We just launched the next generation of xAssist in beta that takes virtual assistance to the next level. xAssist 2.0 utilises advanced domain language models (DLMs) that are proprietary to LexX, combining task execution, reporting and multilingual language support yet offering humans unparalleled control, honouring enterprise data/domain boundaries and mitigating for hallucinations. So here are the benefits of xAssist2.0 over the previous version.
| xAssist 1.0 | xAssist 2.0 | xAssist 2.0 Benefits |
| Utilises responsive AI i.e looks for exact or similar phrases and ranks them by relevance | Utilises Generative AI i.e generates most logical response closest to exact/similar results | Has the ability to provide logical responses even when there are gaps in data sets |
| Query by Alarm or Fault code or Observed phenomenon e.g How do I fix an oil spill from the gear box casing. | Granular queries are possible e.g what types of wrenches are needed to overhaul a specific gearbox. Results references all data sources for deep dives. | The system has the incredible ability to look for all the wrenches appearing in all the data sources but filtered for gearbox type in a single view |
| Results are ranked based on relevance. User still needs to ascertain what the most useful result is by reading through the document or section of document returned. | User is offered a result the system thinks is the most logical and consistent but the system can be challenged as you would do a human by saying “that is not the result I was looking for”. | Users can asks the system to clarify, substantiate, augment and improve the result suggested by the system until they get to the useful response |
| Does not support language translation on-the-go. Users need to use external tools to translate results | Accepts queries and generates results in over 100 languages without losing context in the workflow | Has the incredible ability to support a multi-lingual workforce without interrupting the experience |
| Heavily relies on steps and workflows embedded in document(s) to offer steps by step guides | Generates the steps from available documentation and orders them logically for users to follow | Overcomes challenges in the ingested data not having specific documentation or instructions |
| A report can only be generated from the logs of system interaction as in classical software systems | You can ask the system to generate the path/steps/actions taken any time in the workflow without interrupting flow | Saves enormous amount of time in generating reports for users. Users can do so on the fly and in a jiffy. |
| The only input the system utilises from the user is the query, including the phrasing of the query to determine intent | The system can be configured to ask users to confirm if it is on the right track to ensure it stays within context and intent. | Including human input along the way prevents DLMs from hallucinations that LLMs are notoriously known for |
| Responsive AI relies only on ingested data and it is stonewalled from interacting outside the network. Therefore the risk of enterprise data leaking to the internet is zero. | Data leakage is prevented by ensuring that the only interaction enterprise data has is with our proprietary domain language models (DLMs) which do not interact externally. | xAssist2.0 is designed uniquely to prevent enterprise data leaks. LLMs are used to train our DLMs We prevent enterprise data from interacting with LLMs directly by design |
Experience the new way to troubleshoot with xAssist 2.0 Beta. xAssist’s Gen AI capability radically improves the way engineers, technicians, schedulers and planners undertake maintenance of their critical infrastructural assets. The capabilities are real, so are the benefits. See an aviation example for yourself. Leave a comment below.
