There are many integration opportunities between Sitecore’s composable solutions and artificial intelligence. Sitecore has begun to integrate this technology into their product line and will continue to enhance their SaaS solutions with these capabilities, having committed to an open approach that allows you to bring whatever AI models and technologies to enable your use cases.
When we look across all their composable solutions we see many opportunities to drive innovation with artificial intelligence. When you think about creating an AI roadmap, here are some common areas to consider:
Content Creation & Revision
Using large language models like ChatGPT to generate headlines, write articles and revise content already written is a common use case for AI. Integrating this with Sitecore is a natural extension and an area we think Sitecore will natively support in XM Cloud in the near term. (They are already supporting this in Content Hub One and Sitecore Send).
Although Perficient has experimented with integrating prompts into XM Cloud, we recommend waiting for official support instead of building a custom solution.
Content Classification
A strong content taxonomy can really help with discovery and search relevancy but also with analytics and gathering insights on what type of content is resonating with your audiences. Actually classifying your content however can be tedious and error prone when left to content authors to manually configure. Leveraging generative AI to classify content, identifying tags that are appropriate for that content can make it much easier to ensure that taxonomy is classified and accurate.
These types of classifications can not only be applied to textual content but images and videos. AI can classify objects in these assets, determine colors and even analyze the sentiment of the content itself. For videos it can identify the timestamps where objects appear and even create transcriptions of audio found in the video.
Content Hub actually can facilitate much of the image and video analysis leveraging Azure cognitive services. For XM Cloud, the content classification needs will be highly customized to your content taxonomy. To support that a custom integration that leverages Sitecore’s workflow capabilities, webhooks and an AI service would be our typical approach.
Language Translation
Creating localized versions of content can be a strong candidate for generative AI with the right safeguards in place, particularly for industries like healthcare where accuracy is paramount. While there are many third-party solutions that will extract content from your CMS and farm it out for both automated and manual translations and review, you should consider operationally how new content will be translated after the initial localization of the site is completed.
This too can be integrated with Workflow, leveraging a custom webhook to manage leveraging AI services to get the translation and add or update the language version in Sitecore as needed. While these custom webhooks could be implemented on any platform you like, a low code integration platform like Sitecore Connect could be used to easily connect services together to facilitate the integration. Alternatively, services like Azure functions can provide a low maintenance custom approach for hard coding integration business logic.
SEO Optimization
Analyzing your content from an SEO perspective using generative AI can be a helpful exercise to optimize your organic traffic. Using services like ChatGPT you can analyze existing content for keywords or even generate revisions of that content that better align to keywords that you are targeting.
To support these use cases, we’d recommend including targeted keywords as part of your content taxonomy and running batch reports that analyze and make recommendations for improving results on a periodic basis. You can also consider a workflow action that triggers an analysis and report being generated.
Media Creation & Optimization
Tools like DALL-E and Midjourney are well known for their ability to generate images based on prompts. Video generation has also been improving, with tools like Sora demonstrating impressive results. The natural integration approach is to leverage their API’s to integrate directly into your CMS or DAM tool, and there are plenty of examples of doing so with Sitecore.
And while that does have obvious benefits, there is additional value in leveraging this technology with Content Hub and its ability to generate renditions automatically. While normally this features is used to automatically crop and scale an image to meet common formats, it could be extended to call an AI service capable of making revisions to an existing image, and automatically creating variations that align to themes that make sense for your content strategy.
Personalized Content
One of the barriers to personalization is actually writing the content that will actually improve the experiences you are trying to create. This becomes particularly laborious when you have many segments and many opportunities for personalization across your website and other channels like email and social media.
Leveraging large language models to help generate that content reduces the effort considerably, allowing you to focus more on where and how to apply the strategy than on crafting the content to plug in. While copy paste is usually the easiest way to leverage this, you can integrate your LLM with Sitecore Personalize to generate content on the fly as needed. If you rather review and control the content before it is used in your experiences or experiments, you could source your personalized content from Content Hub One as demonstrated in one of our blog posts. Since Content Hub One already has integrated Generative AI, this approach could be used to quickly create content variations that could be consumed by a Sitecore Personalize decision model.
Customer Segmentation
AI can be useful in recognizing patterns in user data to identify segments that exhibit similar behaviors, preferences, and characteristics. These insights enable marketers to create more targeted and personalized marketing campaigns, messaging, and product recommendations. AI can uncover unexpected relationships between seemingly unrelated variables, revealing new customer segments based on nuanced behavior or preferences. This capability may uncover hidden opportunities and target niche audiences that traditional segmentation methods might miss.
While Sitecore CDP allows marketers to define segments using simple rules or even an advanced SQL editor, if you want to leverage the power of AI to create segments, you will need to rely on Sitecore CDP’s batch API’s to export, process and import the data with additional extended attributes. You could leverage Sitecore Personalize decision models by invoking them from that batch process to provide decisioning capabilities to that process.
Next Best Action
When you are ready to scale your personalization strategy beyond “if this then show that,” a Next Best Action approach can help you think through the possible offers and when and how they can best be presented to your visitors to help them advance through their customer journey’s. I’ve written about how to conceptualize this approach in a recent article in CMSWire on mastering your personalization strategy.
In the end you need to build out logic to determine the best offer to present to a visitor at a given time and touchpoint. While this can be a hard coded algorithm, this can also be an opportunity to leverage AI to determine the best offer for your visitor. If you have enough data about visitors and their behavior, this can lend itself well to an AI based propensity model. This can analyze each of the possible offers and determine their probability of conversion. Taking that into account with the value of the different offers and related conversions can ensure you optimize for not just conversion but outcome.
Search Relevancy
Website Search is a perfect use case for leveraging predictive AI to help improve relevancy and get your visitors to the products or content they need to continue in their customer journey faster. Composable Search solutions like Coveo and Sitecore are making it easier to create personalized search experiences using predictive AI. By looking at aggregate search behavior, navigation, and conversion data it makes it possible to not only move the best results to the top of the search results, but it makes it possible to find “look alike” audiences and move the best results for that specific visitor to the top, creating a truly personalized search experience.
Search Summarization
Another way to leverage AI in search is combining the relevancy boosts that predictive AI brings with the summarization capabilities of generative AI by taking those top results and using a large language model to create an easier to understand summary of those results. This is how Search Engines like Google have embedded generative AI into their search results. Using a composable search solution and a language model you can do the same for your site search.
Development
There are plenty of opportunities to gain efficiency by leveraging AI in your development process. Tools like “Co-Pilot” are just beginning to transform the way we work, from creating tickets, defining acceptance criteria, scaffolding components, fixing defects to deploying and performing QA testing, we’re only beginning to see the value that AI can bring to our development processes.
At Perficient, we’re looking into this closely, evaluating tools that can help us migrate MVC solutions to Next.js and React. We’ve looked at tools from our Partners like V0 from Vercel which can generate React components based on a textual prompt. We’ve even built some proof of concepts to think through how it could work.
Perficient can Help
Perficient has a dedicated AI practice that specializes in delivering innovative AI solutions. And as a platinum Sitecore partner, we have expertise across all of Sitecore’s composable solutions, as well as many complimentary platforms like Coveo, Microsoft and Salesforce, not to mention deep relationships with AI platforms like Writer. If you are interested in creating your own composable AI roadmap, we’d love to help. You can reach out to me on LinkedIn, Twitter or fill out our contact form.
