Fully Autonomous is the hands-off mode: the agent watches your search, works loads in batches, and drafts and sends under the send gate without checking in on every step. It's built for volume. This article covers what "autonomous" actually does — the polling, the batch behavior, persona playbooks, and the guardrails that keep it from double-sending or running away — and where the human stays firmly in the loop.
If you're deciding between modes first, start with AI modes.
How an autonomous run behaves
Instead of running once when you search, Fully Autonomous keeps a watcher on the tab and wakes up to work new matches as they appear. You'll see autonomous runs surfaced in the UI with poll controls and a run card, so you can watch what it's doing and stop it when you want.
An autonomous run in progress with poll controls A product screenshot will be added here.
By default the agent works a fixed batch — roughly 10 to 20 loads at a time — rather than trying to email everything at once. That keeps volume predictable and your outbox sane. The batch behavior changes when a persona is active (below).
Persona playbooks
A persona is an advanced capability Numeo can turn on for your account — a playbook tuned to your operation and truck type. It's selected internally by equipment type behind a feature flag, not something you author, create, or add yourself. When a persona is active, the agent follows the playbook instead of the default batch behavior.
A persona shapes the same four things the defaults handle:
- Without a persona: the fixed 10–20 load batch, default ranking, default negotiation rules.
- With a persona: load selection, ranking, volume, and negotiation all follow the playbook.
Because a persona is tuned to a specific operation, it's a path Numeo enables for you rather than a setting you toggle. If a persona would fit how you run, ask us to turn one on for your account.
The safe-send guardrails
Autonomous mode does more on its own, so it leans on several guardrails that keep it safe. You don't configure these — they're always on — but it helps to know they're there:
- One broker, counted once. Outreach is deduped by broker MC number, so a re-run or restart won't cold-email the same broker again. This is the same mechanism behind the "already contacted" tooltip.
- A lease and an action log. Autonomous side-effects are recorded, so a re-triggered search or a service restart can't re-send an email that already went out.
- Anti-loop protection. Built-in run gating, active-run tracking, a circuit breaker, and a loop detector stop the agent from spinning or repeating itself.
- Send acknowledgement recovery. If a send drops mid-flight, a watchdog recovers it rather than silently firing twice.
- Pause means pause. You can stop the agent, and it stops picking up new work.
Together these are why "autonomous" doesn't mean "unsupervised risk" — the failure modes that would matter most (double-sends, loops, re-sends after a restart) are guarded against by design.
What autonomous still won't do
This is the important part. Fully Autonomous automates the outreach and negotiation drafting — it does not remove you from the commitment:
- Booking a load is still your call. The agent never sends a booking commitment on its own; that's always Book It, your explicit pre-authorization.
- Driver long-term memory is off in this mode for the send gate — the agent won't lean on remembered numbers to make broker-facing claims.
- Voice money-actions require a spoken "yes." Booking, bidding, or sending anything to a broker over voice always gets a read-back and an explicit confirmation.
So even at full automation, the money decisions stay with you. The agent handles the volume of typing and chasing; you keep the commit.
Related
- AI modes — Manual, Auto-Supervised, and Fully Autonomous compared.
- How negotiation works — the rules every draft follows, autonomous or not.
- Book It and the booking flow — the explicit commit the agent can't do for you.
Need help? Email support@numeo.ai or use Report a problem in the app.