Part 2: Website Renewal in the AI Era — Defining the Role of Digital Touchpoints and Focusing on the Bottleneck

In Part 1, we said that the first question in a website renewal is not “What kind of site should we build?” but “Which management issue should we solve?” We also said that AI has lowered the baseline for production—the “nice to have” threshold—and that the real contest has shifted from improving the “I” in ROI (Investment) to designing the “R” (Return).

So how should we design that R? Part 2 begins there.

Part 1: Website Renewal in the AI Era: Before “Rebuilding,” Ask Which Management Challenge You’re Solving

The role of a touchpoint is to “move a state”

What is the lever that moves R? It is defining the role of each touchpoint. But what exactly do we mean by “role”? We need to be precise here.

We define the role of a touchpoint as: “Whom do we move, from what state, to what state?” In other words, it is a state transition. If we think about it, management issues always appear somewhere in the journey as a state that is not moving. People do not know us. We do not remain in the shortlist. People consider us, but we lack the deciding factor. Candidates do not apply. That single point where movement has stopped is the bottleneck. The role of a touchpoint is to move the state at that bottleneck. Nothing more, nothing less.

When we define it this way, the meaning of R becomes clear. The size of R depends on how much we move the bottleneck state. Conversely, no matter how much we polish an area that is not the bottleneck, R will not increase. If one narrow section determines the total flow, widening a pipe that is already wide does not change the amount that flows through. The same principle applies here.

Then how should we describe that “state”? “Can” or “cannot” is not enough. People do not act only because an action is possible or impossible. We need to describe what they know, what they believe, and how they feel. Only then have we captured the state. Take the state of “not being known,” for example. If we are not cited in AI search, the opportunity to be known does not arise in the first place. GEO—optimization that makes content easier for generative AI to cite—is one way to move that state.

At the same time, in our experience, the bottleneck in B2B often lies deeper. The information is sufficient. The features are not a problem. And still, people do not feel, “We can trust this company.” Unless that state moves, we will not be chosen by name. What moves that state is not simply adding one more page. It is designing the experience itself.

You may think that states cannot be measured. But states appear in behavior. Whether we are known appears in branded searches. Whether we are trusted appears in direct traffic and return visits. Reputation appears in reviews and external mentions. In daily operations, we track these behavioral signals. At milestones before and after a renewal, we also measure perception directly through surveys, asking whether people feel we are the right fit. Once we can describe the role this clearly, the meaning of a standard renewal process changes completely.

“Fit” is a decision about allocation

For a long time, we have worked with the keyword “fit.” In the AI era, the meaning of this word has become even more important. Fit is the degree to which these three things align: the management issue, the role assigned to the touchpoint, and the implementation. If they are misaligned, no amount of optimization will make R work for management. If they align precisely, a small investment can create a large return.

AI has made the range of things we can create practically limitless. That is exactly why, paradoxically, the decision of where to go all in and what to let go of has become the scarcest value. We do not try to make every touchpoint perfect. We place a bold bet on the one point that is the bottleneck. And if that bottleneck sits outside the website, the most appropriate answer may be: “For this issue, you should invest in another touchpoint before the website.” Being able to decide not to build is the value we want to protect most.

Five steps to design R

So where should we start when we actually design R? We generally proceed in five steps. It is a shift from “production first” to “role definition first.”

First, we translate the management issue into the language of states. “We want to increase new inquiries” is not yet something we can design around. Whose state must change, and how, for us to say the issue has been solved? Only when we bring the issue down to that level does it become measurable.

Second, we lay out the journey and identify where the state has stopped moving. Are people unaware of us? Do they know us but fail to keep us in the shortlist? Do they shortlist us but not fully trust us? The single point we find here is the bottleneck. This step also includes the question: “Should we invest in the website at all?” If the bottleneck is outside the website, the answer may be, “Not the website this time.”

Third, we define the role of the touchpoint for that one point. We write, in one sentence, “Whom do we move, from what state, to what state?” If we cannot write that sentence, we can also decide that we do not need to create that page.

Fourth, we decide the allocation. We do not try to make everything perfect. We concentrate budget and time on that one point. In many cases, a full rebuild is unnecessary. Improving the part that works as leverage creates a larger R with the same budget.

Fifth, we measure whether the state moved. If the state is “being known,” we may use branded search. If the state is “being trusted,” we may use direct traffic or return visits. We decide these indicators in advance, aligned with the states identified in step two. That becomes the starting point for the next cycle.

In other words, if we skip these five steps and jump into “let’s rebuild every page for now,” we will not be able to explain afterward what moved and what did not.

The detailed methods for each step will be covered in separate articles. Changes in the AI search era and the basics of GEO/LLMO, priorities for avoiding unnecessary rebuilds, reviews of information architecture and UX, and technical foundations such as page speed and structured data—these two articles are a map that connects those topics through the question of why we do them.

Why an experience agency?

Everything we have discussed so far only works when we can start from management issues, define roles, judge the right allocation, and carry the work through design, implementation, and operation as one connected effort. The key is whether we can take responsibility for that entire sequence, not just for making something.

We have a cross-functional diagnostic methodology called Experience Assessment. It evaluates the experience of the stakeholders within the target scope, identifies problems, and defines issues. It is not a tool for “checking the website first.” It is a tool for diagnosing which touchpoint we should invest in, and how much, against a management issue.

Our company-wide use of AI supports the production and operational efficiency needed to go further upstream. By combining AI with service design, we can keep teams compact, improve quality, expand the scope we can handle, and deliver 1.2 to 1.5 times the value within conventional cost expectations. We redirect the value created by more efficient production into role definition and appropriate allocation. That is our position as an experience agency.

First, measure the current state

Finally, we have one proposal. Before rebuilding, measure where the bottleneck lies in the journey against your company’s management issue, and how much the website as a touchpoint is moving that state. We offer a diagnostic service that measures touchpoint allocation and return. Starting with a current-state check from a GEO perspective for the AI search era is also fine. Before asking “Should we rebuild?” why not make visible, together, what should be solved through which touchpoint?

For consultations and inquiries about website renewal or web strategy in the AI search era, please contact us here.