What happens to digital analytics when there are no more clicks?

A few weeks ago, while camping at 2,000 meters in the Swiss Alps, I was chatting with an 80-year-old fellow traveler. When I mentioned my work in in the field of digital analytics and AI, he paused and said, “This AI thing will replace the internet someday.”

At first, I laughed it off. But as I lay in my tent that night, I started thinking — maybe he wasn’t entirely wrong. Maybe AI won’t replace the internet, but it could transform it beyond recognition, or how think of the internet.

This post is not a prediction. It’s a provocation. A thought experiment that follows the signs we’re already seeing. What if the website is on its way out? What does that mean for digital analytics — and for us as analysts?

Definitions

To get our hands around this topic, we need two important definitions:

Agent

An AI assistant that interacts on behalf of the user — understands their goals (not just commands), asks questions, makes decisions, and takes action. For example if you are booking a flight, the agent asks questions like “why are you travelling” or “whats your airline preferences” and then makes the booking.

MCP (Model Context Protocol)

A translation and orchestration layer that connects the agent to real-world services via APIs and adds structured context.

Think of it like a travel adapter — it helps the agent “plug into” different systems in a standardized way.

From pageview to intent: A new analytics era

Let’s break this down through a real-world use case — from my industry, but you can apply this to any other industry: Jasper, 34, wakes up on a Monday morning with a sore throat, fatigue, and a mild fever. He wants quick clarity:

• Should he see a doctor?
• Can he get help online?
• How fast can he get an appointment?

His journey helps us understand how digital analytics has evolved — and where it’s heading. I framed this into four different stages:

• Yesterday
• Today
• Tomorrow
• Future

YESTERDAY: Web Analytics Classic

The era of clickstreams and proxy metrics

Jasper feels sick, he googles his symptoms. And then he may finds his health insurer’s website, a clinical website or the doctors website. Let’s assume he is navigating to the health insurance site. He uses the doctor search on the website and then clicks through to a phone number and calls his doctor.

As analysts, we tracked:

• Entry pages and marketing sources
• Pageviews and sessions
• Conversions (e.g. entry > phone number click)

The problem in the good old days was, that we missed the why, we only covered the what and where. Did Jasper get the help he needed? Was the experience frustrating or smooth? We had data — but not insights. These were at least proxy metrics, not direct indicators of success. And worse, they only covered the digital portion of an on- and offline journey. Digital search leads to an analog action. Lot’s of friction for Jasper.

TODAY: Chatbots & Self-Service

The rise of guided flows and in-app journeys

Now Jasper skips Google. Instead he opens directly his insurer’s app, uses a symptom checker (a guided questionnaire), and gets advice or self care tips. Then he books an appointment directly within the app.

We now track events like:

• symptom_check_started
• booking_form_opened
• booking_confirmed

These events give us much better signals of user intent and progress. We move beyond pageviews and funnels and drop-off rates tell us more. But we’re still bound to predefined flows — linear, structured, and app-specific. We still infer Jasper’s goal. We don’t always confirm if it was fulfilled.

TOMORROW: AI Agents & Intent Recognition

From doing the task to delegating it

Jasper doesn’t open an app anymore. He simply tells his voice assistant: “Hey Google, I feel sick.”

An AI agent takes over. It asks clarifying questions, checks his insurance coverage and then schedules an appointment (Agent > MCP > API) – no clicks, no screens. With the help of the MCP Layer, the agent can act, and book an appointment (is the agent knows the structure of the doctors/insurance API).

Analytics here becomes radically different:

• Intents like book_doctor, symptom_throat_pain
• Context tokens (e.g., insurance status, location)

And KPI’s move upstream:

• intent resolution rate
• escalation to human (fallback)
• dialogue length

This isn’t UX anymore. It’s conversational orchestration. We’re measuring the quality of interactions, not clicks. We’re no longer tagging pages. We’re instrumenting agents and tracking intent flows across APIs and language models.

FUTURE: Proactive Health Agents

What if Jasper never even asks?

This is where things get fascinating — and a little unsettling. Jasper’s wearable detects abnormal sleep. His voice sounds different. His calendar is open in the afternoon. Without prompting, his health agent says:

“Hey Jasper, you seem unwell. I’ve reserved a 3pm slot with your GP. Shall I confirm?”

Now we’re in the invisible era. The agent:

• Gathers real-time contextual signals from the human data collection
• Acts on Jasper’s behalf
• Only asks for approval

Our metrics become:

• Autonomy score (agent-initiated resolutions)
• Agent trust score
• Goal fulfillment rate

There’s no funnel. No path to trace. Only intent → outcome, mediated entirely by machines. The human is only receiving the benefit.

Agent ecosystem: Who owns the interaction?

This future opens a new strategic question: Who builds the agent? And who owns the user relationship?

I see three emerging models:

1. Big Generalist Agents
   Like ChatGPT, Gemini, or Apple Intelligence — one agent to rule them all.
2. Specialist Agents
   Domain-specific AIs like an Insurance Health Coach, deeply integrated into business logic, contracts, and medical networks.
3. Public Infrastructure Agents
   Created by governments or alliances (e.g., a Swiss Health Agent), built on privacy-first principles.

In my opinion it will be a hybrid model. You will have a dominant user-facing meta agent (ChatGPT / Gemini) that delegates tasks to trusted specialist agents in health, finance, etc.

Each of these scenarios requires very different data strategies, API interfaces, and analytics frameworks.

Websites: What will happen?

Not all websites will disappear. But their role is shrinking fast. I see four different stages:

Likely to disappear

• Transactional flows (e.g. booking, form submissions)
• Repetitive tasks (e.g., quotes, appointments)

Likely to shrink

• Information-heavy sites (e.g., FAQs, calculators)
  Agents retrieve answers faster

Likely to stay

• Trust- and emotion-driven journeys (e.g., onboarding, storytelling)
• Social discovery (e.g., LinkedIn, Reddit)

Likely to evolve

• eCommerce, entertainment, and exploration-based platforms will merge browsing with agent augmentation.

The analyst’s role is changing

This shift demands new skills, new tools, and a new mindset.

From	To
Page-based analysis	Intent flow tracking
Tags & pixels	Prompt metrics & vector embeddings
UX journeys	Conversation orchestration
GA & GTM	Langfuse, PromptLayer, Vector DBs
Session analysis	Goal fulfillment & agent confidence

You’ll need to:

• Understand LLM architectures and prompt metrics
• Measure agent success, not user clicks
• Collaborate with AI product teams, not just marketing

Tools like:

• Langfuse (analytics for AI workflows)
• PromptLayer (managing LLM prompts)
• Vector databases (the new CMS for AI)

This is not a bolt-on change. It’s a ground-up redefinition of digital value creation.

---

So What Now?

This shift isn’t years away — it’s already happening. And as analysts, we can either adapt or be left decoding metrics no one cares about anymore. So start small: follow AI tools, join prompt design discussions, and think bigger than the page.

• Think in intents, not sessions
• Track outcomes, not conversions
• Learn how to instrument agents and APIs
• The ultimative stress test. Ask yourself: What does my analytics look like without a website?

---

Final Thought

“The users are still there. The clicks are not. Are you ready to track meaning — not movement?”

We are entering a world where frontends shrink, where conversations replace clicks, and where the only metric that really matters is:
Did the user get what they needed?

Let’s build analytics for that world — one agent at a time.

My presentation (pdf) from the Analytics Pioneers Summit 04.07.2025: