The AI That Represents You: Inside Zinley’s Bet on the Personal Extension
Focus Tips:
- Zinley is pitching a different category from the AI copilot: an AI that represents you — with its own name, phone number, email inbox and always-on computer — rather than one that helps you type faster.
- The product’s most consequential design choice is disclosure. It never pretends to be you. Every call opens with “this is Nova, Ha’s AI assistant,” and the person on the other end can look up whose Zinley it is.
- For executives, the real question is not “can it do the task” but “what am I willing to let it decide” — and Zinley builds that boundary into the setup.
Every productivity tool of the last three years has made the same promise: you will get more done. Zinley, a young company that spent its early life under the name Orion, is making a stranger one. Its pitch is that you can get things done without being there at all — and that the entity doing them will be honest about the fact that it isn’t you.
The company’s homepage puts it in a single line: “Every AI so far made you faster. You still had to be there.” What follows is a product built around a simple but unusual idea. Your Zinley is not a chat window. It has its own name, its own phone number, its own email address and a computer of its own. It compares the twenty suppliers, then calls the three worth calling — as your Zinley, never as you.
What a “personal extension” actually is
Zinley calls this a personal extension, and the phrase is doing real work. A copilot sits beside you inside an app you are already using. An extension goes where you cannot. It answers the phone at 6:40 PM when you are driving. It replies to the buyer at 11:40 PM in the buyer’s language. It gets CC’d on an intro thread and books the meeting while you get on with the day.
Three things make that possible, and the company lists them plainly:
- Its own memory. It keeps a record of the people in your life — who they are, how you work with them, and what is still open with each of them — updated after every call, so you never explain the backstory twice.
- Its own number. A real phone line and inbox that belong to the assistant, not to you. It calls and answers under its own name.
- Its own computer. An always-on machine of its own, plus any device you approve, so it can work inside your apps while you are somewhere else.
The list of “twelve things to ask it” reads like a small operator’s week: compare twenty suppliers, read forty PDFs, rename and file 400 documents, clear tonight’s inbox, screen two candidates, answer my customers, book the demo, handle the insurance call, deploy the branch, make the launch video, build the campaign, send a file off my Mac. It works across Gmail, Slack, Google Calendar, Notion, Linear, GitHub, Drive, Figma, Stripe, iMessage and Discord, and it reaches you by web, mobile app, email, phone, text or Telegram.
The design decision that matters most: it never claims to be you
Plenty of tools can now place a phone call with a synthetic voice. Most of the anxiety around them comes from one question: does the person on the other end know? Zinley’s answer is unambiguous, and it is baked into the product rather than left to a policy page.
In the company’s own demo, the assistant opens a call like this: “Hi, this is Nova, Ha’s AI assistant. I can check your order or book time with Ha — she reads everything I write down.” The assistant has a card that says whose Zinley it is. It has a lookup page. And it operates inside a three-line permission model that the owner sets up front:
- May decide: answer questions, check an order, book time.
- Must ask: refunds, discounts, anything signed.
- Always: every call is written back to the owner, in the owner’s words.
For anyone who has sat through a compliance review of an AI rollout, that framing is the interesting part. The disclosure is not a disclaimer; it is the identity. And the “stops at money and comes to get you” rule turns the hardest governance question — where does the machine’s authority end? — into a setting the user configures in a few minutes and can change later.
Four weeks it is built for
Zinley does not describe itself in terms of features so much as in terms of whose week it is meant to absorb. Four scenarios anchor the pitch:
The online seller. Amazon, TikTok Shop, and a “where is my order” at 11:40 PM. The assistant answers buyers in their language at their hour, checks the order, sends tracking, starts the return, and fills in the order sheet before the owner wakes up. The company’s own tagline for the scenario: nineteen messages overnight, three left for you.
The small-business owner. The phone rings while you are already driving. It picks up, knows who is calling, books, reschedules, quotes from your price list — and stops at money. The voicemail becomes a booking; you call nobody back.
The hiring manager. Four hundred applicants and time to call thirty. It calls every one with your six questions, says it is an AI and offers a human instead, and sends a shortlist with the reasons attached. You read one comparison, not four hundred résumés.
The creator or brand. Ninety brand emails a week, four of them real. It asks about budget, rights and timeline in the first reply, sends the rate card, holds the call slot, and chases the invoice on day thirty-one.
What links the four is not the industry but the shape of the problem: high-volume, low-stakes-until-suddenly-not, and happening when the human is unavailable. That is exactly the work that has historically been done badly, late, or by an underpaid assistant.
The email handoff, and why it is the tell
The most mundane use case on the site may be the most persuasive one. You are on an intro thread with a prospect. You reply, “Putting Nova on this to find us a time,” and CC the assistant’s inbox. It asks for availability and time zone, proposes Tuesday 15:00 Berlin, sends the invite to both parties, and the meeting lands on your calendar. You are copied on every step, in your tone.
This is the tell because it demonstrates the product’s actual claim — not that it can do a task, but that it can hold a thread with a third party, in public, under its own name, without you supervising each turn. Scheduling is the low-risk proving ground for that behavior; supplier negotiations and candidate screens are the same behavior with the stakes turned up.
What executives should ask before they hand it a phone number
Zinley says it is free to start, with a number and inbox included, and that setup is three steps: the assistant gets its number and inbox; you tell it who matters and what it may decide; then you simply ask, in any language. That is a low barrier. It is not, on its own, a reason to deploy. Leaders evaluating this class of tool — Zinley or its inevitable competitors — should be asking a short list of questions:
- Who does the record belong to? Zinley says the memory it keeps about your people is “yours alone, and never shared with the people it is about.” Confirm what that means for retention and export.
- Where is the money line? Decide, before the first live call, which actions are “must ask.” Refunds, discounts and anything signed are the company’s own defaults; your list may be longer.
- How does it disclose? Listen to a real call. The opening sentence is the whole trust model.
- What is the audit trail? “Every call written back to you, in your words” is the promise; check how that log looks in practice and who inside your company can see it.
The bigger bet
The category Zinley is trying to name — the AI that represents rather than assists — will be judged less on model quality than on manners. Can it introduce itself honestly, know when to stop, and report back accurately? The company has clearly decided those are product features, not policy footnotes. If it is right, the next wave of workplace AI will not be measured by how much faster it makes you, but by how many rooms you no longer have to be in.
Zinley is at zinley.com; the product formerly went by Orion. Details in this article are drawn from the company’s published materials as of August 2026 and were not independently tested by Executive Central Weekly.
