Configure Growth Assistant agent autopilot

Updated May 2026

You'll learn how to: Pick how aggressively each AI agent runs autonomously, and which actions need your explicit approval before they take effect.

Time: 5-15 minutes per agent.

What "autopilot" actually means

Growth Assistant ships several AI agents (Reply Coach, Daily Briefing, Approvals agent, Network Sweeper, etc.). Each one has a configurable autonomy level:

Manual only: agent runs only when you ask. No background activity.
Suggest: agent runs in the background, queues drafts / suggestions, but never acts on them. You approve / reject each one from the Today Approvals queue.
Suggest + auto-low-risk: same as Suggest, plus auto-fires actions classified as low-risk (e.g. background contact enrichment, marking inbound emails as read).
Full autopilot: agent acts autonomously within its remit. Approval queue is only for explicit edge cases (e.g. outbound to unknown contacts).

The right setting depends on how much trust you've built with the agent and how high-stakes its actions are.

Steps

Open Settings (gear icon, bottom-left).
Click the Growth Assistant tab (the Bot icon).
Find the Agents section. Each agent is listed with its current autonomy level.
Click an agent to expand its config.
Pick a new autonomy level from the radio buttons / dropdown.
(Optional) Customise per-action permissions:
- "Send emails" -> approval required.
- "Run enrichment" -> auto.
- "Schedule meetings" -> approval required.
- "Update contact fields" -> auto.
(Optional) Set per-agent credit caps so an over-eager agent doesn't drain your balance.
Save.

The new config applies on the agent's next run cycle (typically within a few minutes).

Common autonomy profiles

Brand new to Nynch: every agent on Suggest for the first 2 weeks. See what they propose. Reject what's wrong. After you're seeing 90%+ acceptance, escalate.
After 2 weeks of light use: Reply Coach + Daily Briefing on Suggest. Background enrichment and field updates on Auto-low-risk. Outbound and meetings stay on Suggest forever (high-stakes).
Heavy power user / solo founder: Suggest+auto-low-risk for everything except: outbound email to never-contacted prospects, calendar invites to senior buyers. Keep those on Suggest.
Lockdown: every agent on Manual. Use this if you want zero AI activity without explicit clicks.

Tone and skills config

Below the agents section, the Growth Assistant tab also configures:

Default tone (formal / conversational / direct).
Style notes: free-text guidance to all agents about your voice ("avoid overly American phrasing", "always sign as the company, not me").
Skills access: which saved skills each agent can use when drafting.
Knowledge sources: which corpora the agents ground in.

Tune these together. Agent autonomy without tone calibration produces fast-but-wrong output.

View activity logs

Each agent has an activity log:

What it did, when, on which entity.
What it proposed but never acted on (Suggest queue history).
Credit spend per run.

Useful for spot-checking that the autopilot isn't doing anything weird.

If something goes wrong

Symptom: "Growth Assistant tab isn't visible." → Fix: Your plan may not include it. Check Settings > Account.
Symptom: "I set Full autopilot but the agent isn't acting." → Fix: Some actions are gated regardless of autonomy level (outbound to never-contacted prospects, calendar invites to senior people). Check the per-action permissions inside the agent.
Symptom: "Agent kept running after I changed to Manual." → Fix: In-flight actions complete before the new setting takes effect. After the current batch, the agent goes manual.
Symptom: "Credit cap exhausted but agent isn't paused." → Fix: Refresh the page. Cap enforcement runs on the next agent cycle.