AI-driven customer experiences: How do you choose the right CMS?

The only reliably predictable thing about the future of customer experience is its unpredictability. All in all, it’s a difficult place to be if you’re trying to work out how to keep the technical platforms in your digital experience stack both today-fit and tomorrow-ready.

The capabilities that got you here – content creation, personalisation, multi-channel delivery – remain important. But AI fundamentally changes what 'content management' means. Your CMS needs to evolve from managing pages to orchestrating conversations, from serving content to enabling actions.

The hard truth is that you can’t afford to evaluate content management and delivery platforms on their current features only. However, due to the lack of industry consensus on long-term solutions, platform vendors lack clarity on what their clients are going to be asking for after the immediate rush to deliver marketer AI toolkits, which makes it hard to evaluate platform roadmaps too. 

It may be hard, but it’s going to be critical. So here’s a view, based on our current thinking about the future of customer experience that you can use to start asking the right questions when you’re looking at content management and delivery vendors: about what they have available now, and when and how they’ll deliver other features critical to the agent-to-agent-driven experience in the near future.

Internal toolkits

Platform vendors typically address their direct customers first, and so most of the major CMSs now offer some form of AI assistance with some content tasks that your internal teams can use to enhance their marketing operational effectiveness and efficiency. These help with tasks like content creation, translation, layout management, task orchestration, data analysis, and ideation assistance. But the breadth and quality of implementation vary widely.

Ask how they deliver these features now:

· Content gap ideation and research

· Copy creation based on a provided brief

· SEO metadata generation

· Accessibility tag generation

· Automatic asset and content tagging

· Layout creation assistants

· Workflow orchestration assistants

· Integrated machine translation

· Full APIs for all management operations

· MCP or other AI agent-accessible functional interfaces

· Delegated (rather than application) agent permissions

Ask when and how they’re planning to deliver these features:

· Multi-modal prompting (image, audio and video)

· Multiple AI model support

· Agent instruction tuning

· Agent context tuning

· Agent chaining and tool usage

· Custom agent tool development

· Model tuning for brand tone of voice

· Model tuning for audience segments

· Automatic variant creation for personalisation

· Automatic targeting for personalisation

· In context analytics interrogation and recommendations

· Multi-channel campaign co-ordination

· Multi-modal prompting, including transcript analysis

· Virtual meeting attendance

· Content model generation using design files

· Content experience generation using design files

· Multi-modal content repurposing and promotion (video to written article)

1st party content functionality

Brands increasingly and urgently need to create AI-driven content interfaces and functionality that allow customers to control how they discover and engage with content incrementally, at their own pace and following their own interests.

Realistically, to deliver any reasonable level of quality of AI-driven content within your own digital experiences, that content has to be programmatically accessible in a structured way. That doesn’t mean you necessarily need a dedicated Headless CMS, but to give your AI features proper context for the tasks they need to deliver, they need to be able to select and address content with a lot more nuance than just at the webpage level – and without all the design markup designed for human readers.

Chatbots are the obvious functionality example, alongside semantic and summarised search. However, AI-driven micro-audience build and targeting features are needed to deliver on the promise AI creates to bring hyper-personalisation functionality within the reach of every brand: personalised content digests, landing pages and content discovery agents.

Note that some of these questions don’t apply to pure headless CMSs as they relate to functionality provided elsewhere in the content delivery pipeline – ignore the questions marked (*) in this situation, but start asking the relevant supplier in your experience stack.

Ask how they deliver these features now:

· Programmatic access to the content repository (ask for GraphQL rather than RESTful interfaces)

· Behavioural content recommendations *

· Semantic search (supporting customer intent understanding without excluding keyword search) *

· Search reranking *

· Audience build and content targeting

Ask when and how they’re planning to deliver these features:

· Native semantic chunking for long-form content (breaking articles into meaningful segments AI can understand and recombine)

· Tuneable prompt analysers

· Native, fine-tuneable agent completions and embeddings *

· Core repository native content graph

· Custom content graphs

· Dynamic audience context using customer data platforms

· Micro-audience persistent content caching

· Generative content quality and accuracy metrics *

1st party functional interfaces

The ability for customers to instruct AI agents to carry out actions on their behalf is in its relative infancy, especially in the consumer space: ChatGPT agent is the pioneer, released only in July 2025. But it’s the logical destination that customers will expect brand experiences to reach, that stops the abrupt lurch from full-service to self-service digital journeys just when the customer is ready to commit to something.

1st party functional toolkits will allow your customer-oriented AI agents to complete tasks on their behalf, based on their instructions, behaviour, conversational context and history with your brand: form-filling, profile management, transaction processing, order management, and more.

Fundamentally, there is little available out of the box right now in any standard platform, and your key focus will be on ensuring that your ability to build the future experiences that customers will demand within the lifetime of your platform investment will be supportable when the time comes.

None of these considerations apply to dedicated headless CMSs, but they do apply to the relevant functionality suppliers in your experience stacks, across form providers, ecommerce engines, account management platforms, and so on – you’ll want to ask those providers instead. For everyone else, ask these questions.

Ask how they deliver these features now:

· Single Sign On identities

· Delegated permissions and access management

Ask when and how they’re planning to deal with:

· Conversational data collection

· Distributed ID support

· Core function MCP servers for internal functions (e.g. basket management functions for CMSs with integral ecommerce functionality)

· Custom MCP server support

· Custom MCP client support

· Persistent agent context/state

· Prompt injection protection

3rd party agent interfaces 

When we talk about 3rd party agents, we’re talking about the AIs your customers are already talking to: ChatGPT, Gemini, Copilot. Scraping your content and rewriting it as they talk to your customers about you; or even worse, not scraping your content and talking about your competitors to them.

And in the very near future, they won’t be limited to talking. They’ll be acting as full personal assistants to your customers, with access to their calendars, email, finances: booking, buying and managing your products on their behalf, or even worse, working through mega-marketplaces taking yet more cuts from your revenue. Or worst, doing that with your competitors instead.

To deal with 3rd party agents successfully, you’re going to need all of the functions from your CMS that you need for your own 1st party agents. But you’re also going to need a bit more.

CMSs are only just starting to address this space. In our experience, you'll find three types of vendor responses: those with immediate roadmap commitments (6-12 months), those 'monitoring the space' (translation: 18-24 months minimum), and those who haven't started thinking about it.

Ask when and how they’re planning to deliver these features:

· Agent-aware interaction metrics

· Machine readable microformat markup

· Dynamic content structuring (summaries and Q&A pairs)

· Dynamically generated AI crawler files (llms-full.txt, llms.txt)

· Agent-source and type access tuning 

· Agent-to-agent interfaces / servers

We should just say – one of these features is unlike the others. Agent-to-agent interfaces are by far the most uncertain feature on the list, and you’re unlikely to get realistic commitments from any vendors yet. The rest of the features though, are immediately important and should be well-within reach.

Getting from questions to decisions

Picking out the platforms that are going to deliver the experiences your customers demand has always been a complex and important decision. Making sure that those platforms can provide the pivot to a new world of digital experiences only makes it harder, especially given the uncertainty about what that world will look like and how platforms and brands will adapt.

The key is to map these technical capabilities back to your immediate action plan. For fixing visibility issues, prioritise platforms with strong semantic search, microformat support, and AI crawler optimisation. If you're investing in AI-enhanced interactions, you need robust APIs, semantic chunking, and prompt analysers today. Your measurement strategy demands platforms with agent analytics and quality metrics built in. Start with assuring the capabilities that support your immediate plan of action, then look to make sure your chosen platform has a credible roadmap for the rest.

Remember: your customers have already shown you where they want you to be - talking to their AIs instead of directly to you. The platforms you choose now will determine whether you can meet them there, or whether you'll be left behind having one-sided conversations. Focus on capabilities that deliver immediate value while building towards the agent-to-agent future, because the one thing we know for certain is that standing still isn't an option.

If you need help with planning how your customer experiences could change, and what capabilities you’ll need to support that, we’d be delighted to talk to you.