The AI-ready commerce ecosystem starts with your architecture, not your chatbot

But while marketing teams chase flashy, customer-facing AI pilots, the real enabler, or bottleneck, sits at a far more foundational level.

The hard truth is that no amount of AI-powered personalisation, recommendation engines, or digital agents will work if the underlying data and architecture aren’t ready. Without flexible pipelines, orchestrated systems, and a clean data foundation, AI initiatives collapse under their own weight—clunky, costly, and underperforming.

That’s why the real question isn’t “which AI tools should we adopt?” but “is our commerce ecosystem AI-ready?” 

And the bar for readiness is rising fast. New approaches, such as Model Context Protocol (MCP) servers, are emerging that take the concept of a composable AI ecosystem even further. These systems make it possible to orchestrate multiple AI models like modular services and enable the kind of agent-driven commerce experiences that are starting to take shape. 

For most organisations, though, those future AI capabilities will only be accessible once the foundations are solid. Similarly, until then, CMOs will never be able to deliver the kind of personalised experiences customers now expect.

For years, personalisation in commerce meant targeting broad segments with pre-defined campaigns: think “back-to-school shoppers” or “holiday gift buyers.” But today’s customers expect interactions that respond to their individual behaviours and signals in real time. Whether that’s a search for running shoes at 9 a.m. or abandoning a cart at 9:15.

Meeting that expectation requires moving from a reactive model to an agentic model. In an agentic approach, intelligent systems don’t just wait for a customer to act, but proactively curate, recommend, and even transact on their behalf. This is what marketing leaders envision when they talk about “AI-driven personalisation.”

Here’s where MCP servers start to matter. They extend the idea of composable commerce into the AI layer itself, allowing multiple specialised models to be orchestrated like modular services. Instead of relying on a single recommendation engine, MCP servers can dynamically select the right model for the context, whether that’s predicting a customer’s intent, optimising a price, or generating a piece of content, and then swap it out as new capabilities emerge.

The catch is that none of this works on top of brittle, monolithic platforms or nightly batch jobs. A retailer that only updates customer segments once a day can’t take advantage of real-time orchestration. A DTC brand hardwired into a single commerce suite won’t be able to plug in new agentic capabilities without major disruption.

This is why foundational readiness matters. Treating commerce as a living system, powered by data and capable of adapting in real time, is what separates the brands that can operationalise AI from those that just experiment with it.

AI might look like magic on the surface, but underneath, it’s nothing more than a machine that runs on data. If your data is fragmented, stale, or inaccessible, no model, no matter how advanced, will produce meaningful results.

This creates a balancing act: more data fuels better personalisation, but it also drives up storage and processing costs, especially with usage-based cloud billing. While not enough data limits AI effectiveness, leaving personalisation shallow and generic.

Take a common scenario: a global retailer invests in a new personalisation engine, only to discover that customer data is locked in regional silos, updated in overnight batches, and riddled with gaps. The result? The AI makes irrelevant recommendations, customers churn, and marketing loses faith in the “AI investment.”

To be AI-ready, your team needs a machine of its own. That encompasses processes, systems, and skilled people who can quickly pull new data into your pipelines and make it actionable. With that flexibility, you can test and deploy emerging AI capabilities as they arrive, without being crippled by missing or poorly maintained data.

Three simple questions can reveal whether your organisation is ready to harness the advantages of AI (or if you’re just dabbling with expensive tools):

• Can your architecture react to customer behaviour in real time, or are you still relying on nightly jobs?
• Can you pivot your data feeds to new channels and agent-driven experiences, or is everything hardwired?
• Do you know which data is worth the cost to store, and which is superfluous?

Having the right data is one thing. Making sure it flows to the right places, in the right format, at the right time is another. That’s where orchestration comes in, and it’s often the most overlooked layer in enterprise commerce architecture.

On the back end, data pipeline tools and integration platforms (think MuleSoft, Boomi, or modern iPaaS solutions) let you route information across systems using pre-built connectors. Without this orchestration layer, marketing might stand up a new loyalty app, only to find that it can’t talk to the customer database without months of custom integration.

On the front end, a Digital Experience Platform (DXP) gives business teams the flexibility to plug new capabilities into the commerce stack without leaning on developers for every update. For example, a merchandising team might want to test AI-generated product recommendations on the homepage. With orchestration, that can be done in weeks or less. Without it, the request often gets buried in an IT backlog.

To close the loop, those real-time customer signals need to feed back into your system of record—CRM, CDP, or ERP—so every downstream system has access to the same intelligence. We’ve seen global brands invest in advanced data solutions, but because orchestration was missing, the marketing team still exported CSVs manually to trigger campaigns.

When orchestration is absent, AI initiatives stall. When it’s in place, business teams can experiment quickly, IT retains architectural integrity, observability, and security, and the entire commerce ecosystem moves as one.

Most commerce brands’ AI initiatives are hindered because their foundations can’t support them. Data is fragmented, orchestration is missing, and architecture is too rigid to adapt in real time. 

When teams and systems aren’t aligned, AI projects become expensive experiments. Marketing spins up pilots, IT resists half-baked integrations, and customers see little impact. The only way forward is to create a shared roadmap where marketing’s strategic goals are mapped onto the data, architecture, and orchestration decisions that IT controls.

The real work of becoming “AI-ready” is building a flexible commerce ecosystem in which clean data flows in real time and can support whatever AI capability comes next.

At DEPT®, our focus is on helping brands align their flashy front-end ambitions with foundational back-end readiness. By unifying strategy and creative with data and tech, we enable CTOs to deliver the architecture and orchestration their CMOs need to turn AI from a cost centre to a driver of durable, enterprise-wide growth.