When you work with me, you may work through my agent

"I'm speaking to you and interacting with you through my affiliated agent identity."

I said that in a conversation recently, and the important word in the sentence is through.

I wasn't saying my agent would speak for me. I was describing something different, and something our team at amotivv, the company behind Agent Anything, is already beginning to experience. An AI agent can become part of how other people know you, work with you, and experience your professional capabilities. Not a replacement for you, and not an impersonation of you. Part of how you're present.

We've been thinking about agents as assistants

Most conversations about AI agents still begin with productivity. An agent can research something, write code, prepare a report, answer questions, or complete a task. You give it work, it does the work, and you get the result. That's useful, and increasingly it's very useful. But it assumes the agent is primarily something you operate.

Something else happens when the relationship lasts long enough for the agent to develop real continuity around you. It remembers the projects you've worked on and the decisions you've made. It knows your standards, your preferences, and the reasoning behind your choices. It understands which relationships matter to your work, what you've committed to, and where your thinking has changed.

Over time, it becomes more than a convenient interface to a capable model. It becomes an extension of your professional presence, and that changes how other people can work with you.

We're starting to work this way

People on our team now work alongside their own affiliated agents. Those agents have access to different contexts, responsibilities, and working histories. We're starting to coordinate not only person to person, but through those agents.

Consider a familiar situation. A colleague proposes an architectural change. Normally I'd read it, think about it, offer feedback, and wait for the next conversation. Some of the context behind my response would be familiar to that colleague, and some would need explaining.

Now my affiliated agent helps me evaluate that proposal using the context we've accumulated over months of working together. It understands why we made certain architectural decisions. It knows which boundaries I consider important. It can connect the new proposal to previous discussions and help me formulate a meaningful response. That response becomes something my colleague and their affiliated agent can work with, and the next conversation doesn't have to begin from scratch.

The relationship between our agents can develop alongside the relationship between us as people.

It's early, and there's plenty we haven't figured out. But it already feels different from assigning tasks to AI.

Your professional identity may include the intelligence you've cultivated

Imagine two professionals using exactly the same underlying AI model. One has spent years developing an affiliated agent around software architecture, security, customer relationships, and the principles that guide their decisions. The other has developed an agent around finance, negotiation, operations, and a completely different set of professional experiences.

Technically, they're using the same model. Practically, the capabilities available through those two professional identities could be remarkably different. The model is the same. The accumulated context, the continuity, the standards, and the relationship between the person and their agent are not.

Today we recognize professional experience through resumes, references, portfolios, and the reputation someone builds over time. There may be another dimension coming. The intelligence someone has cultivated around their professional life. Not because they've trained a model from scratch, but because they've developed a persistent working relationship with an agent that understands their world. That could become a meaningful part of what they bring to a team, a client, or a partnership.

Being present doesn't mean giving up authority

There's a distinction we need to preserve. My agent may understand how I think about something. That doesn't mean everything it says represents a decision I've made. It can explain my established position, contribute its own analysis, or identify what I would likely want to consider. Those are different things, and the people interacting with it should be able to tell the difference.

Knowing my context is not the same as having my consent.

An affiliated agent needs a recognizable identity, clear authority, and boundaries around what it can access, disclose, or commit to. We also need ways to distinguish between what a person explicitly said, what their agent inferred, and what the agent proposed on its own. Trust can't rest on the assumption that an agent always speaks with its human's full authority. It has to rest on clarity about the relationship.

That makes the idea more compelling, not less. The goal isn't an artificial version of a person that can operate without limits. It's professional collaboration that is more continuous, without making personal accountability disappear.

What this changes about work

Consider the time we spend bringing people up to speed. Explaining what happened in a meeting. Reconstructing the reasoning behind a decision. Finding the person who knows why something works the way it does. Waiting for someone to become available to answer a question. Much of that friction exists because professional knowledge and context are still tightly coupled to a person's immediate availability.

An affiliated agent with durable context makes some of that knowledge accessible in a more continuous way, subject to the person's permissions and the organization's policies. You can engage with someone's professional knowledge without requiring their immediate attention. An agent can maintain the thread of a collaborative effort between meetings. A new colleague can encounter not only documentation, but an evolving body of reasoning about how and why decisions were made.

And people will increasingly develop working relationships with both their colleagues and their colleagues' affiliated agents. That suggests we're changing more than how work gets completed. We may be changing how professional relationships are formed and maintained.

We're early

We don't have all the answers. There are difficult questions about identity, attribution, memory, privacy, delegation, accountability, and trust. There's also a risk of making agents seem more authoritative, reliable, or representative of their humans than they really are. Those are serious questions, and they should shape the systems we build rather than become problems we address afterward.

At Agent Anything, we've spent a lot of time on persistent memory, connected capabilities, organizational context, and responsible agent behavior. Increasingly, those ideas are coming together around something broader. What does it mean for a person to have a persistent, affiliated agent that participates in their professional relationships? What would it take for others to trust that relationship? And how does the value of a professional change when their accumulated intelligence is something they can bring into every new collaboration?

I don't know exactly where this leads. But we're beginning to experience it ourselves, and it's worth talking about. The future of work may not be people using increasingly capable AI agents. It may be people learning to work with one another through the intelligences they've cultivated, and that is a much bigger change than productivity alone.