Supplier Documentation Thanks to Artificial Intelligence

The problem of supplier documentation

Brückner specializes in film stretching machines and has been a family-owned business since its founding in 1960. The company's 2900 employees in 13 countries design, build, and manage some 1500 plants in Asia, North and South America, and Europe. And each of these 1500 plants has a unique setup and ends up having many different versions and combinations of instructions when it comes to the documentation. This is because each plant needs to have operating instructions that are specific to its components and the plant itself, along with a wide range of standard documents (such as ATEX certificates). Then there are the spare parts catalog, various translations into the relevant target languages, and, of course, the supplier documentation.

Above all, the supplier documentation presents a particular problem, simply due to the sheer quantity of documents. In addition, the supplier documents cannot be edited and are also more difficult to manage than Brückner's own documentation. Given that the plants can have a service life of over 50 years and provisions have to be made for more than 20 languages, it was clear that innovative solutions were the only way forward.

A systematic solution

Brückner addressed this situation by making an early move to invest in professional authoring tools. SCHEMA ST4 provides a powerful basis for creating the company's own documentation, Across supports terminology and translations, CatalogCreator (a Quanos product) automates the spare parts catalog, and SCHEMA CDS offers a platform for content delivery that makes all documents relating to a product available online.

But it's important to note that “provide” doesn't necessarily mean users can find and use documents. With 278,000 relevant documents, limiting search results was key. After all, who wants to carry out a search and find that a set of instructions causing a severe issue is somewhere among 3000 documents included in the search results?

On the other hand, a massive quantity of documents like this can't be managed manually using editorial analysis and tagging. The only things that can really be found automatically in the supplier documentation are the file name and path, and these include information on the content in only the rarest of cases.

Solution with (artificial) intelligence

When people are unable to manage the sheer quantity of information, it makes sense to solve the problem using computers. The idea at Brückner was to use artificial intelligence to analyze supplier documentation on an automated basis and (after suitable training) provide it with metadata and taxonomies. This metadata could then be used to improve access to the supplier documentation, for example, to limit search results in the content delivery system.

After conducting extensive research, Brückner opted to go with an AI system offered by plusmeta. The main factors behind this decision were the broad range of documents that plusmeta's AI could analyze and the seamless connection to SCHEMA ST4. In addition to the documents in the supplier documentation, this also offered scope to evaluate and process the node contents of Brückner's own documentation and the existing taxonomies.

After training the plusmeta AI system with the available data and existing taxonomies, the company was able to evaluate the material in the supplier documentation. This led to a number of surprises. For example, there were significantly more maintenance intervals for the respective plant components than had been provided for in the existing documentation. Thanks to AICube, returning the ST4 node content (that had now been enriched with the taxonomic data determined by the AI) was also an issue that could be easily solved. All that had to be done was to make sure the taxonomy of the AI was already known in the CCMS as well.

Using AI to obtain metadata and taxonomies on an automated basis proved to be an effective solution for Brückner. This approach makes it easier to find supplier documentation in the content delivery system, simplifies translation workflows, and creates the basis for variant filters. Thanks to artificial intelligence, supplier documentation is no longer a problem at Brückner.