You decide how much of the work Maya owns. Levels run from L0 (Maya is off) to L5 (Maya can publish on opted-in surfaces autonomously). You always retain ownership — you choose the level, and Maya operates at that level until you change it or until quality drift triggers an automatic step-back.

The level model

Level

Maya's role

Surfaces unlocked

Tier

L0 Off

Nothing runs — no research, no recommendations, no proposals

None

All

L1 Research & Recommendations

Researches and surfaces recommendations in your Strategy Hub; you pull, decide, and create

Strategy Hub (read-only intelligence)

All

L2 Propose

Maya proactively proposes opportunities; one-click promote

Briefs (propose tier)

Starter+

L3 Brief Autopilot

Full briefs autonomously, lands as draft. Turning a brief into an article is your call

Briefs (commit tier)

Professional+

L4 Content Autopilot

Briefs + their articles + social drafts autonomously, all land as drafts for your approval

Briefs + articles + social drafts

Enterprise

L5 Surface-Specific Distribution

Auto-publishes on the surfaces you opt into (per-surface enrollment)

Blog, and sharing those articles on social

Enterprise

Where Maya's automation begins

There are two lines that matter, and they're nested:

  • L1 → L2 is where Maya gains agency. L0 and L1 are read-only platform intelligence — the platform researches and surfaces recommendations on your Strategy Hub, but you decide and act. L0 turns even that off. L2 and up are Maya automations — Maya takes the initiative: she decides what's worth surfacing and proposes it (L2), drafts briefs (L3), writes content (L4), and ships to enrolled surfaces (L5).
  • L2 → L3 is where Maya creates the artifact. At L2 Maya proposes, but you still create the brief from her proposal. At L3+ Maya creates the artifact herself (you review the draft). This second line is where the Professional tier unlocks more of Maya's hands-on work.

So: L0 = Off · L1 = Research & Recommendations (you act) · L2+ = Maya automations (Maya acts, you approve).

What never automates

Email blasts, cold outreach, and paid ad spend stay manual at every level. These surfaces involve real spend or real reputational risk and always require your explicit approval — there is no level at which Maya sends an email blast, fires off a cold email sequence, or launches paid spend on her own.

Maya checks brand fit before every brief

Before any brief is researched or counted against your plan, Maya asks one question: does this topic serve the people your brand sells to? She reads your brand settings, your personas, the words your own site uses and your excluded topics, and answers fit, unfit or unsure, with a short reason.

  • A topic Maya or the platform came up with that she judges off-brand is declined. It shows in your briefs list as Declined by Maya, with her reason, and costs nothing.
  • A topic you asked for that she doubts is held, not refused. Open the brief to see her reason, then choose Research it anyway (tell her why in a sentence) or Agree and delete. Held briefs you created also appear in Maya Desk under Waiting on You.
  • When she is unsure the brief is held for you the same way, as Needs Review.

Your answers teach her. Each override and each deletion becomes a precedent she reads the next time she judges a topic for your brand. A brief Maya questioned never turns into an article on its own: at L4 and L5 she writes articles only for topics that passed this check. The weekly digest tells you how many topics she declined that week.

This check runs at every level, including L0 and L1, because it covers briefs however they start, not only Maya's own. The one exception is the first article written while your account is set up, which is built from your own top keyword.

How drafts vs publish works

At L3, Maya drafts briefs; her briefs wait for you to generate the article, so no article credit is spent without your click. At L4, Maya also writes the article for each brief, up to her nightly draft limit. Either way, Maya creates drafts. Drafts wait in your inbox for approval. Nothing publishes until you click publish. The autonomy level controls how much Maya creates without asking, not whether she publishes without asking — that gate stays manual until L5, and L5 is per-surface (you opt in surface-by-surface).

L5 surfaces

Live today: auto-publish to blog, and sharing those articles on social. Maya's autonomy covers articles and social posts only. Lead magnets and landing pages are part of your campaigns, so Maya never creates or publishes them without your approval.

L5 is the only level where Maya can publish without your prior approval, and even then it's surface-by-surface. You decide which surfaces Maya is allowed to ship on. The two surfaces:

Surface

What Maya does

What you need first

Auto-publish to blog

Publishes the articles she writes overnight to your blog as live posts, once they pass her checks. Anything that fails a check waits for your approval.

A connected WordPress, Shopify or Payload blog.

Share auto-published articles on social

Once an auto-published article has been up for a day, writes a post about it and shares it to the accounts you tick. She never posts anything else on her own.

Blog auto-publish, plus a connected LinkedIn, Facebook, X, Threads, Bluesky, Pinterest or Reddit account.

When you elect L5, the Settings tab shows a checkbox for each surface. Surfaces whose prerequisites aren't met (e.g. no CMS connected for "auto-publish to blog") render disabled with a hint about what to connect first. You must check at least one surface to save L5; an empty selection is treated as "stay at your current level."

How blog auto-publish works

Maya only auto-publishes articles she wrote overnight, after you enrolled the blog surface, and only once:

  • the article's images have finished,
  • the topic passed Maya's brand-fit check,
  • the article sticks to its brief (at least 70% of its claims trace back to the brief),
  • it does not repeat a post your brand already published,
  • the article review passed it (an AI editor reads the whole article and its images against your brand settings and your Article Review Rulings; anything it would hold shows on the article page with the quote, the problem and a suggested fix),
  • it has never been sent to your blog before,
  • the article has no translation (a translated pair waits for you, so you can check both), and
  • your brand has one clear WordPress, Shopify or Payload blog to send it to.

Anything that misses a check stays in your Article Hub waiting for approval, exactly as at L4. Each auto-published article gets a notification and a line in Maya's activity. Maya also skips auto-publishing while you use an approval workflow with review stages. An auto-published article counts toward Maya's keep rate only after it has stayed up for 30 days; Maya checks your blog once a day during those 30 days, so unpublishing a post there counts even if you never open ILLIXIS.

Email blasts, cold email, and paid ad spend remain manual at every level — they're never on the L5 surface list.

Quality drift and automatic step-back

Maya watches what happens to the briefs and articles she writes on her own. A brief counts as kept once an article from it is approved, scheduled or published, and as kept with heavy edits if you changed more than a fifth of that article's words in the editor first. Maya's own finishing passes, like adding takeaways or links, never count as your edits. An article Maya generates automatically does not count on its own: a person has to sign it off. A brief you archive or delete counts as discarded. Anything left without a sign-off for 30 days, including a brief whose article you deleted, counts as ignored. If fewer than 60% of her drafts are kept over a 30-day window (with at least 10 drafts resolved), Maya automatically steps back one level. You'll see an in-app notice that she has stepped back.

This isn't a punishment — it's a circuit breaker. If Maya isn't matching your judgment, she should do less, not more. When you start keeping more of her drafts again, you can re-elect a higher level.

L2 is the floor. Maya never demotes you below L2 on drift; the propose-tier loop is cheap and the dismissals are themselves training data.

At L5, drift is per-surface. If you unpublish or delete something Maya auto-published within 30 days (in ILLIXIS, or on your WordPress, Shopify or Payload blog), that surface is removed from your L5 enrollment list. Editing a post is not a strike, but heavy edits lower Maya's keep rate. You stay at L5 for any other surfaces still enrolled. Only when the last surface gets removed does Maya step you down to L4. A strike against social-auto-publish doesn't revoke your blog-auto-publish authority.

Liability framing

When you elect L3+, you're accepting that Maya creates content without your prior review. You see drafts after she's done; you don't see the prompt or the intermediate work. That's a real shift in ownership compared to L1/L2, and the level-up confirmation modal makes it explicit. We keep an acceptance record so the agreement is auditable.

Selecting your level

Settings → Maya Autonomy. The radio shows only levels your tier permits. Picking L3+ opens an acceptance modal with the level-specific framing; you have to confirm before the new level takes effect.

Higher levels are earned. L0–L2 are always available within your tier. L3 unlocks once Maya has trained on at least 30 of your preference signals (briefs accepted/rejected, opportunities selected/rejected) with a confidence floor; until then the L3 radio is disabled and a hint shows your progress. L4 unlocks once at least 60% of Maya's drafts have been kept, over 10 or more drafts resolved in the last 30 days while at L3. L5 unlocks (Enterprise tier) once at least 75% of the drafts Maya wrote at L4 were published without heavy edits, over 20 or more resolved in the last 30 days. The L5 bar is stricter because L5 means autonomous publishing, so it counts only clean sign-offs and only Maya's work at L4.

The unlock gates are action-based, not calendar-day-based. An idle tenant 100 days in doesn't progress; an active tenant who clears the action thresholds in a single week does. If the share of her drafts you keep later drops below 60% over 10 or more drafts in a 30-day window, Maya automatically steps back one level. That is the same threshold she had to clear to ascend, applied symmetrically for descent.

How the meter works

In the proposal inbox you'll see a learning meter once your tenant is more than 14 days old. Each Promote counts as 1 high-quality decision; each Dismiss counts as 0.5 (a dismiss is signal, but a thoughtful Promote is more signal). The 100-point target is a UI proxy, not the underlying ML threshold — Maya's preference model trains on its own schedule and graduation to L3 depends on both the model state and your tier.

The meter is hidden in the first 14 days so early signals can land organically. It's also hidden once you're already at L3+ — at that point the meter has done its job.