Our work often involves cases that do not fit neatly into familiar procedures. We study the relevant law, develop a position and prepare supporting documents. The colleague discussing the case with an official or counterparty then needs to understand the reasoning well enough to explain it and respond to questions.
This is why we began bringing Seraphim, my personal AI assistant, into team discussions. Colleagues can add facts, question an answer and ask for another explanation. I remain involved in assessing the reasoning and deciding what we act on.
In this exchange, Khun Pin raises a Thai work-permit question, clarifies the signing problem and asks for short steps explaining what to do and how.
The recording illustrates the interaction. It does not establish the legal correctness of the AI’s proposal or the outcome of the application.
Read the English transcript
PIN: I would like to ask about the work permit. Do we need a Thai person involved, such as a director or shareholder? I encountered a requirement at the employment office. The officer said we needed one. Could you check how we can address this, or what documents we could submit instead?
AI: When they said a Thai person was needed, did they mean a shareholder, director or employee?
PIN: A Thai director and shareholder. The four Thai employees required are not a problem.
EUGENE (OFF CAMERA): Our focus is specifically the work permit, the visa, and a Thai shareholder or director.
PIN: But I am asking about a case where this foreigner is the only director, and no other authorised person or other Thai person can sign for him. What can we do?
AI: Khun Pin, the answer is to use a power of attorney in the company’s name.
PIN: Please give me short steps, each explaining what to do and how. Short and easy to understand.
AI: Understood. One: draft a power of attorney in the company’s name, so you can sign and submit the application for the company. Two: take the draft and company certificate to the receiving office to check the scope of authority first.
How the team shapes the exchange
Pin supplies the specific circumstances.
The assistant asks whether the requirement concerns a shareholder, director or employee. Pin clarifies the roles, then explains that the foreigner is the sole director and that no one else can sign for him.
I keep the scope explicit.
I bring the discussion back to the work permit, visa and the question of a Thai shareholder or director.
Pin asks for a different answer format.
“Please give me short steps, each explaining what to do and how.” The assistant responds with two proposed steps.
What interests me here is how a personal AI assistant can become a shared resource within a working conversation. A colleague supplies the circumstances and asks directly for an explanation she can work with. My responsibility is to keep the discussion focused and decide what we act on.
Recording notes and media enquiries
This is first-party footage supplied by Eugene Prudchenko, a participant in the meeting, and edited from a longer working session. The English subtitles were reviewed and approved by Eugene and Khun Pin.
For source footage, context or an interview about this working practice, contact Eugene Prudchenko.
Original recording and context
The original excerpts introduce the broader practice behind this session: bringing an AI assistant into a working discussion while people retain judgment and responsibility.
I do not only use AI to prepare for meetings or summarise them afterwards. I also bring it into the discussion while we are working out what to do.
These edited excerpts show parts of my working meetings: an operational question, an AI response, and a request to address the people in the room in Thai. I direct the exchange rather than handing the meeting over to the assistant.
The interesting part is not that AI can speak. It is where its contribution sits: inside the discussion, before we settle on a next step.
In one excerpt, I use Profiler Plus, my working instruction for structuring an explanation around the recipient's concerns, constraints and next decision. It is not a separate AI model. Seraphim is also a name for my assistant, not a model I developed.
The arrangement is simple. People provide the situation and constraints. AI contributes an analysis or explanation. People question, redirect and decide. Important factual claims still need checking before anyone relies on them.
This is one visible part of the Compound Operator approach I describe elsewhere: a person working through an AI system with explicit context, rules and responsibilities. The Personal AI Layer explains how I organise that context and those boundaries.
The closing artwork uses Homo Cyberis for the broader idea of a sustained human–AI working relationship. Human and AI remain distinct. The aim is to extend human capability without surrendering human judgment.
The human still decides. But no longer thinks alone.
Apply the approach to your work
Start with a workflow, the context it needs, and a person responsible for reviewing the result.