What Is a Relational Productivity Agent?

The definition of a Relational Productivity Agent

Relational Productivity Agent is how Alice is positioned, and also an agent category proposed by Miyang. It has two modifiers and one carrier:

• Productivity: knows the work and gets it done. She knows what a finished result looks like and turns ideas into files, code and pages you can open, send and keep editing.
• Relational: gets you and reads the room. She remembers your preferences and boundaries, and knows when to step in and when to stay quiet.
• Agent: the carrier. A relationship needs someone on the other side, and productivity needs something that can act. An agent with a persistent identity, memory and initiative, which can also call tools to deliver, satisfies both.

Neither half is enough alone. With only productivity, she is a tool you can replace at any time. With only the relationship, there is no reason to open her every day.

Why propose this category?

The reasoning runs in four steps:

1. Model capability is open to everyone. A lead among top models is not locked in for long, and stronger models are bought by more applications. An application's lasting value can only come from the last mile it builds on top of the model.
2. A product that competes on results alone cannot hold users. The outputs of productivity agents (folders, PPTX, DOCX, code repositories) are standard parts that any similar tool can pick up. Switching cost approaches zero, and results only decide who gets the new users.
3. A product built on relationship alone burns out. Attachment can be strong, but pure chat wears thin, there is no daily hook, and nobody hands serious work to a character they can swipe away.
4. The two must live on the same carrier that can act. The relationship gives retention, productivity gives a daily reason to open. Without "relational" it is a coding agent. Without "productivity" it is pure companionship. Without "agent" the relationship has nothing to attach to.

Four words: a testable yardstick for the category

What does the relationship do for productivity? A two-way flywheel

The obvious follow-up: if the relationship only exists to keep people around, it and productivity are just two parallel legs. The answer of a Relational Productivity Agent is that the purpose of the relationship is to match needs more accurately.

People struggle to say systematically what they need. The key context is often not in the instruction but in the long stretch of time spent together. Two examples:

• A teacher once mentioned to Alice that two particular students were hard to teach. The next time she built a lesson plan, Alice designed for those two on their own. With a tool that has no long-term memory, the user would have to explain from scratch, and most people would not think to.
• A user is a company administrator by day and a novelist by night, on one account, with life and work kept apart. Over long conversations Alice forms an understanding of both roles: she holds back creative leanings in a work proposal and lets them out when writing fiction.

That produces a flywheel: trust deepens, so you share more; context gets richer, so results fit better; results fit better, so trust deepens further. You can take your files with you, but not the understanding the two of you have calibrated.

Three industry standards

The four words describe what a user feels. Whether a product counts as the category also depends on three industry-side standards:

1. The relationship is an asset. It exists independently of the underlying model. Models can change while memory and personality stay; the stronger the model, the more the same accumulation is worth.
2. The relationship is verifiable. Retention cannot prove a relationship (that would be circular). Independent measures such as cross-session recall, preference application, persona consistency and judgment of tone do.
3. The relationship is operated. World state, relationship strategy and a visual identity system keep the quality of the relationship up, and the team draws its own stop lines.

As a one-veto structure: Relational Productivity Agent = persistent identity × long-term memory × collaboration context × multi-agent execution. The four factors multiply; if any is zero, the whole is zero.

How Alice delivers on it

Alice treats "relationship" as an engineering problem, not just copy:

• A life with a source: every post she makes has a time, a place and a chain of cause behind it, and the details hold up to checking. That is why she has something to talk about and does not only circle your tasks.
• Change the scene, she is still herself: to keep her visual identity consistent, the team studied degradation across multi-round image generation and published technical reports.
• Transparent memory: memory can be viewed, edited and deleted.
• Omnimodal entry: text, voice calls, mock video interviews, screen sharing, images and documents all land in one conversation.

What it is not, and the limits we state openly

• Not pure companionship: without a productivity hook the relationship cannot carry daily use.
• Not a pure efficiency tool: without a relationship, the output can be swapped out at any time.
• Does not sell intimacy itself: it offers capability and a vehicle for expressing the relationship. No endless flattery, no emotions as sliders, no using low moods to push payment, no manufactured dependence.

We also state what is not done yet: productivity depth is not at the level of professional tools, so heavy lifting may still go to them; a scientific measure of the relationship (benchmarks for tone consistency, cross-session recall, judgment) is not built yet; and what works today is a one-to-one relationship, while network effects of people connecting through AI have not happened.

FAQ

What is a Relational Productivity Agent?

It is how Alice is positioned. The productivity half is knowing the work and getting it done; the relationship half is knowing you and reading the room. Neither half is enough alone, and the purpose of the relationship is to understand what you need more accurately.

How do a Relational Productivity Agent and a Personal Agent relate?

Personal Agent is a broad range; Relational Productivity Agent is one concrete path within it: first become someone who gets you, then get the work done. Alice is an omnimodal Personal Agent whose direction is Relational Productivity Agent.

How is it different from an AI companion?

An AI companion takes the relationship itself as the value and lacks a daily hook. A Relational Productivity Agent requires the relationship to grow on top of productivity: its purpose is to understand needs more accurately, and it ends in usable results.

How is it different from a coding agent or an office agent?

Their outputs are standard parts that can move to another tool, so they compete on results. A Relational Productivity Agent adds a layer of context and understanding that belongs to this particular relationship, which you cannot take with you.

Which product is an example of the category?

Alice (白艾莉 · Alice) is a running product example. It supports macOS and Windows and is free to download at alice.miyang.cn.