Unclassified

25 — Twin Story

Status: CANONICAL · Commercial OS
Owner: Product Marketing
Last verified: 2026-07-21

Purpose

How to tell the Twin — the concept most at risk of being misheard as "chatbot." This story exists to make the difference unmissable.

Definition (verbatim): A person's governed AI counterpart that works alongside them on the same workbench — it prepares, suggests, and proposes; it never owns truth and never acts without approval.

1. The Story

Everyone has met the other kind of AI. It lives in a tab. It knows nothing about your morning until you paste your morning into it. It answers with confidence, sometimes about things that aren't true. It forgets you by tomorrow — or worse, remembers something it invented about you. You learned its real place: emails, summaries, nothing that matters. The quiet demotion.

The Twin is built from the opposite premises. It is yours — one Twin per person, scoped by your role and authority, at your bench. It already knows the context, because it shares your workbench and reads the same Spine of truth you do — it never asks you to re-explain your own work. It prepares the hard parts: the meeting brief before you open it, the analysis behind the risk signal, the draft grounded in what was actually said. It shows its evidence — every claim links to sources; and when it has nothing grounded to offer, it is silent. No filler, no vibes, no invented renewal dates.

And the line it cannot cross, by architecture: the Twin proposes; you approve. It cannot write to organizational memory — TruthLayer requires your approval. It cannot execute a consequential action — governance requires your approval. It is a colleague with excellent preparation and no authority — which is exactly why you can finally give it consequential work.

The experience promise, verbatim: "My work already knows its context. AI prepared the hard parts. I decide."

2. What the Twin Does (grammar, not feature list)

NOTICE what changed and matters · EXPLAIN this work from its context · RELATE this to what the org knows · PREPARE the brief/analysis/draft · SUGGEST the next step · PROPOSE the reviewable action.

And what it never does (doctrine): own truth or memory · own identity, policy, or authorization · execute unapproved · occupy a permanent chat destination that competes with the work.

3. The Hard Distinctions (say these before the prospect asks)

They think	The correction
"So it's a chatbot on my data"	A chatbot is a destination — you go to it and narrate. The Twin is at the bench: its help lands inside the work, grounded, with evidence.
"Like a copilot?"	Copilots are sidecars to one app and see that app. Your Twin works over your organization's assembled truth — every connected system, one context. "Copilots visit your apps. IntegrateWise is where work lives."
"An autonomous agent then"	The opposite bet. Agents seek autonomy; the Twin's contract is preparation + proposal, with your approval gating anything real. That contract is why it gets real work.
"Will it replace my team?"	It's one-per-person, not one-instead-of-person. The accountable human is the design center — the Twin amplifies them.

4. Trust Mechanics (why belief is not required)

Evidence chains on interventions · TruthLayer approval on memory · action approval + audit trail · confidence honesty (silence over speculation) · model-agnostic (swap the AI; the Twin's knowledge stays, because knowledge lives in your Spine).

Trust line, verbatim: "Every AI remembers. Only IntegrateWise remembers the truth."

5. Language Rules

Always "your Twin" (possessive, singular, capital T). Never "the AI," "the bot," "our assistant," "agents."
Verbs of preparation and proposal, never verbs of autonomous completion ("your Twin drafted" ✓ · "your Twin handled it" ✗).
The human decides in every sentence where something consequential happens.

KPIs: Approved Twin Actions per seat/week (north star); intervention acceptance rate; "chatbot" misreadings in discovery (should trend to zero as messaging lands) (73).