discovered significant interest from the oil and gas industry, where the cost of inefficiencies in maintenance planning can be an order of magnitude higher, and our solution can improve the margins even better.
Whatever the use case, a perpetual challenge for us is deciding whether to focus more on the AI back-end or the user experience. While watching the AI magically extract various types of data from PDFs and organize them in the correct order is infinitely satisfying for the geeks, it is the Apple-inspired simplicity and elegance of the process that blue-collar workers and their managers are more likely to appreciate. The artificial intelligence components we use, such as locally deployed LLMs and convolutional neural networks, are improving every few months and we treat them as plug-and-play. This allows the design and responsibility for the overall user experience to remain firmly on our side.
Meanwhile, speaking as geeks, we’re enjoying integrating MariaDB’s Vector technology on the back-end, as it will significantly streamline the rollout of several important features like semantic search. Plus, the estimated $20-50K hardware cost for our locally deployed AI infrastructure looks appealing to our target customer base compared to the risk of trusting their intellectual property to mostly ethical AI cloud providers.
The last rule of the Matrix is that even when you become aware of the Matrix, unless you’re a liberal arts major, you still have to produce. In our dimension, despite all the innovation happening in engineering and information technology, manufacturing and subsequent maintenance operations are quickly becoming the ultimate frontier for Western companies’ competition in the face of supercharged China’s industrial powerhouse. Hence, leave your fears behind. Call Senticore to discuss your technical publications processes, and make your downstream manufacturing dreams a reality.
