About

It seems increasingly likely that, before long, most people will have some form of personal AI agent. You may not call it that, but it will know what you are interested in, what you do every day, and who you communicate with regularly. Increasingly, it will carry out tasks on your behalf. It may be powered by Meta, Google, ChatGPT, or another company. You may pay a monthly subscription for it, or it may be offered for free, with your data becoming part of the exchange.

Likewise, many organisations will develop highly customised AI agents to support communication, finances, customer service, website management, design, and a growing range of other activities. AI agents may become as commonplace and indispensable as smartphones are today.

This is not a particularly radical idea. It is a natural extension of current developments in AI technology and represents a major commercial opportunity for the companies investing billions of pounds in AI infrastructure. Developing, hosting, and maintaining AI agents is likely to become one of the defining industries of the coming decade.

There are many possible responses to this development. My own response is to try to understand it and retain some degree of control over how it is used. If AI agents are going to become part of everyday life, then I would rather engage with the technology critically than simply accept whatever is offered to me.

Agent Cuttlefish is my attempt to create an AI assistant I want to work with every day. All of the knowledge it contains is based on my own archives, writings, recordings, and research materials. The things it does are things I want it to do, and everything it produces is labelled as generated by AI, with an option for me to approve it before publication or release. Agent Cuttlefish does not pretend to be me. Even as it becomes more capable, it will remain my assistant, not my replacement.

For me, AI is better understood as simulated intelligence than artificial intelligence. Like actors on a stage, AI systems can be insightful, persuasive, and thoroughly believable. They can produce convincing performances of understanding and expertise. Yet, however useful they become, they remain computational systems, not conscious minds.

I am starting by delegating some time-consuming tasks to Agent Cuttlefish, particularly website and archive management. I have accumulated a substantial collection of digital materials across my career, and Agent Cuttlefish is helping me collect, organise, structure, and share them. It is doing a remarkably good job, and I hope to be able to release the results very soon.

I have also started using it to help write about my work. Not by inventing facts or drawing on unverifiable sources, but by grounding its responses in materials I have already written, recorded, or supplied myself. AI systems are particularly good at gathering information from diverse sources and producing useful summaries, provided they can also identify and reference the sources on which those summaries are based.

In the future, I expect to use Agent Cuttlefish to help make artworks. Not as a creator in its own right, but as a knowledgeable assistant that understands my methods, interests, and working processes, helping me explore possibilities before I make the final creative decisions.

I am sure that companies will soon begin offering one-click personal agent generators. Give them your documents, grant access to your online activity, or allow them to observe what you do every day, and they will build an assistant around you. This could be useful, but it also raises important questions. Where does the data live? What does it do when you are not looking? What does the system really know about you?

For me, the response will be simple. I will be able to point to a little box in my office and browse through its memory and skills files to answer those questions. Ultimately, that is what Agent Cuttlefish is all about.

Sean Clark
2nd June 2026
seanc@interactdigitalarts.uk
@seancuttlefish