At IBM, the patent attorneys aren't a part of the process for approving patents. Rather, depending on the division, there is a panel consisting of various representatives from the department. Some are engineers, usually one is an attorney. Inventors pitch their idea and the panel asks questions and decides to either ask the inventors to return with more details or additional information or may approve moving forward. Usually, the next step is to pass a search of prior art. Only then is the disclosure rated "file," where the inventors can start working with an assigned attorney to prepare the actual application. In some divisions, the panel may be just a few people, but always predominantly from the scientist/engineering side, not the legal. Legal is usually just one representative, providing guidance so no time and money is wasted on ideas that fail basic patentability criteria. A key consideration through all this is how important the idea is to IBM business. Another consideration is how discoverable the idea is. If it's too hard to detect whether or not someone is violating the patent, it's not worth going through the process. Ideas that relate to an upcoming product tend to be prioritized over those that are just potentially connected to product. This is based on going through this process a few dozen times when I used to work there.