Agentic Experiences Don’t Replace Digital Experience. They Raise the Bar.

We recently wrote about what happens when the website stops being the only interface between a business and its customers. Anthropic’s commerce-agent blueprint, and platform moves like Optimizely’s Virtual Teammates and CMS MCP server, made the shift concrete: AI agents can now search, compare, decide, and act on a customer’s behalf, not just answer questions about a product.

That raises a more useful question than “should we build an agent?” It’s this: what has to be true about a business’s digital experience before an agent can act on top of it reliably?

The honest answer is that agentic experiences don’t remove the need for good digital experience work. They expose exactly how much of it was already missing.

The customer journey stops being a path

A conventional journey is a sequence: search, browse, filter, compare, add to cart, check out. A business can optimize each step because each step is visible.

An agentic interaction skips most of those steps. A customer describes what they want, and the agent is expected to get there directly. That means the journey still has to be designed, just not as a sequence of screens. It has to be designed as a set of outcomes and the information needed to reach them, regardless of which interface a customer or an agent uses to ask for it.

UX has to design for delegation, not just navigation

Traditional UX assumes a person is making every decision, one screen at a time. When an agent is completing steps on a customer’s behalf, the real design questions shift: where should the agent be allowed to act without asking, and where does a person need to stay in the loop? What does the agent need to know to make a good recommendation instead of a generic one?

Those aren’t cosmetic questions. Answering them badly is how a business ends up with an agent that either annoys customers by asking too much, or damages trust by deciding too much on its own.

Content and platform architecture has to be legible to machines, not just people

Most content today is still built for a person reading a page: a headline, some body copy, an image. An agent working on a customer’s behalf needs something different, content that’s structured, labeled, and connected in ways a system can reason about, not just render.

This is where composable architecture stops being a technical preference and becomes a practical requirement. If product information, policies, and account data live in disconnected systems with inconsistent structure, an agent inherits that inconsistency, and so does the customer relying on it.

Integrations become the actual product

An agent is only as capable as the systems it can reliably connect to. Catalog data, cart and checkout, order history, support systems: if those integrations are fragile, partial, or held together with workarounds, that fragility shows up immediately in what the agent gets wrong.

This is different from a website with a broken integration quietly degrading in the background. An agent surfaces the gap in real time, in front of the customer, while it’s trying to complete the task.

Transaction systems need explicit boundaries

The moment an agent can create a cart, apply a discount, or complete a purchase, a business needs clear answers to a different kind of question: what can this system do without asking, and what always requires a person to confirm? Those boundaries have to be built into the transaction layer itself, not left to the agent’s judgment or a support team catching mistakes after the fact.

Where this leaves AI in the conversation

None of this starts with “which AI should we add.” It starts with whether the experience underneath can support an agent acting on it at all. The organizations that will do this well aren’t the ones racing to launch an agent first. They’re the ones who already have a customer journey, an architecture, and a set of integrations that can support one without falling apart.

What to assess now

• Does your customer journey depend on someone knowing where to click, or does it hold up if someone simply describes what they need?
• Is your content structured well enough for a system to reason about it, not just render it?
• How reliable are the integrations connecting your content, commerce, and customer data today?
• Where would you want a person to stay in control if an agent could act on your customer’s behalf tomorrow?

An agent doesn’t fix a fragmented digital experience. It reveals one.

Ready to see where your digital experience actually stands?

Talk to our team about your customer journey, your architecture, and where agentic experiences genuinely fit, before deciding where to add one.

Last Updated: September 2026