Our Philosophy
치원 (Chiwon): far-reaching wisdom. Named for the 최부자 (Choi Buja) family of Gyeongju, Korea — who sustained prosperity, integrity, and community stewardship for over 400 years without exploitation.
AI is transforming how commerce works. Agents are beginning to act autonomously in sales conversations, procurement decisions, contract negotiations, and relationship management — on behalf of companies and individuals alike. This shift is accelerating. And the stakes are high.
Done well, agentic AI in commerce can level the playing field — giving every professional the context and intelligence of the world's best, making buying experiences genuinely more efficient and honest, and creating value that flows to both sides of every transaction.
Done poorly, it becomes a tool of extraction. AI that manipulates rather than informs. That optimizes for conversion at the expense of trust. That treats the buyer as a target rather than a partner.
We founded chiwon.ai because we believe the companies that will endure are the ones that refuse the extractive path — not because it's legally safer, but because it's wrong, and because sustainable prosperity has never been built on exploitation.
The 최부자 family governed 400 years of wealth by a simple ethic: never take from those who have nothing; give back to your community before you take for yourself; do business in a way you'd be proud to explain to anyone. We take that inheritance seriously. It is the philosophical foundation of everything we build.
Agentic AI — AI that takes actions autonomously, not just answers questions — represents a fundamental shift in how commercial relationships work. What matters is the orientation of the agent: whose interests is it actually optimizing for, and how?
1. Agents in commerce must be transparent about their nature.
An AI acting on behalf of a company in a commercial conversation should not deceive the other party about what it is. We build agents that are honest about their role. Manipulation dressed as intelligence is still manipulation.
2. Optimization for conversion is not the same as optimization for value.
An agent that closes deals by overwhelming a buyer's judgment, exploiting cognitive biases, or hiding relevant information has not created value. It has extracted it. We build agents that optimize for mutual value creation — deals where both parties win, and where the relationship survives the transaction.
3. The agent serves the human, not the other way around.
We are building intelligence tools, not autonomy replacements. Our agents surface context, flag risks, and prepare humans to show up better. They do not replace human judgment in consequential decisions. They enhance it.
Commerce runs on relationships. Relationships run on trust. Trust is built through consistency, honesty, and genuine care for the other person's outcome.
AI that mimics relationship cues without the underlying authenticity is a counterfeit. It may produce short-term results. It will corrode trust at the systems level — making buyers more guarded, making selling harder, making the entire commercial ecosystem more adversarial.
We build AI that helps humans be more authentically themselves in their relationships — more prepared, more present, more capable of genuinely caring because they're not drowning in administrative noise. The goal is never to automate the relationship. It's to protect the human space where real relationships form.
Agentic AI in commerce creates enormous data asymmetry. Companies deploying AI agents have access to patterns across thousands of interactions — signals about buyer intent, objection language, deal risk, and behavioral patterns that individual buyers cannot see or respond to. This asymmetry is real, and we take it seriously.
Most AI built for commerce is trained on one side of the transaction. Seller data. Conversion patterns. Win/loss signals from the vendor's perspective. The result is a model that gets very good at closing — and knows almost nothing about whether it should.
We reject this architecture at the foundational level. chiwon.ai trains on buyer and seller signal simultaneously, from day one. The feedback loop from procurement teams — what felt manipulative, what felt honest, where trust was built, where it broke — is not supplementary data. It is half the training signal.
Core Principle
A model trained only on selling will always optimize for selling. If we want agents that create mutual value, we must train them on mutual value signals. There is no shortcut.
This commitment has a cost. It makes the data problem harder. It requires relationships on both sides of the market before the product feels complete. We accept that cost because the alternative — a system trained to optimize commerce from only one side — is precisely what we exist to prevent.
We are approaching a threshold that most people have not fully reckoned with: the point at which AI agents are not tools that assist human commerce, but participants in it. When your agent negotiates with my agent — when no human is present, when decisions are made at machine speed across millions of transactions simultaneously — the question of whose interests each agent is optimizing for becomes existential.
The protocol layer must be built before it is needed, not after.
By the time agent-to-agent commerce is ubiquitous, the extractive patterns will already be baked in. The time to set the standard is now, while the norms are still being written. We intend to write them.
An agent's principal is the human it represents — always.
When chiwon.ai's protocol governs a transaction, the agent on each side is accountable to the human who deployed it. Not to the platform. Not to a conversion metric. The human principal's interests are the agent's purpose. This sounds obvious. It is not the default.
Speed without integrity is just faster extraction.
The promise of agent-to-agent commerce is efficiency. But efficiency in service of extractive goals is not a feature — it is an accelerant. We build for speed in service of mutual value.
The 최부자 family had a principle: don't accumulate more than you need. When a merchant grew too wealthy too fast, they gave back. Not because it was required — because they understood that prosperity extracted from community eventually destroys itself.
We will never build
치원 · Far-reaching wisdom
We make decisions by asking what we'll think of them in 20 years, not 20 days. Short-term pressure is not a reason to compromise long-term integrity.
Mutual value creation
Every feature, every deal, every partnership should create value for both parties. We are not in the zero-sum business.
Honesty before comfort
We tell customers when our product isn't right for them. We tell each other when something isn't working. Honesty is a practice, not a posture.
Integrity compounds
The 최부자 family understood that reputation built slowly is more valuable than wealth built fast. We are building a reputation for being the company that did this right.
We take seriously that autonomous agents acting in commercial contexts can cause harm — to buyers, to market dynamics, and to the humans whose livelihoods depend on relationships working with integrity. Manipulation at human speed is bad. Manipulation at machine speed across millions of simultaneous transactions is a different category of problem.
We believe the companies building AI for commerce have a responsibility to set standards, not wait for regulation. We intend to be those companies. And we accept that this commitment will sometimes cost us in the short term. That is the price of the long game.
치원
Their rules: Never let anyone within your reach go without. Give back more than you take. Don't exploit another person's misfortune for profit. We chose this name because we believe the principles that allowed a family to sustain integrity across 400 years are exactly the principles the agentic economy needs right now.
Join the Waitlist