🎥 The YouTube version is also available:
The Relationship Economy: How AI Is Transforming Value, Trust, and Business
Introduction
Until now, AI has primarily been used to determine what to recommend.
People who bought this product also purchased another product.
This service is popular with users in the same age group.
Show this advertisement to people who previously viewed this page.
These mechanisms are typical examples of Recommendation AI: systems that predict user behavior from purchase histories and attribute data.
Yet what people need is not determined by their past history or attributes alone.
The same person behaves differently on weekdays and weekends. What they seek changes when they are alone versus when they are with family. The support they need while traveling for work can be entirely different from the support they need while on vacation.
What becomes truly important in the AI era is not:
“What has this person bought?”
but:
“What state is this person in right now?”
We call this way of thinking State Understanding.
Traditional AI Has Been Product-Centered
Traditional marketing and recommendation systems have placed products at the center. They analyze data such as:
- Which products sell well
- Which products are purchased together
- What similar users choose
- What a user has viewed in the past
- Which coupons a user has used
They then predict the product most likely to be purchased next.
This approach is highly effective for e-commerce sites and retail services with many products. But when a user’s real objective is not the purchase of a product itself, its recommendation accuracy has limits.
Suppose a user is looking for a jacket. Traditional AI might recommend products based on:
- Brands purchased previously
- Popularity rankings
- Purchasing patterns among similar age groups
- Browsing history
- Price range
But a jacket may not be what the user truly needs.
They may have an important presentation the following week. It may have suddenly become cold. They may be looking for clothing to wear on a business trip. Or a family member may have asked for a jacket as a birthday gift.
The background purpose differs even though the visible behavior—“looking for a jacket”—is the same. As long as AI sees only products, it cannot fully understand these differences.
What Is State Understanding?
State Understanding is an approach to comprehensively understanding the current state of a person, organization, community, or system by integrating multiple sources of information.
The main elements that compose a state include:
- Context: the current situation
- Goal: the outcome to be achieved
- Intent: the intention behind an action
- Constraint: conditions that must be respected
- Role: the role currently being performed
- Location
- Time
- Relationship: relationships with surrounding people and entities
- Past Behavior
- Capability: available capabilities and authority
State Understanding does not view these elements separately. It integrates them as a single state.
In other words, it estimates the support a person needs now from:
attributes + history + current situation + goal + intent + constraints.
The important point is that a state is not fixed. The state of the same person changes with time, place, purpose, and companions. AI must therefore continuously understand a changing state—not merely a predefined profile of the person.
Context: What Is Happening Now?
Context is the foundation of State Understanding. It is the background information available when AI makes a decision.
It can include:
- Current date and time
- Location
- Weather
- Companions
- Services currently in use
- Immediately preceding actions
- Events occurring nearby
- Inventory and congestion conditions
- Roles within an organization
- Applicable rules
The answer to the same question can change when the Context changes.
Consider the question: “Where should I go next?”
For a traveler, the AI may suggest a tourist destination. For a sales representative, it may suggest the next customer visit. In a disaster, it needs to direct the person to an evacuation site.
The wording of the question alone does not determine the best answer. To make an appropriate decision, AI must understand the circumstances in which the question is being asked.
Much of the reason generative-AI answers feel “not wrong, but not usable as they are” is a lack of Context.
Goal: What Does the Person Want to Achieve?
A Goal is the state that a user or organization is seeking to achieve. Traditional systems have processed activity primarily around the operations users perform:
They searched. They viewed a product. They opened a document. They made an inquiry.
But an operation is not the purpose itself.
For example, the action of “searching for a hotel” may reflect very different Goals:
- Securing accommodation for a business trip
- Planning a family vacation
- Staying near an event venue
- Finding a long stay within budget
- Responding to a sudden flight cancellation
When the Goal differs, the optimal recommendation also differs. If AI can understand not only search terms and activity history but also what a person ultimately wants to achieve, it can provide support that goes beyond products and services.
This is a shift from AI that introduces products to:
AI that supports the achievement of goals.
Intent: Why Is This Action Being Taken?
Intent is related to Goal, but it describes the intention behind an action. A Goal is the outcome to be achieved; Intent is closer to the reason why a particular action is being chosen.
Suppose a user is looking for a coupon. Their Intent could be:
- To purchase something as inexpensively as possible
- To try a new store
- To find a place the entire family can use
- To use points before they expire
- To find somewhere to stop after attending an event
The same action can reflect different Intents. Analyzing only visible operations therefore fails to reveal the support that is truly needed.
When AI understands Intent, it can make proposals aligned with the meaning of an action rather than merely react to it.
Intent, however, is not always stated explicitly. It must be inferred from conversation, search, past behavior, and current Context. An inferred Intent must never be treated as an established fact. It should be managed as an uncertain hypothesis and, when appropriate, confirmed with the person.
How Do Proposals Change Through State Understanding?
Imagine that a family opens a local app on a Saturday morning.
Traditional AI might display:
- Recently used stores
- Popular products
- Recommended coupons
- Past purchase history
State Understanding, by contrast, integrates information such as:
- It is a holiday
- The family is together
- The weather is good
- A sporting event takes place in the afternoon
- They are near the venue
- Lunchtime is approaching
- Children’s events are also taking place nearby
- Several participating stores lie along their route home
It can further infer that the Goal is “to enjoy the day as a family,” and the Intent is “to spend time comfortably before and after the event.”
The AI can then propose not simply products, but a coherent experience:
- Family-friendly restaurants available before the game
- Routes to the venue that avoid congestion
- Nearby activities for children
- Stores to visit after the game
- Shopping timed to the family’s journey home
What the AI understands here is not products.
It is the family’s current state.
Organizations Have States Too
State Understanding does not apply only to individuals. Companies, organizations, projects, and communities also have states.
Consider a project:
- The deadline is approaching
- An important specification remains undecided
- Team capacity is insufficient
- The customer has submitted additional requests
- A quality risk has been found
- Executive approval is required
When these are integrated, we can understand that “the project is in a state of heightened delivery risk and requires prompt decision-making.”
Traditional business systems display tasks and progress rates. AI with State Understanding can infer the current state from multiple sources and suggest:
- Which decisions should take priority
- Who should receive an escalation
- Which constraints should be checked
- Which knowledge should be consulted
- What actions can reduce risk
This is the difference between a simple business-support AI and a Decision AI.
From Knowledge Flow to State Understanding
In the previous article, we introduced Knowledge Flow, which transforms enterprise documents and data into knowledge that AI can use.
Through Knowledge Flow, AI can understand an organization’s rules, past cases, products, customers, operations, and structure.
But possessing knowledge alone is not enough. AI must select from that knowledge what is relevant to the present situation and compose the current state.
If Knowledge Flow is the mechanism that delivers enterprise knowledge, State Understanding is the mechanism that integrates that delivered knowledge with real-time information to understand the present.
The flow is as follows:
Enterprise Knowledge
↓
Knowledge Flow
↓
Context
↓
Goal
↓
Intent
↓
State Understanding
↓
Decision
↓
Action
Knowledge Flow supports “what is known,” while State Understanding reveals “what is happening now.”
From State Understanding to the Relationship Economy
The Relationship Economy emphasizes building ongoing relationships among users, companies, communities, and AI, rather than maximizing a single transaction.
To do this, we must understand the other party’s current state. If we continue to make one-sided proposals without understanding that state, the relationship weakens. If, instead, we understand their Context, Goal, and Intent, and provide appropriate support at the right time, trust emerges.
This creates the following cycle:
State Understanding
↓
Relevant Support
↓
Better Decision
↓
Trust
↓
Relationship
↓
Value
Economic value in the AI era will be determined not by how many products can be presented, but by how deeply a state can be understood and how consistently appropriate support can be provided.
State Understanding Requires Boundaries
State Understanding has significant potential. But that does not mean every kind of information may be collected to understand a user’s state.
Location data, health information, family information, purchase history, and conversation content all require extremely careful handling.
For this reason, implementing State Understanding requires mechanisms such as:
- Consent from the person concerned
- Clear statement of purpose
- Use of only the minimum necessary data
- Access controls
- Distinction between inference and fact
- Records of decision rationale
- Human confirmation
- Data retention limits
Improving the accuracy of State Understanding is not the same as excessively monitoring people. The scope AI may understand and the scope within which it may intervene must be explicitly designed.
The mechanism that supports these boundaries is the Trust Infrastructure, which will be introduced in a future article.
Conclusion
In the AI era, value will not be created merely by AI that recommends products.
It will be created by AI that understands the Context in which people and organizations currently exist, the Goals they pursue, and the Intents behind their actions.
Through State Understanding, AI evolves from a system that reacts to past history into one that understands the present situation and supports the achievement of goals.
This understanding produces continuous support; support produces trust; and trust strengthens relationships.
It is a critical foundation for realizing the Relationship Economy.
Next, we will move beyond the state of individuals and organizations to examine how the connections among people, companies, stores, events, communities, and AI can be represented: Community Graph — Relationships Become Assets.
Chinoba
Intelligence as Relationship
Research Platform
founded by
Masao Watanabe
AI Systems Architecture
Decision Trace
Human–AI Coordination
Algorithmic Governance
Related Research
This topic is part of the Chinoba Knowledge Base.
コメント