#### There are 3 types of connections in **Sitecore CDP**: 1. ***Data System*** - can be used in Decision models as external source of data. 2. ***AI*** - can be used in Decision models as a logic block that performs any calculations on external service side. 3. ***Destination*** - can be used in Full stack experiences to send data to external services. ![Sitecore CDP - Connection types](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/connection_types.jpg) In this article I`ll show how to create AI connection to **Azure Cognitive Face API** service and use it decision modeling to provide personalized offers to customers. For my demo I`ll use customer photo to predict his age and gender and provide personalized offer based on this information. *If you don`t have Azure account, scroll to the bottom of the article where I share links to my own Face API implementation that you can use instead of Azure Face API.* ####   #### How to get access to Azure Face APIs? You need to login to your Azure account, find Face APIs in Azure services and create new one: ![Sitecore CDP AI - Azure Face API](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/add_face_api.jpg) Free **Face APIs** service has some limitations like 20 requests per minute, so try to make requests in reasonable places *(user uploads/changes photo)*: ![Sitecore CDP AI - Azure limits](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/azure_limits.jpg) When **Face APIs** service is created, navigate to **Keys and Endpoint** settings and copy one of the keys: ![Sitecore CDP AI - Face API keys](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/azure_keys.jpg) #### How to setup AI connection in Sitecore CDP? Login to your Sitecore CDP application account, Connections -> Add Connection-> AI. Enter Name and Description of your connection, choose None on Athentification tab and configure request parameters: - **Method**: POST - **URI**: [copy url](https://westeurope.api.cognitive.microsoft.com/face/v1.0/detect?returnFaceId=true&returnFaceLandmarks=false&returnFaceAttributes=age,gender&recognitionModel=recognition_04&returnRecognitionModel=false&detectionModel=detection_01&faceIdTimeToLive=86400) - **Add header**: Ocp-Apim-Subscription-Key: your\_Azure\_KEY1\_or\_KEY2 - **Request**: ``` { "url": "${url}" } ``` *${url}* variable will be subsctituted in Decision model with url of user photo. For **URI** endpoint chose closest one for your location from [Face API documentation](https://westus.dev.cognitive.microsoft.com/docs/services/563879b61984550e40cbbe8d/operations/563879b61984550f30395236). ![Sitecore CDP AI - Azure Face API request configuration](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/request_configuration.jpg) Click **Test Now** button, insert any url with person photo and click **Run Test** button. If connection is OK, you will see response from Azure: ``` "response": { "body": { "result": [ { "faceId": "65d8afbb-96ab-477c-8100-a61a3c54235c", "faceRectangle": { "top": 100, "left": 140, "width": 196, "height": 196 }, "faceAttributes": { "gender": "female", "age": 23 } } ] } ``` Once connection is ready, it is time to modeling. Create new Decision model and populate it with blocks like this *(**MS Face API** block is our created AI connection)*: ![Sitecore CDP AI - Decision model](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/decision_model.jpg) For my demo I store customer photo in profile extended properties (see [this article](https://www.brimit.com/blog/cdp-3-how-to-pass-extended-data-to-sitecore-cdp) if you don`t know how to do it). In **Extract photo** block I extract user photo url from profile extented properties and pass it to recently created MS Face API connection: ``` (function () { for (var i = 0; i < guest.dataExtensions.length; i++) { if (guest.dataExtensions[i].key === 'logo' ) { var photo = guest.dataExtensions[i].values; print(photo); return photo.url; } } throw new Error('Stop execution'); })(); ``` Output for this block is: **url** - string. ![Sitecore CDP AI - Extract user photo](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/bloak_1.jpg) In **Artificial Intelligence** block I just set **url** parameter from previous block as an input: ![Sitecore CDP AI - Artificial Intellicense](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/block2.jpg) In **Parse AI result** block I parse AI response and provide output data in format *{age: 0, gender: 'male'}* to decision block: ![Sitecore CDP AI - Parse AI results](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/block3.jpg) Before continue with decision we need to create Offers for our future experience. Navigate to **Library -> Offer Templates** and create new one: ![](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/offer_template.jpg) I will use this attributes to substitute offer values in my web experience *(corner popup window)*. Next, navigate to **Decisioning -> Offers** and create offers based on your offer template *(as much as you need)*. In my example I create 2 offers: one for women with age in range 25-35, the second one for retiree customers. Example of offer: ![Sitecore CDP AI - Offer](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/offer.jpg) When offers are ready it is time to come back to our decision model and configure last **Suggest Offer** block: ![Sitecore CDP AI - Decision model block](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/block4.jpg) **Column settings**: ***Age***. **Format**: Expression, **Column Input**: result.age, Type: integer. ***Gender***. **Format**: Expression, **Column Input**: result.gender, Type: string. ***Offer***. **Format**: Offer, **Choose Offer Template**: your recently created offer template. Configure rules logic that your need for your offers. Decision model is ready and you can test it with **Test Canvas** tool: ![Sitecore CDP AI - Test decision model](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/test_decision.jpg) The last thing to do is create Web experience. Navigate to **Experiences -> Web** and create new one. In *Decisioning* section link to created decision model and setup other sections how you need. In *Content* section choose predefined template **"Corner Popup"** and replace **Image URL**, **Title**, **Description**, **CTA destination URL** parameters with offer template variables: ![Sitecore CDp AI - Web experience configuration](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/popup_substitution.jpg) Go to "Advanced Edit" and configure variables: #### API tab *(Freemarker)*: ``` { <#if (offers)??> "decisionOffers": ${toJSON(offers)}, } ``` #### HTML tab: ``` {{#decisionOffers}} {{#attributes}}

[[Title | string | {{title}} | {required:true, group: Title, order: 1}]]

[[Description | text | {{description}} | {required:true, group: Description, order: 1}]]

[[Dismiss Text | string | No Thanks | {required:true, group: Secondary Button, order: 1}]] [[CTA Text | string | Yes Please | {required:true, group: CTA Button, order: 1}]]
{{/attributes}} {{/decisionOffers}} ``` Demo is ready! Start Web experience, login to website and populate guest photo - you will see personalized offer if user matched decision criteria: ![Sitecore CDP AI - Demo](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/demo.jpg) If you don`t see popup-offer you can debug your experience. Click on **Preview** button and populate inputs with your demo page and context user: ![Sitecore CDP AI - Debug](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/debug_1.jpg) In **QA Tool** you will see detailed response with validators and it can be useful for troubleshooting: ![Sitecore CDP AI - Demo](https://www.brimit.com/-/media/project/brimit/blog/2021/x3mxray/4/demo2.jpg) #### See also: My meetup demo about Sitecore CDP and custom AI service: Source code of custom Age/Gender ML service is available on [my Github](https://github.com/x3mxray/Sitecore.CDP.ML). Here goes the table of contents for my Sitecore CDP blog series: - [Introduction](https://www.brimit.com/blog/cdp-0-intruduction) - [How to connect Sitecore CDP to website](https://www.brimit.com/blog/cdp-1-how-to-connect-to-website) - [Debugging tools](https://www.brimit.com/blog/cdp-2-debugging-tools) - [How to pass extended data to Sitecore CDP](https://www.brimit.com/blog/cdp-3-how-to-pass-extended-data-to-sitecore-cdp) - [How to use Data System connections (Decision model with Content Hub products)](https://www.brimit.com/blog/cdp-5-how-to-use-data-system-connections) - How to use AI connections (Web experience with Azure Cognitive Services) - [How to use Destination connections (Triggered experience webhook)](https://www.brimit.com/blog/cdp-6-how-to-use-destination-connections) - How to use External service connections (Flows with External Service) - Remarketing with Sitecore CDP and Facebook Ads ###### Author [!\[sergey-200\](https://www.brimit.com/-/jssmedia/feature/blogs/authors/sergey-200.jpg?h=197&iar=0&w=200&hash=0D636570F87F1C13C39E06EB19D9DF06) Sergey Baranov Sitecore MVP/ Senior Sitecore Developer](https://www.brimit.com/blog/author?authors=Sergey%20Baranov) #### More on Sitecore [!\[How Vercel Will Help You Save Effort When Deploying Sophisticated Sitecore Projects\](https://www.brimit.com/-/jssmedia/project/brimit/blog/2024/vercel_cover-image.png) #Guides#How-toDXPE-commerce ##### How Vercel Will Help You Save Effort When Deploying Sophisticated Sitecore Projects Optimize and accelerate the development and deployment of complex multisite Sitecore projects. Alexei Vershalovich on July 17, 2024](https://www.brimit.com/blog/how-vercel-will-help-you-save-effort-when-deploying-sophisticated-sitecore-projects) [!\[Training Up Tomorrow's Sitecore MVPs: a Mentoring Success Story\](https://www.brimit.com/-/jssmedia/project/brimit/blog/2023/sitecore-mentoring---cover-image.png) #How-toDXP ##### Training Up Tomorrow's Sitecore MVPs: a Mentoring Success Story How to participate in the Sitecore Mentor program and help younger colleagues jump-start a career in Sitecore development. Sergey Baranov on October 2, 2023](https://www.brimit.com/blog/training-up-tomorrows-sitecore-mvps) [!\[Going Headless. Part 2: When a Headless CMS Is Your Best Bet (if you have Sitecore)\](https://www.brimit.com/-/jssmedia/project/brimit/blog/2022/headless/adobestock_456986731.jpg) #How-toDXPE-commerce ##### Going Headless. Part 2: When a Headless CMS Is Your Best Bet (if you have Sitecore) Discover how a headless CMS can benefit organizations that use Sitecore. Daniil Raschupkin, Palina Trokhautsava on September 15, 2022](https://www.brimit.com/blog/going-headless-part-2-when-a-headless-cms-is-your-best-bet-if-you-have-sitecore) ![](https://bat.bing.net/action/0?ti=187017043&tm=gtm002&Ver=2&mid=02516337-e1cc-4b92-be0a-9bd81637278f&bo=2&gtm_tag_source=1&pi=0&lg=en-US&sw=800&sh=600&sc=24&nwd=1&tl=Dive%20into%20Sitecore%20CDP%20-%20How%20to%20use%20AI%20connections%20(Web%20experience%20with%20Azure%20Cognitive%20Services)&kw=sitecore%20CDP&p=https%3A%2F%2Fwww.brimit.com%2Fblog%2Fcdp-4-how-to-use-ai-connections&r=&lt=336&evt=pageLoad&sv=2&asc=D&cdb=AQAY&rn=897794)