Your AI Agent Does Not Know Who You Are

Pillar: AI and AI Integration

Sundar Pichai stood on stage at Google I/O 2026 and declared the arrival of the “agentic era.” Not an AI that answers your questions. An AI that acts on your behalf. Gemini Spark, the centrepiece, runs 24/7 in the background, booking restaurants, managing your inbox, buying your shopping, and building custom tools on the fly. The Gemini app has passed 900 million monthly active users. Google is processing 3.2 quadrillion tokens a month, seven times what it was processing a year ago. The company is spending up to $190 billion on infrastructure this year alone.

This is not a small announcement. This is the moment one of the world’s most powerful technology companies formally transitions from “AI that responds” to “AI that decides.”

And nobody is asking the obvious question: how does it decide?

The Problem With Agents That Don’t Know You

An AI agent that books your holiday, filters your emails, and curates your news feed is making decisions on your behalf. It is exercising preference, judgement, and priority. It is, in effect, performing a compressed version of what psychologists call appraisal: evaluating what matters, what can be ignored, and what deserves attention.

But here is the catch. Appraisal is not a generic process. It is deeply personal. It is shaped by your motivations, your identity, your tolerance for risk, your need for connection, and your desire for control. Two people given the same situation will appraise it differently because they are different people. Not because one is right and the other is wrong, but because their psychological needs are not the same.

The STAR Framework identifies four fundamental motivational profiles. Socialisers are driven by relatedness. Thinkers by competence. Adventurers by autonomy. Realists by security. These are not quirks. They are the architecture of human motivation, rooted in Self-Determination Theory and supported by decades of research.

When Gemini Spark decides which emails are “important” and which can wait, it is making an appraisal. When it curates your news feed, it is making an appraisal. When it books a restaurant, it is making an appraisal about what you value: convenience, atmosphere, novelty, or proximity to friends. But the agent does not know whether you are a Socialiser who wants the busy place where everyone goes, a Thinker who has already read every review, an Adventurer who wants the new one nobody has tried, or a Realist who wants the reliable one that is close to home.

It treats you as a generic user with generic preferences. It optimises for the average. And the average is nobody.

3.2 Quadrillion Tokens, Zero Emotional Intelligence

The scale is staggering. Three point two quadrillion tokens a month. Eight-generation TPUs. $190 billion in capital expenditure. Google is building the most powerful inference infrastructure in human history.

But processing volume is not understanding. Appraisal Theory of Emotion tells us that emotional responses are generated not by events themselves but by how those events are evaluated in context. An AI agent can process a trillion tokens about a calendar conflict. It cannot evaluate that conflict the way you do: with the knowledge that Tuesday’s meeting is with a colleague you trust, Thursday’s is with one you don’t, and the school play is on Wednesday evening and you have already missed two this term.

That is not a data problem. More tokens will not solve it. It is a motivation problem. The agent does not know what drives you because it has never asked, and it was never designed to ask.

What the “Agentic Era” Actually Means

Pichai described a future where AI agents “do, not just answer.” That sounds like liberation. Delegation. Efficiency. And for some tasks, it will be.

But delegation only works when the delegate understands your priorities. A human assistant who booked the cheapest restaurant every time, ignoring the fact that you value ambience and proximity to friends, would not be a good assistant. They would be a bad assistant with good intentions. An AI agent that optimises for the wrong priority is the same thing, except it operates at the speed of 3.2 quadrillion tokens a month and you cannot have a quiet word with it over coffee.

The risk is not that AI agents are incompetent. The risk is that they are competent at the wrong thing. They will optimise aggressively, consistently, and at scale for a set of assumptions about human preference that may not match yours. And because they work in the background, silently, you may not even notice the divergence until you find yourself in a life that has been curated by a system that does not know who you are.

The STAR Response

This is not an anti-AI argument. It is a design argument. If the agentic era is going to work, agents need to understand motivation, not just behaviour. They need to know not just what you did last time, but why you did it. They need a model of human psychology that goes beyond click patterns and purchase history.

The STAR Framework provides exactly that. Four motivational profiles, each with distinct decision-making patterns, communication preferences, and emotional triggers. An agent that knows you are an Adventurer will curate differently from one that knows you are a Realist. Not because it guesses from your browsing history, but because you told it, and it understands what that means.

The technology to build motivational awareness into AI agents exists today. The psychological models exist today. What is missing is the will to build it, because the current approach, optimise for the average and hope for the best, is cheaper, easier, and looks impressive on a keynote stage.

But 900 million users are not “the average.” They are 900 million individuals with different motivations, different needs, and different ideas of what a good decision looks like. The agentic era will only succeed when the agents catching the billion-dollar infrastructure understand the people they are supposed to be serving.

Until then, you are being optimised by a system that has never asked what you actually want.

David Chadderton spent his twenties and thirties teaching people how to make life-or-death decisions at forty thousand feet. He now applies the same principles to consumer psychology, which, depending on the brief, can feel equally high-stakes. He’s the creator of the STAR Framework and the author of The STAR Framework: Rewriting the Rules of Consumer Engagement (NYC Big Book Award 2025), The STAR Operating System: Decode Mindset, Understand Motivation, Transform Human Behaviour, and Dear Algorithm, It’s Not Me, It’s You. By day, a Chief Marketing Officer. By night, a behavioural science obsessive who writes The Unoptimised Human because he can’t stop thinking about why people do what they do.