Adobe Experience Platform Agents for improved marketing efficiency and better business outcomes

Marketing and customer experience teams are under immense pressure to deliver faster, smarter, and with greater context. Enter Agentic AI, the breakthrough approach enabling teams to meet these demands.

This exclusive session features an insightful Q&A session with Anjul Bhambhri, Senior Vice President of Adobe Experience Cloud and Gina Casagrande, Director of Experience Cloud Evangelism. Discover Adobe’s journey and bold vision to provide innovative Agentic AI capabilities that empower marketing and creative practitioners to automate complex steps, accelerate decision-making, and expand their teams’ capacity without compromising control or oversight. Be inspired by early customer adoption success stories and learn how Adobe supports your onboarding and adoption process, giving you the confidence to embark on your own Agentic AI journey.

To continue the discussion about this show/topic, please visit the Experience League community discussion.

Questions from the show

Below are some of the questions asked during the show.

If a company could start today to begin helping their organization find meaningful ways to implement agentic AI into their framework, who would need to be included and where could they start?

The most important thing is to get started with practical use cases! Agentic capabilities enforce permissions and data handling across underlying systems, giving IT and security teams confidence without compromising user experience. Adobe provides role- and task-based sample prompts to help teams adopt and scale agents faster.

So everyone that has AEP has access to the agent, or is there a rider to sign with a license?

Yes, you can get access if you are licensed for AEM as a Cloud Service, Real-Time CDP, Journey Optimizer, Customer Journey Analytics (and soon Workfront). If you don’t have it yet, please request it through your Adobe contact. A GenAI rider is not required for the try-buy program.

Are the agents able to build out the underlying data like schemas, datasets, etc.?

Yes! Agents are evolving to help you understand and create these artifacts. We will soon launch a Data Engineering Agent that can guide data engineers and architects to more easily build schemas, manage and run SQL jobs, all using natural-language prompts.

Does Adobe recommend a specific data structure in order for AI agents to work effectively?

AI Agents work across a variety of data. That being said, if your audiences/journeys have descriptive names and your XDM fields include extra descriptions, it can help drive better responses.

Assume brand guidelines and copy guidelines are loaded so the work is compliant. Does Adobe’s brand team have solid art-direction/copy-direction skills, or is creative looped in?

Please stay tuned for Adobe Summit, where you will hear about ensuring created assets are brand-compliant!

Can I use agents to explain results in CJA? And can I export these results to a dataset?

Data Insights Agent can help users explain results in CJA to answer the “why” question. It outputs responses using data viz and text summarizing the insights. Datasets in AEP store a more comprehensive view of the data.

We need to understand: is redesigning required to enable AI Agent capabilities? What’s Adobe’s approach to ensure new designs remain compatible with future updates?

Redesigning datasets is not needed for AI Agents. There are several tuning parameters as well as provisions for providing additional context so that AI Agents can navigate query data.

How do the agentic AI capabilities help with analyzing large data sets in CJA when compared to what Gemini does for GA4?

Adobe’s proprietary grounding models are layered on frontier LLMs for accuracy and trust.

We are in the process of standing up our Adobe omnichannel tech stack. When does it make sense to adopt GenAI?

As Anjul mentioned, as early as possible! You can learn about product features and best practices, and as you bring the data in, other agents will start to help make sense of it as well.

Would we have GPT models in the agents as Anjul mentioned?

Please view this documentation: https://experienceleague.adobe.com/en/docs/platform-learn/tutorial-one-adobe/agents/agents1/ex2

Working with the AEM MCP server is compounding the value of these agents. Are MCP server community options that offer specific coverage (like OSGi config management or log analysis) an option too?

Adobe is exploring agent skills and a local MCP server for code generation and debugging in an IDE. We’re interested in your use cases and feedback, which you can send to aem-devagent@adobe.com.

Please could you demo how to use the agents from the resources you just mentioned, Anjul?

Please view this documentation: https://experienceleague.adobe.com/en/docs/platform-learn/tutorial-one-adobe/agents/agents1/ex2

Can I use agents in CJA to explain offer results, for example?

Data Insights Agent can help users explain results in CJA to answer the “why” question. It outputs responses using data viz and text summarizing the insights. Datasets in AEP store a more comprehensive view of the data, so we do not support exporting Data Insights Agent results at the moment.

Can you actually show us in another episode how you would integrate with ChatGPT, Perplexity, Claude, Gemini, etc.?

Please view this documentation: https://experienceleague.adobe.com/en/docs/platform-learn/tutorial-one-adobe/agents/agents1/ex2

Our 4 year old backend datasets follow clean formatting standards and support BAU operations effectively, but don’t work with the AI Agent. What should we do? Do we need to redesign our datasets?

Redesigning datasets is not needed for AI Agents. There are several tuning parameters as well as provisions for providing additional context so that AI Agents can navigate query data