What are AI agents and how will they impact my CMS or DXP?

As the unstoppable AI juggernaut continues to move at a remarkable pace, you will undoubtedly have heard the term “AI agent” or “agentic AI”. The latest wave of AI innovation is already here with AI agents being launched across different workplace technology products, including within leading Content Management Systems (CMSs) and Digital Experience Platforms (DXPs). AI agents have even reached the world of TV advertising with Idris Elba mentioning them in the ServiceNow platform.

But what exactly are AI agents? What is agentic AI? And how will AI agents impact the work of digital marketing teams? In this article, we’re going to explore what AI agents are, what they mean for your CMS or DXP and what the future might hold.

What are AI agents?

Not everybody knows what AI agents actually are, so here are some useful definitions.

IBM describes an AI agent as “a system or program that is capable of autonomously performing tasks on behalf of a user or another system by designing its workflow and utilising available tools.”

McKinsey defines AI agents as “a software component that has the agency to act on behalf of a user or a system to perform tasks.”

SAP refers to agents as “artificial intelligence-applications that make decisions and perform tasks independently with minimal human oversight.”

What is agentic AI?

Meanwhile “agentic AI” is a linked term and concept to AI agents but is more to do with the approach, framework and coordination of the individual AI agents within an individual Sitecore has a nice explanation of the difference between AI agents and agentic AI:

“Think of agentic AI as the conductor in an orchestra and AI agents as the musicians. Agentic AI manages the AI agents in a coordinated way, ensuring they play their part or complete their tasks seamlessly.”

What are the characteristics of AI agents?

Essentially AI agents have a number of characteristics:

• They are powered by Large Language Models (LLMs).
• They are autonomous and have the authority to carry out actions that require no or minimal human intervention.
• They complete specific tasks and act within workflows.
• They are “intelligent” in the sense they can make decisions that define their actions to complete that task.
• Agents act on behalf of users but can also act on behalf of a “system” so potentially an agent acting on behalf of a user, could then instruct another agent in order to complete the original task, so different agents can then work together to achieve more complex outcomes.

What are the benefits of AI agents?

AI agents have been around a while but tended to be focused on a developer audience. However, their ability to be deployed at scale within commercial products that can be activated by business users is relatively new.

The benefits of AI agents are clear in that they will allow users to complete specific and more complex tasks that are geared towards popular use cases. In some cases, an agent might be able to interact with another platform too through what is effectively an integration.

For resource-challenged digital marketing and website teams this means they can:

• achieve more tasks with fewer resources.
• get more value as AI agents embedded into products are geared towards common use cases and completing specific tasks.
• unlock both the potential of AI and the fuller range of DXP features such as the ability to introduce personalisation.
• enable business users to do more without the help of their IT colleagues.

Are there any downsides to AI agents?

While we think AI agents are an exciting new phase for AI, there are some considerations to bear in mind.

Firstly, the marketing machine is in full swing so you can expect “AI agents” to feature heavily in promotional material and product descriptions. Functionality that was previously described as an app, a feature, a workflow, a copilot or an automation many now be described as an AI agent. Some will be true AI agents, but others might stretch the definition a little, so it will be hard to distinguish what is an agent and what is not.

Secondly, there is likely to be a likely learning curve with using, managing and orchestrating AI agents. Similar to getting used to effective prompting with generative AI, using a series of semi-autonomous agents to complement your work will take a little getting used to, may need a change in how you carry out core activities, and will require a mindset shift

Thirdly, AI agents are going to be coming at us thick and fast over the coming months and will be evolving at breakneck speed. Like all things AI, it is going to be difficult to keep up.

Finally, on a wider scale the introduction of AI agents may lead to job losses or make the business case for additional headcount that much harder. There is a lot of debate about this, but this is potentially automation at scale brought to knowledge work and it seems inevitable that some roles will change. The common presentation of AI agents as additional members of your “team” or as part of your “workforce” seems to be actively acknowledging this direction of travel.

Are AI agents coming to CMSs and DXPs?

AI agents have already launched in leading CMSs and DXPs. The extent to which these are currently “true” agents, or a rebranding of existing AI-powered features is hard to ascertain – and perhaps it does not really matter at the moment – but the likely evolution will mean agents become more sophisticated and more autonomous. Optimizely Opal AI launched recently and promises a team of “AI specialists” that act as “your infinite workforce” that are brand-aware and are embedded within existing workflows.

Within the offering there are currently six agents advertised:

• Keyword Researcher: Get recommendations based on your target keywords.
• Video Transcriber: Transcribe videos with text, subtitles and translations.
• Campaign Manager: Build a brief and launch a set of campaign activities.
• Market Researcher: Get deep market insights to inform your strategy.
• Personalization: Design and measure high-opportunity segments.
• Industry Marketer: Tailor content with industry insights and trends.

Within Optimizely, you can initiative these agents using prompts but perhaps more excitingly, you can:

• Create custom (“Specialised”) purpose-built agents incorporating inferences, prompt templates and domain-specific knowledge.
• Create potentially very sophisticated workflows by connecting agents together.

Sitecore’s main AI offering – Sitecore Stream – also incorporates AI agents. However, agents are less prevalent in the positioning for the product, which has tended for focus on “Copilots”. However, Sitecore Stream says it offers:

• “Copilots and agents” that “help you think and act faster across critical marketing tasks so you can strategize, execute and scale without the busy work”
• “Agentic workflows” that “harness the power of AI automation and human creativity to optimize the entire content lifecycle.”

It seems likely we will get more focus on agents with Sitecore Stream in the coming months. It’s also exciting to note that a version of Sitecore Stream will be activated for all Sitecore customers over the summer with options also to subscribe to the premium version of Stream.

Looking to the future

Just as it is impossible to fully predict where we are going with AI overall, the same applies for agentic AI. The products are evolving rapidly, and agents have the potential to significantly change the way marketing teams operate. At the same time, the hype might turn out to be just that, and the impact may be less profound or slower. Whatever happens, we’ll try and cover it on the blog. If you’d like to discuss AI agents, Sitecore Stream or Optimizely Opal, then get in touch!