Maya Proposes Briefs Before Drafting Them

Why Two Tiers

Building a full brief is not free — Maya runs SERP analysis, pulls Google Search Console data, profiles competitors, and generates the article outline. That's roughly $1–2 of platform cost per brief plus a quota slot. If a brief gets rejected, that cost is sunk.

Maya's two-tier autonomy fixes that:

| Tier | What lands | Per-event cost | When it fires |
|------|-----------|---------------|--------------|
| Propose | A topic + 2-sentence rationale in your inbox at /maya/proposals/ | ~$0.05 (LLM only) | When Maya isn't yet confident the insight matches your preferences |
| Commit | A full draft brief in your briefs list, same as today | ~$1–2 | When Maya's preference model scores the insight high enough (≥70/100) |

A brand-new tenant has no preference model trained yet, so every nightly insight lands in Propose. As you click ✅ Promote or ❌ Dismiss, the model learns. After ~30 signals the model trains, and high-fit insights start skipping Propose and going straight to Commit.

The Proposal Inbox

Visit Maya → Proposals (or /maya/proposals/) to see what Maya has surfaced. Each proposal shows:

• A topic title (e.g., "How to compare zone 2 training plans for runners")
• A 2-sentence rationale explaining why Maya chose it
• A link back to the source insight
• ✅ Promote to Draft — runs the full brief workflow (1–2 minutes, lands in your briefs list with a notification)
• ❌ Dismiss — closes the proposal and tells Maya you don't want this kind of topic

Pending, Promoted, and Dismissed tabs let you review prior decisions.

How Maya Learns from Your Clicks

Promote and Dismiss are the most precise training signals Maya gets. They're better than brief approval/rejection because you're rating the topic itself, before any drafting work happens. A brief might be rejected because the writing was off — that's noise. A topic dismissal is a clean signal that the topic isn't right for you.

After enough signals accumulate, Maya's nightly run will:

1. Rank insights by your preference model
2. Send high-fit ones (≥70/100) directly to the briefs list as full drafts
3. Send borderline ones to your proposal inbox for a quick review
4. Stop suggesting topic patterns you've consistently dismissed

Per-Night Caps

Two separate caps protect your budget and inbox:

• Max drafts per night (Commit tier) — default 2. Set it in Settings → Maya Autonomy. This is the expensive cap.
• Max proposals per night (Propose tier) — default 3. Cheaper, so the cap is higher.

Both caps respect a nightly cost ceiling — Maya stops dispatching once the budget is exhausted, even if caps aren't reached.

Disabling or Adjusting

In Settings → Content → Advanced, the Maya Autonomous Drafts toggle turns the whole nightly loop on or off. The toggle is gated to Professional and Enterprise tiers.

Granular controls — per-night cap on full drafts, per-night cap on proposals, nightly cost ceiling, and the tool whitelist — currently default to safe values and require admin assistance to adjust. Contact support if you want Maya to propose more or fewer topics per night, run with a different cost cap, or restrict her to a narrower set of tools.

If you'd rather review every topic before any drafting work happens, ask support to set Max drafts per night to 0 and Max proposals per night > 0. Maya will only ever propose, never auto-commit.