What is a digital twin?

You’ve probably heard the term. Here’s the plain answer: no jargon, just the idea and why it’s suddenly possible.

A digital twin is a living version of a person, one that knows what they know, sounds like them, and can hold a real conversation when they can’t be there.

First, what it isn’t

It’s not a clone.

It’s not a profile, it’s not a recording, it’s not a chatbot on your documents.

A digital twin is not…

  • A robot or a copy of a person
  • A video or a voicemail that just plays back, frozen the moment it was made
  • A generic chatbot that answers like everyone, from whatever it can find on the internet
  • Something that replaces the real person

It is…

  • Built from a real person's knowledge and voice
  • Able to answer new questions, in their way
  • A specific someone, their taste and judgment
  • An extension of them, always in their control

It is a context graph.

Most tools index files and retrieve the closest match. A twin works differently: it follows how one thing led to another.

Every node carries the story of how it got there.

Follow the orange line above: a conversation influenced a decision, that decision led to an outcome, the outcome taught a lesson, and the lesson now informs a future decision. Each arrow is an edge, the graph doesn’t just store “decision” as a fact. It stores why the decision happened and what it led to next. That connective tissue is the tacit knowledge that usually walks out the door when a person does.

The two kinds of knowledge

Search solved half of it. The graph holds the rest.

Explicit · already searchable

What’s written down

Lives in documents, emails, wikis, chat. It’s indexable, and tools already retrieve it well.

“What’s our refund policy?” → find the doc
Tacit · what a twin captures

What lives in your head

Why a decision was made, which exception to make, the pattern an expert recognizes on instinct. It isn’t in any file, it’s in the edges.

“Should we make an exception here?” → traverse the reasoning
How it answers a question

Grounded, not guessed.

Every reply is anchored to your knowledge and shaped by your rules, so it sounds like you and stays honest.

01
Input · text or voice

Understand the question

The question is interpreted in context, who's asking, what they've asked before, and what they're really after.

02
Retrieve · graph traversal

Pull the right context

Instead of a blind keyword match, it walks your context graph for the nodes and edges that actually bear on the question.

03
Reason · grounded generation

Reason over it

A reasoning model composes an answer grounded in what was retrieved, applying your judgment, not the internet's average.

04
Constrain · guardrails

Check it against your rules

Guardrails filter for tone, scope, and safety. Outside what it knows, it says so and offers to bring in the real you instead of making something up.

05
Respond · in your voice

Answer as you

The reply comes back in your phrasing and tone, as text, spoken voice, or a talking avatar.

06
Learn · owner review

Improve from your review

You review, correct, and approve. Each edit refines the graph, the twin is curated by the expert, not left to drift.

Interfaces & capabilities

One graph. Every surface.

The same twin is reachable through whatever channel fits, each backed by the same grounded core.

Text

Type a question, get a grounded written answer, any hour.

streaming chat · citations to source

Voice

Speak and listen, in a voice modeled on yours.

speech-to-speech · real-time streaming

Avatar

A talking, on-brand face that makes it feel present.

lip-synced video · voice-driven

Rehearsal

Build a twin of who you'll face and practice the real conversation first.

role-play mode · scenario prompts

Embed & API

Drop the twin into your site or product with a snippet.

web widget · SDK · webhooks

Human handoff

The real you steps into the same thread when it matters.

live takeover · no restart

Different surface, same twin. Same graph, same reasoning, same guardrails, every time, and none of it leaves your control.

“But is it really me?”

Yes, because your twin is simply you, just digital. Your twin knows what you teach it. It learns from what you tell it, what you share with it, and how you guide it. The more you train it, the more it becomes like you. It gets better as you spend time with it, because everything it knows comes from you. It’s yours to build, shape and make your own.

The rest is easy: it’s just you.

Building your own digital twin takes about ten minutes.