this video shows how to use natural language understanding workbench to improve your user's search experience in AI search an nlu model Maps users phrases or utterances to specific intents the model also identifies details of an utterance or entities like a specific software app the utterance how do I access Miro might nap to the intent find catalog item and the entity mirror AI search uses those intents and entities to display genius results in response to user searches creating an nlu model for AI search in nlu workbench has three phases first you build and train your model to recognize the utterances you expect to get from users next you test and refine it when it's ready you publish to start using it with AI search and finally umap specific intents to specific genius results to display for users we'll start on the nlau workbench homepage on the AI search tab there are a few different ways to build your model you can start with the pre-built model for AI search and tailor the intents and utterances to your business you can also create a model by importing data from a CSV file or start from scratch for our demo we'll open this existing model and start the first phase build and train here are the intents in this model let's look at catalog fine the intent to find a catalog item we create a set of utterances for each intent phrases that users might actually enter in AI search here are the utterances for catalog find we want at least 15 distinct utterances for each intent while we're doing this we can try specific utterances to see how our model handles them as we update the data for our model we train the model to incorporate our changes we also create a set of test utterances to help us see how our model is performing later on when your model is in use you can come back to the build and train phase to refine the model or add new utterances when we've built and trained our model we go on to the next phase test and publish we test our model using our test set the test score shows how well the model handles the test utterances we'll use the test results to refine our model and then train and test it again train test repeat when we're satisfied with the test score we publish the model to make it available to AI search finally we go to genius results to map each intent in our model to a specific genius result to do that we choose the genius result configuration we want to use our nlu model for we set the trigger condition to nlu and select the active option to make the genius result available to users then down on the nlu model tab we'll open our model and add the corresponding nlu intent to the genius result when that intent is detected in a user's search AI search will display this genius result now we're ready to start using AI search with our nlu model that is how natural language understanding workbench helps you build test and publish your nlu model for AI search to improve your user's search experience for more information see our product documentation or knowledge base or ask a question in the servicenow community foreign foreign