Who we are
first90.work is the marketing surface for Aparté, an onboarding-context service operated by Unsold Group. An Org — one company's curated knowledge space — is the customer. The Onboarding Owner (the person accountable for onboarding as the org grows) administers Aparté on the org's behalf. A Joiner is a new hire whose own AI companion connects to the org's context so they can learn how the company works while they deliver.
What we collect
Aparté holds four kinds of data on behalf of the Org:
- Context items
- Curated units of the org's institutional knowledge — docs, policies, runbooks, decision records, FAQs, people directory entries — each with a source, freshness date, audience scope, and tags. Orgs supply these directly, through connectors (Google Drive, Gmail, Linear), or through Public-Site Inference, which reads an org's public website to draft provisional context. Nothing enters the live store without an admin curating it.
- Plan & progress
- The Joiner's Onboarding Plan — 30/60/90 milestones, tasks, required reading, intro meetings — and progress recorded against their Individual JD.
- Reflections & narrative
- Session bookends, elicited reflections, and the questions a joiner logs during real work. This is developmental data about how onboarding is going, not a performance record.
- Query telemetry
- Each context read is logged with the joiner's role and tenure-week so Aparté can adapt teaching depth and help the Onboarding Owner see where onboarding needs work. Telemetry is about the fit between question and context, not surveillance of the person.
Visibility and the wall on use
Aparté's handling of developmental data follows a single principle: symmetric visibility, with a wall on evaluative use. A joiner's narrative and developmental data — reflections, session bookends, logged questions, assessed proficiency, org knowledge, onboarding progress, and assistance provenance — is visible to the joiner and to the org's designated onboarding owners and managers, and the onboarding prompt tells the joiner exactly that. The joiner can inspect the evidence and confidence behind any inference and correct it.
This is the disclosure the joiner sees in their onboarding prompt, delivered verbatim:
“What you ask, log, and reflect here is visible to you and to your org's onboarding owner. It is used to teach you better and improve onboarding for the next person — never for compensation, ranking, promotion, or performance evaluation.”
The wall is on use, not on sight. Aparté may use this evidence to adapt teaching and to help HR or managers target coaching. It is never a bonus, ranking, promotion, or punitive performance instrument, and Aparté does not collapse it into an opaque global score. No surface renders this data inside anything compensation-, ranking-, promotion-, or punishment-shaped.
Retention
Context items persist for as long as the Org keeps Aparté active; admins can edit, re-scope, or delete any item at any time, and stale items surface for review rather than lingering silently. Plan, progress, reflection, and telemetry data are retained for the duration of a joiner's onboarding and a limited period afterward so the Org can improve onboarding for the next person. When an Org closes its account, its data is deleted or returned on request within a reasonable period, subject to legal retention obligations. A joiner may request correction of developmental inferences at any time (see Your rights).
Subprocessors
Aparté relies on a small set of infrastructure providers who process data on our behalf:
| Provider | Purpose |
|---|---|
| Convex | Backend database and application hosting — the org's context store, plans, progress, reflections, and telemetry. |
| Anthropic | AI model provider powering Aparté's context reasoning, teaching adaptation, and Public-Site Inference. |
| Firecrawl | Web retrieval for Public-Site Inference, which reads an org's public website to draft provisional context. |
A joiner's own AI companion (for example, Claude, ChatGPT, or Codex) is chosen by the joiner or org and connects to Aparté over MCP; that tool is governed by its own provider's terms, not this policy.
Your rights
Joiners have a symmetric view of their own developmental data, including the underlying evidence, confidence, and assistance provenance behind any inference. A joiner can:
- See what has been recorded about their onboarding and how confident an inference is;
- Correct a mistaken inference, so the record reflects reality;
- Update personal strengths and learning preferences directly (changes to mandate, authority, or decision boundaries require manager approval);
- Ask the Onboarding Owner to export or delete their data.
Because Aparté is administered by the Org, requests to access, correct, or delete data are handled together with the org's Onboarding Owner. We help orgs meet these requests.
Contact
Questions about this policy or a data request? Email support@first90.work. See the support page for details.