This is the second part of a blog post, where we demonstrate 10 real world examples. Showing how we have incorporated AI into our every day Sitecore workflows.
6. Data Analysis
Generative AI models excel at analysing large, complex datasets. They can pull out key summaries of what matters most to users.
The Search Report Module
In the video below, we look at our Search Report Module. This module tracks all searches on our site. More importantly, it tracks subsequent engagement metrics such as time spent on a page and number of visits. It links this data back to the keyword used in the original search.
AI-Generated Insights
We use a GenAI model to analyse the search data for each page. It produces a short 50-100 word insight for the user, highlighting any key trends or results. If the user wants more detail, we perform a deeper analysis.
The Takeaway
By collecting and preparing data in a tailored way, we can use GenAI models to make analytics more accessible to more users.
7. Exception Analysis
Exception analysis is another example of using GenAI for data analysis. It delivers powerful results when monitoring a web application.
The Application Insights Module
The example in this video shows our Application Insights module. This module periodically calls the Application Insights API for each of our Sitecore and XConnect roles. It stores data about each exception in a database and tracks their occurrence over time.
AI-Powered Analysis
We added AI to the module, which we believe brings great value. We can now package up all exception data over a given time period and send it to a GenAI model. We ask it to identify trends in the data.
The model provides a detailed analysis. It highlights new exceptions, groups together those that are likely linked, and points out which need immediate attention.
Follow-Up Questions
We can ask follow-up questions of the dataset. We might ask it to focus only on certain types of exceptions, such as "focus only on Solr related" or "focus only on memory related".
Automated Alerts
We built in an AI Alert feature. It automatically analyses the last hour of data and compares it against the previous 48 hours. It identifies anomalies and sends out an email to highlight issues.
8. Exception Resolution
In this example, we look again at the Application Insights module. This time, we take it one step further. We ask AI to help us analyse and resolve exceptions.
Grouping Stack Traces
When a user selects an exception inside the module, it collates all stack traces for that exception over the given time period. This groups exceptions that share the same stack trace. Often an exception with a high count may actually be the same stack trace repeated once or twice.
Analysing Custom Code
We use AI to scan the stack trace and look for any instances of custom code. If found, we call the code repo and pull the associated code. The AI then analyses both the exception and the code together. It provides suggestions on how we might fix the issue.
Taking Action
With the suggested fix, we have two options. We can save the information and details as a backlog item. Alternatively, we can submit the code change to the repo as a PR for someone to review and test.
9. Component Documenation
As Sitecore developers, we constantly build new components for the business. Ideally, we would provide clear instructions with every component. These would detail how we built it, the different input options, available datasource templates, and any key business rules.
However, we often simply don't have time. We get asked to move onto something else. In the next example, we show a tool that aims to pick up where we left off.
Exploring Component Code
We use a GenAI model in conjunction with the Azure DevOps API to explore the code behind our components.
How It Works
When launched, the tool checks which components are referenced in the presentation layer. It then calls the DevOps API to find the relevant controller action, viewmodel, view, and any other immediate services.
Generated Documentation
With that collated information, the tool creates a detailed set of instructions. These highlight the purpose of the component, the types of input it accepts, available parameters, and any constraints the user needs to be aware of.
Video showcasing this in operation to be follow.
10. Component Interogator
This next example is still a work in progress at this point in time. I wanted to add it here, to highlight the idea more than anything else... and also I really wanted 10 examples!
The idea for this tool, came from all the times I get contacted by people in the business asking questions like... why is this component not showing in this particular way. This often involves a component with complex business rules, that dictate what gets displayed based on a variety of factors. Often including data from one or more external services.
For example, a course page that takes fees from one location, modules from another, intakes from another...
The plan would be for the new tool to simulate calls to each of the different external data sources. Collate the viewmodel. Review the code for the component (much the same as the documentation tool). Then allow business users to ask their questions.
It might then answer... the reason the course fee is not showing in that box, is because the external service only returns a value if X setting has been applied. Our recommendation would be to contact xxxx to investiagate this data issue.