Autonomy Levels — Manual to Aggressive
Manual, Assisted, Auto, Aggressive — what each level actually means, how to pick one, and when to graduate.

Lesson 1: The Four Autonomy Levels
Bolta ships four discrete autonomy levels. Each one is a real behavioral change, not a vibe.
Manual
What happens: The agent drafts. You decide whether it ever gets posted.
- Agent generates the draft on schedule.
- It lands in your inbox.
- Nothing happens until you approve, edit, or reject.
Use this when: You're brand-new to Bolta and you don't yet trust the output. Or when stakes are unusually high (e.g., a regulated industry, a sensitive launch week).
Assisted
What happens: The agent drafts and recommends — but still waits for your approval.
- Agent generates the draft, picks the best posting time, picks the right platforms.
- Recommends a specific schedule.
- Waits for you to approve before anything goes live.
Use this when: You want to spot-check 100% of output but offload the decisions around timing, platform, and structure. This is the right starting point for most people.
Auto
What happens: The agent drafts, posts, and reports back.
- Agent generates the draft, posts on its schedule, and notifies you.
- You can intervene before the scheduled time, but the default is "publish."
- Approval becomes opt-out instead of opt-in.
Use this when: You've reviewed 30+ outputs from this specific agent and the failure rate is below 5%. Auto is where the real time-savings show up.
Aggressive
What happens: The agent posts, replies, engages, and adjusts strategy without checking with you.
- Schedule shifts based on performance signals automatically.
- Agent will reply, follow, and engage on your behalf in real time.
- You're notified after the fact via the activity feed.
Use this when: You have a months-long track record with the agent and you've defined hard guardrails in the persona. Aggressive is for trusted agents in mature workflows, not for week one.
Lesson 2: Picking the Right Level
The single most common mistake: starting at Auto on day one. The second most common: staying at Manual after you trust the agent.
A simple framework
- First 7 days — Manual or Assisted. You're learning what your agent's voice sounds like.
- Days 8–30 — Assisted. Approve everything. Edit when needed. The agent is being trained by your edits.
- Days 31–60 — Move trusted agents to Auto. Keep less-trusted ones at Assisted.
- Day 60+ — Aggressive only for agents that have shown bulletproof judgment. Most agents stay at Auto forever, and that's fine.
Match autonomy to stakes
- Drafting a Threads post — Auto is low stakes. Worst case, you delete it.
- Replying to a customer DM — Manual or Assisted. The blast radius is real.
- Posting on a Facebook business Page during a launch week — Manual. Always.
Don't conflate autonomy with trust
A high-autonomy agent isn't a "better" agent — it's just one with more rope. Some jobs deserve more rope (publishing routine content). Some don't (responding to a complaint thread).
Lesson 3: Graduating Up (and Sometimes Down)
Autonomy isn't a one-way ladder. You'll move agents up and down as conditions change.
When to graduate up
- You've approved 30+ outputs without meaningful edits.
- Performance metrics are trending in the right direction.
- You catch yourself rubber-stamping approvals — that's the sign you're ready to skip the step.
When to drop back down
- The agent goes off-tone (often after a Voice DNA refresh on stale data).
- You're entering a sensitive period (launch, fundraise, controversy).
- You hire a new agent that hasn't earned trust yet — keep it on Assisted regardless of what your other agents are doing.
Common myths
- "Higher autonomy means better quality." No — it means less oversight. Quality comes from Voice DNA, persona, and feedback loops, not autonomy.
- "I should set everything to Aggressive once it works." No — Aggressive carries real risk. Most workforces hit a sweet spot at Auto for content + Assisted for replies.
- "If I move to Auto, I'm done with this agent." No — you should still skim outputs weekly. Auto is "check less," not "check never."
Action Steps
- Open every agent you've hired and write down its current autonomy level.
- For each one, ask: have I approved 30+ of its outputs without meaningful edits?
- If yes — graduate it up one level. If no — keep it where it is and review again in two weeks.
Autonomy is the dial that turns a Bolta workforce from "AI assistant" into "team member." Tune it deliberately. Next lesson: blueprints, the shortcut to skipping all of this thinking on day one.