The question comes up after almost every keynote now. Someone waits until the room has thinned out and asks, half seriously: should I build an AI version of myself?
My answer is a question back. Does your work have two layers, one that repeats and one that only you can do? If it does, an AI twin deserves a place on your list, and building even a small one will teach you more about AI than another year of briefings. If you cannot yet tell your two layers apart, that is the real work, and it costs nothing to start.
I can offer more than a theory here. I built two AI twins in 2026. One coaches. One runs my front desk. This is what building them taught me, including the mistake that cost me a lead.
The Two Layers Test
Every senior role I have looked at closely, my own included, contains the same hidden split.
There is the layer that repeats. The same questions answered again and again. The first contact. The routing of a request to the right place. The follow-up that never happens because the day is full. From the inside it rarely feels like repetition, because each instance arrives dressed as something new, with a real name and a real deadline attached.
Then there is the layer only you can do. The relationship. The judgment call. The conversation where someone needs to trust you before anything can move.
For an executive, the repeating layer rarely looks like admin. It looks like the same strategy question answered for the fourth stakeholder this month. The same context explained to every new advisor. The briefing that precedes every decision meeting, rebuilt each time from the same sources. None of it is beneath you. It is simply not the work your judgment is for.
The test is simply this: which parts of your work would look interchangeable if you recorded yourself doing them three weeks in a row? Those parts can go to a twin. The parts that depend on who you are, what you notice, and what you decide cannot, and should not. Not because the technology is unready, but because handing them over removes the reason people come to you at all.
Most executives I work with across Switzerland guess that their repeating layer is small. Then we write down one real week, and it is not small. Nobody sees the front desk inside their own role until they look for it.
What I Learned Building Two
The coach twin came first. CoachTünde coaches leaders every day, in their pocket, on my frameworks and my material: career questions, leadership questions, the thinking a full calendar keeps postponing. She works for people who want coaching from an AI and know that is exactly what they are getting.
The front desk twin came second, and she taught me more. She lives on my website, built on ElevenLabs and Claude, trained on my keynotes, my articles and the way I actually work. She answers in English, German, Italian and Hungarian, in chat or in voice, at three in the morning if that is when you think of your question. Her job is deliberately narrow: she knows my speaking, my coaching and my advisory work, and she guides you to the right one. She does not coach. That is her sister's job.
Keeping the two apart was the most important design decision, and I nearly got it wrong. My first instinct was one twin that does everything. A front desk that also coaches does both jobs badly: the event enquiry gets a coaching session, the coaching client gets a receptionist. Scope is not a limitation you reluctantly accept. Scope is what makes a twin trustworthy.
The second lesson cost me a lead. In April, someone who had seen me speak at an event in Vienna asked my early twin for the slides. She answered politely, and the conversation ended there. No email, no follow-up, nothing I could act on. A twin without a designed next step is not a front desk; she is a very friendly dead end. Every conversation now ends somewhere deliberate: a booking link, a resource, or a good reason to leave an email address that gives the visitor something real in return.
Where the Judgment Stays
Here is the part I did not expect. Building the twins changed how I evaluate every other AI decision, in my own business and in the organisations I advise.
Before, my views on AI were assembled the way most senior leaders assemble them: from briefings, conference talks and other people's conclusions. Building something, even something small, replaced borrowed opinions with experience. I now know what training material actually does to an answer. I know where the line runs between capable and confidently wrong. I know how much design hides behind the word "just", as in "the AI can just answer that". A vendor pitch sounds different once you have built a small version of what they are selling.
And one boundary became very clear. My twins do the work. The judgment stays with me. I read every conversation my front desk twin has, and where a question deserves a human, that person hears from me directly. She does the answering; I decide what matters. The day I stop reading is the day I have not delegated the work but surrendered the judgment, and with it the relationship.
That is the line I would give any executive considering a twin of their own, or approving one for their organisation. Decide in writing, before anything is built: which layer does she get, and which layer stays yours. If that question has no answer yet, the problem is not the technology.
The same split scales up. When a vendor or an internal team proposes an AI system, the two layers give you the first question worth asking: which layer of whose work is this taking over, and who keeps the judgment over what comes out? A surprising number of business cases have no answer to it.
You do not build an AI twin to replace yourself. You build one to be more yourself, more often. Mine started as a front desk; she freed the hours, and the judgment never left the building. If you want to see what that looks like in practice, she is at www.thechangerepublic.com/ai-twin, and she will introduce herself.
Private AI Advisory for executives and senior leaders at The Change Republic, working with leaders across Switzerland and Europe who want to build their own AI judgment before they sign off on everyone else's. Find out more at www.thechangerepublic.com/private-ai
Tünde Lukacs is an executive coach, AI advisory specialist, and founder of The Change Republic. She works with senior leaders on AI literacy, leadership development, and strategic decision-making across Switzerland and Europe.