Você deixaria uma IA negociar o seu Salário?
Read the Portuguese translation →
Author: Sandeco · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Você deixaria uma IA negociar o seu Salário? · Channel: Sandeco · 02:08 · Watch on YouTube
Topics: AI agents, Enterprise AI applications, Claude, Markets and economy
Summary
Section titled “Summary”Anthropic is consulting law firms about how AI agents could negotiate on behalf of people, including prices, complaints, contracts, and salaries. In an internal experiment involving 69 employees, agents bought and sold personal items without human intervention, producing 186 agreements in a round involving real exchanges. More capable Opus agents secured better terms than Haiku agents, though participants using less capable models did not clearly notice the disadvantage. The experiment was small and does not prove how the future economy will work, but it raises questions about how AI quality could affect people’s bargaining power.
Key points
Section titled “Key points”- Anthropic is exploring how AI agents might negotiate for people, consulting with law firms on the topic.
- Possible uses include negotiating prices, resolving complaints, and seeking better contract terms.
- Anthropic ran an internal market experiment involving 69 employees and agents buying and selling personal items.
- Once negotiations began, humans did not intervene in the process.
- A round involving real exchanges produced 186 agreements totaling over $1,000.
- Participants represented by the more capable Opus model achieved better terms than those using Haiku.
- In one example, the same bicycle sold for $65 through Opus and $38 through Haiku in different rounds.
- Participants using less capable models did not clearly perceive their disadvantage.
- The small employee experiment does not establish how the economy will work, but suggests AI quality could affect bargaining power.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | AI agents as negotiators | The video introduces the possibility of AI negotiating salaries and other matters, and describes Anthropic’s consultations about agents negotiating for people. |
| 00:32 | Anthropic’s internal negotiation experiment | The video explains how 69 employees used agents to buy and sell personal items without intervening once negotiations began. |
| 01:05 | Different models, different outcomes | The experiment produced 186 real agreements, and Opus agents achieved better terms than Haiku agents in the model comparison. |
| 01:37 | What AI negotiation could mean | The video cautions that the experiment was small while asking whether AI capability could influence people’s bargaining power and future choices. |
Quotes
Section titled “Quotes”Would you let artificial intelligence negotiate your salary? — 00:00
Once the negotiation began, humans no longer interfered in the process — 00:32
the same bicycle was sold for $ 65 by Opus and $ 38 by Haiku — 01:05
the quality of the artificial intelligence could influence our bargaining power. — 01:37
would you entrust this responsibility to an artificial intelligence agent? — 02:08
Concepts and entities
Section titled “Concepts and entities”Anthropic, Claude, Financial Times, Opus, Haiku, agente de IA
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Would you let artificial intelligence negotiate your salary? And what if the company on the other side also used an AI to negotiate with you? That is what Anthropic, the creator of Claude, is currently doing as it prepares for this scenario. According to a Financial Times report, the company is consulting with law firms to understand how artificial intelligence agents should negotiate on behalf of people. We are talking about discussing prices, resolving complaints, and even seeking better terms in contracts. And you can
00:32 see where this is going, right? It’s an AI defending your rights on one side and another AI trying to negotiate with you on the other. And Anthropic has already conducted an experiment that shows this deserves some attention, as it created an internal market with 69 employees. Each participant was given an agent to buy and sell personal items . The agents posted listings, made offers, responded to counteroffers, and closed deals. Once the negotiation began, humans no longer interfered in the process, and in the round that
01:05 turned into real exchanges, there were 186 agreements made, totaling over $ 1,000 in the process. But the most interesting result of all appeared when they set different models to negotiate. Those represented by the more powerful Opus, that more robust configuration, achieved better terms than those using the Haiku model. In one example, the same bicycle was sold for $ 65 by Opus and $ 38 by Haiku in different rounds.
01:37 And do you know what is most curious? The participants with less capable models did not clearly perceive the disadvantage. But stay calm, everyone, because this was just a small experiment among Anthropic’s own employees, and it doesn’t actually prove how the economy will work going forward. But it does raise a huge question, doesn’t it? If agents start negotiating our purchases, contracts, and salaries, the quality of the artificial intelligence could influence our bargaining power. This is true, because today you only compare the price of a subscription based on how
02:08 much you will use it daily, right? But tomorrow, you might need to compare how well it can negotiate for you. So, would you entrust this responsibility to an artificial intelligence agent?