Creating snappy and eye-catching email headers for marketing campaigns can often be difficult. In today’s digital landscape, customers are inundated with offers and announcements, which may feel overwhelming. Standing out from the crowd is essential, and can mean the difference between an email being opened and read or ignored altogether. Starting in the September 2024 CI360 release, there is a new feature available to customers: Gen AI email headers.   How it works   When a user creates an email header, there is now a button available which can suggest Gen AI alternatives. When this is selected, a call is made behind the scenes to the AWS Bedrock API, which is used to run the Anthropic LLM and return a suitable subject header.   In this example, the user starts with their own user-written header: “Now 25% off all sunglasses”. When they click the “Create Gen AI Alternatives” button, it brings up a list of suggested options created by Gen-AI.   Select any image to see a larger version. Mobile users: To view the images, select the "Full" version at the bottom of the page.   If a user didn’t like any of them, they could select “refresh subject line options”, which would re-run the API call and bring back different results.   However, they liked the "Sunny Days Ahead..." suggestion. So it was selected and now becomes the new header. There is also a magic wand symbol next to it to highlight that it has been Gen-AI generated.     Note that, as well as being flagged in the UI, it is also flagged in the underlying data. The field “variant_source_NM” in the “MD_Task_X_Variant” table denotes whether it is a header produced by the user or Gen AI.   A user can also provide instructions to the LLM to help guide the results. For example, to set a certain tone or be targeted to a specific audience.       In the examples above, the user has added the free text instructions in orange. This has then been factored into what the LLM has generated.   Once selected, the Gen AI created header is used in the same way that a human generated header would be. The remainder of the email creation process is exactly the same.   Best Practice   Getting the best results is sometimes an iterative process. The first prompt might not get the best results. If you are not happy with the first set of results, then refresh the list. The new suggestions are still based on your original subject. It is important to be as specific as possible. If there are key words or messages required in your heading, make sure they are provided in the original prompt. Be intentional with the tone of your original subject. Unless tone guidance has been provided, the AI-generated suggestions will match the tone in the original subject.   Merge Tags   Merge tags are supported in this feature, with the following caveats:   If you include a merge tag in the original subject, make sure that the merge tags that the Gen AI produces match your original one. You get better suggestions when the name of your merge tag describes what it represents. GenAI uses merge tags for context clues when suggesting replies. You can use a sample value instead of a merge tag in the original subject. After you select the generated subject to use, you can replace the sample value with the appropriate merge tag. Conditional merge tags are supported. For best results, do not use more than one conditional merge tag in a subject. Ensure that there is enough content preceding and following the merge tag for GenAI to use when figuring out how to place the merge tag in the suggestions.   Language support   The language is automatically detected based on what the user types in. The results are generated based on the language of the input header, and the results can often be quite different due to language nuances and communication differences. This is completely independent of the language settings of the browser or CI360. Most Latin scripted languages are supported as well as Japanese. For example, if the same text is entered in Italian, the following results are produced:     A/B testing support   In addition to being available for standard emails, it can also be used in A/B testing of email campaigns. In this way, you can attempt to beat the computer!   User oversight   We all know that Gen AI can sometimes produce some unusual results! Because of this, a user must manually approve, modify, or reject any AI generated. Note that any modifications to a Gen AI header will result in the header being considered a user generated header rather than a Gen AI one.   Admin settings   If you don't see the Gen AI button, and definitely have the version which supports it (Sept 2024 onwards), you may need to enable it. This is not enabled by default, as the terms and conditions for use must first be accepted. This can be done under general settings.     Find more articles from SAS Global Enablement and Learning here. ... View more

This post will explain some key changes to how CI360 handles event attributes in activity maps. Event attributes from upstream events historically needed to be manually added to nodes in an activity. The trigger event with the required attributes had to be used as a targeting event before the attributes could be used . Additionally, the attributes often didn’t flow down correctly due to timing issues. The attribute values would be uploaded to the customer state vector (CSV) from the trigger event. The downstream tasks would then try and retrieve them, but the CSV may not have been refreshed in time. The next event could have triggered before the CSV had been updated, even by a matter of milliseconds and so the attributes would not be there. To get round this, wait nodes were often added, purely to allow some more time between nodes starting and finishing, and therefore increasing the chances of attributes being propagated correctly.   Select any image to see a larger version. Mobile users: To view the images, select the "Full" version at the bottom of the page.     Changes   Starting in the February 2024 release, the event attributes from trigger events will automatically be available in all downstream nodes. This addresses the two issues that has been discussed:   Usability: This change negates the need for the event to be added as a targeting event. It wasn’t very intuitive that this is what needed to be done, and therefore wasn’t a great user experience. Often, customers would miss this step and wonder why the attributes were not available. Now, they are automatically available for use. The screenshot below demonstrates this:     Timing/ Reliability: There is no longer any need to add “dummy” wait nodes to allow for any timing discrepancies. The timings of the updates to, and retrievals from the customer state vector have been synchronised. It should be noted that the attributes are available downstream, but the attribute values themselves may change as the activity progresses. It is always the most recent value that gets used. For example, imagine a situation where an “add to cart” event is the trigger event. A customer starts an activity when they add a t-shirt to their cart. The customer then adds a pair of trousers to their cart. The “product name” attribute is available for use throughout the activity, but the value itself changes from “t-shirt” to “trousers”.   Further information   There is a live demonstration of this feature on the next release of the CI360 SAS Curriculum, in the “Creating Activities with SAS CI 360” section. Specifically “the Building a Cart Abandonment Activity” demonstration.     Find more articles from SAS Global Enablement and Learning here. ... View more

Making accurate and timely data driven decisions is crucial for companies as they strive to compete in the age of AI. Additionally, it is important to be able to build, test and deploy these capabilities in a user friendly and transparent manner.   SAS Intelligent Decisioning software is a cutting-edge product that delivers all these features. The purpose of this post is to introduce a new decisioning product from SAS, SAS Decision Builder, which builds upon our partnership with Microsoft. Background Fabric is Microsoft’s new solution for data and analytics. It brings together a range of new and existing data and analytics toolsets in a unified environment. Data that is used by Fabric is available in a data lake (Microsoft OneLake). The front page can be viewed below:   Select any image to see a larger version. Mobile users: To view the images, select the "Full" version at the bottom of the page.   The main goals of Fabric are to;   Make data and analytics accessible to everyone.  By offering a combination of low/no code interfaces and coding environments, the solutions are available and usable by both business users and data scientists/engineers. Have all the data and tools in one place. All data and analytics tools are available through the Fabric portal (see below). Data can be accessed in the OneLake lakehouse in a single storage format (Parquet), either directly or via connectors. There is no need to migrate data or convert formats.   Following the partnership with Microsoft, SAS have developed SAS Decision Builder, a brand-new decisioning workload available exclusively on Fabric. It is packaged as a SaaS offering and can be purchased directly from the Azure marketplace. There is no need for a separate SAS license! About SAS Decision Builder SAS are the only Independent Software Vendor (ISV) offering decisioning capabilities on Fabric, and the SAS Decision Builder product integrates seamlessly with customer data already in the One Lake data lake environment. If your team has existing workstreams in Fabric, you can take advantage of the power of AI-driven decisioning on your Fabric workloads.   Core Capabilities SAS Decision Builder is built on the same underlying engine as SAS Intelligent Decisioning and therefore inherits much of the same functionality. Decision Builder enables customers to design and validate decision flows, manage business rules, and test and automate decisions at scale. We will now go through some of the key features in more detail:   Microsoft One Lake Data integration There is full access to all the customer data stored on OneLake. Users can simply drag and drop the data they would like to build decisions on. Data produced from decisions made in Decision Builder can also be written back to OneLake for use in downstream processes, including Power BI reporting.     Building Decisioning Rules Conditional rules can be built, tested and deployed using if/else logic. This utilises a user-friendly interface for developing and testing the rules.     Python Capabilities SAS Decision Builder fully supports the Python programming language. It is therefore possible to develop rules, models, and perform pre/post processing entirely in Python.     Model Integration It should be noted that SAS models and procedures are not the focus of Fabric, and therefore not of Decision Builder. However, it is possible to use models from Azure AI Services and call externally hosted models (including LLMs) using REST APIs. As Python is supported in Decision Builder, all Python model libraries are available for use.   Decision Flows Rules, models, and data can then be combined to build end-to-end decisioning processes. This can include splits, segmentations, and A/B testing.     Fabric Monitoring Hub Integration Deployed decisions can be monitored using the Fabric monitoring hub. This allows full oversight and transparency of what decisions are being run and when, as well as any jobs that may have failed.   Next Steps Customers can sign up for a private preview of SAS Decision Builder, available since June 2024, with an expected general availability release in Q1 2025.     Find more articles from SAS Global Enablement and Learning here. ... View more