An AI clone for strategy consultants is not a deck generator. It is the thing that closes the post-interview loop.

No. The clone in this product is operational, not generative-media. It does not produce a 30-page deck from a prompt. It runs the back office around the deck: it reads the Zoom transcript from your last interview, drafts the synthesis section in your engagement fact pack the way your last eight write-ups read, logs the decisions on the right HubSpot contact, and queues a follow-up email in Gmail. The deck itself is still your work product, written in PowerPoint or Google Slides the way your firm writes decks. The clone closes the loop around the deck.

Those are reasoning surfaces. You bring the data in, you ask a question, you read the answer, you copy the answer into your fact pack by hand. They do not touch the engagement Drive folder, the HubSpot contact, or the Gmail thread that follows the interview. The operational clone does. It runs in the same browser sessions you already have open, files the synthesis section under 01 Interviews in the right engagement folder, and updates the contact properties on the right HubSpot record. The fact pack does not migrate, the data stays where it lives.

Your Mac. The principle that drives this is in src/components/architecture.tsx line 46 of the live marketing site: 'Client files, emails, contracts, and transcripts never leave your computer.' Clone Memory holds your rituals, your synthesis style, your interview structure, and your engagement folder layout under ~/.clone/memory/ on local disk. You can route the planner to a local Llama 3 on an M-series Mac for fully on-device operation, or to a cloud LLM for faster reasoning. The Computer Agent and Memory remain local in either configuration. For boutique strategy work where the engagement letter has a confidentiality clause, the local-only path is the default-on path.

Three to four observed runs of the same shape are usually enough for the pattern-mining UI to fire. Eight is the comfort threshold where the clone reliably preserves the small judgment calls (quote treatment, open-question cap, headline takeaway position). If you write only one or two synthesis sections in a quarter, the clone will still help, it will just propose a less specific draft and ask more questions on the first run.

It samples direct quotes that match your prior italic-treatment pattern, decisions where a stakeholder used decision verbs ('we will', 'we are going to', 'we have signed off'), and next steps where someone said 'I will' or 'we will' with a clear subject. It skips small talk, scheduling chatter, and side conversations that are not on the engagement scope. You see the diff before any of it lands in the fact pack.

No. The default is review-then-approve. The drafted follow-up lands in Gmail drafts, queued for one click. You can opt into auto-send for a specific ritual once you trust it, but the out-of-the-box behavior is 'draft and queue, never send'. For senior interviewees in particular, where the cost of a wrong follow-up is your reputation, the review queue is a feature, not a friction.

Both. The Solo plan is $49 per month for one consultant. The Boutique plan is $129 per seat per month for teams of two to ten. On a boutique plan, each consultant runs their own clone with their own Memory layer; the synthesis style is per-consultant, not per-firm. Two consultants on the same engagement can run different rituals on the same fact pack folder, which usually matches how a senior associate and an engagement manager actually divide the work.

Both, once you save the ritual. The interview-synthesis ritual is the most common one for strategy work because it fires multiple times per week. The partner pre-read ritual fires once a week and reads the engagement fact pack, the prior week's open questions, and the hours-tracker rollup. Each ritual is named, lives as a markdown file, and is editable by hand if you want to tighten the prompt.

Three. First, the clone is only as honest as your last N runs. If your last eight syntheses were rushed, the cloned ritual will be rushed; the fix is to consciously run the next two or three the way you actually want to. Second, a brand-new engagement type (a regulatory diligence after a career of consumer growth work) gets nothing useful from the clone for the first few interviews; you write those by hand. Third, the cost of letting auto-send run on a follow-up to a senior interviewee you have not reviewed is your reputation, so the default review queue is the path. Use it.

The pricing comparison row on the live marketing site puts the closest non-AI alternatives at $3,000 to $6,000 a month for a virtual assistant who could draft synthesis sections under your supervision. The Boutique plan is $129 per seat per month. For a two-person team that runs eight interviews a week and reclaims roughly five hours per consultant per week of synthesis time, the math closes inside the first engagement. The numbers metric block on this page lays out the eight, five, and 129 inputs.