Cloud Pak for Integration

[Transcript of the video https://www.youtube.com/watch?v=_cjSLVU07jo ]

Hi, I’m, Krithika Prakash, AI integration architect in the IBM Automation Development organization.   Today, I would like to take you through two of the exciting AI features that my team and I have been working on this year. The AI / NLP features are now released and available as part of the CP4I 2021.4.x release.  Let's look at these two features briefly.

I assume you are familiar with the CP4I platform,  AppConnect flows, and Automation Assets. If not, please get familiarized at a high level with what these are before proceeding with this video. https://www.ibm.com/docs/en/app-connect/containers_cd?topic=designer-tutorials 

I’m here on the CP4I Platform Navigator. I’m logged into AppConnect Designer. This tile here, where you see the light bulb icon, is where we have infused AI. Previously this tile was “Create from an Asset” which lets you create an AppConnect flow from an existing set of templates available in the Automation Assets - which is another component on the CP4I platform. When you click on this tile, it displays the list of all the templates that are available.  Previously, this list used to be empty. Unless you have uploaded a few assets, you won’t see anything here. But as part of an earlier release this year, we started auto-populating the Automation Assets with a set of templates from our public GitHub repository which has 40+ CP4I specific templates as of Dec 2021.

The AI infusion is implemented in this search bar. What used to be just a keyword match search, is now made to process natural language. You can describe your integration scenario in the English natural language and not just as keywords.

There are two separate AI services and two separate AI features here: Let’s type in an utterance and look at both the features.  If you don't know what to type, you can also click on this bulb icon that takes you to the Knowledge center where we have documented sample utterances. Here my utterance is “ When a new contact is created in Salesforce, create a contact in Insightly”.  See, how I'm describing the intent clearly. When I click "Enter", the first one on the top is the first AI service, smart recommendations, or Watson Insights.  The second  AI service is below here.  Further below is the classic table keyword search results which we have left intact.

Let’s look at the first AI service. It uses Watson NLP to find a semantic similarity between the utterance you typed and the existing template names. It returns the top 3 templates that match with a high confidence score.

The second part here, the second AI service - Which is where I really want to put the spotlight on and the main purpose of my video is also this part. You can click on the “info” icon that takes you to the knowledge center where we have documented lots of information related to this AI feature, guidelines, and a few sample utterances that you can try out.
This feature right here appears small, but the technology behind this is extremely powerful. It is a feature developed by the IBM research team that we have productized here.  If you click on the "view flow", it shows you the flow that is generated from scratch! Yes, you heard it right, it takes the utterance, parses the abstract meaning, maps it to a knowledge graph, plans out the connectors, objects, responses, etc, and then builds the entire flow - all from scratch. 

The good thing about these AI  infusion is that once you like any of these templates - either smart recommendations or the one generated from scratch, you can click on it to view in the flow editor. Then you can process it business as usual. You can use the same set of features available in the flow editor, like for example, the mapping assist - it is another existing AI feature that lets you map fields between the connectors, and then you can edit the flows, save it as a new flow, and from the Dashboard, you can export it back to Automation Assets as a template, and so on and so forth.

To summarize, - describing your intent in English and generating a fully developed flow is exciting and is a big step for us in the direction of infusing AI. 

We are actively investing and working on it. We are looking for feedback specifically on the utterances you and your team commonly use that you would like this AI to autogenerate for you. If you have such data or if would like to collaborate directly with the product team and the IBM Research team on this AI feature, please leave a comment or contact us. Looking forward to hearing from you. Thanks for watching!