AI Hub runs at three levels of automation, set per tab. Manual keeps you at the keyboard, Auto-Supervised lets the agent draft and send under a safety gate while checking in with you on anything that commits, and Fully Autonomous works batches of loads on its own — but still stops short of committing money without you. New tabs start in Auto-Supervised (semi-auto).
Whichever mode you pick, one thing doesn't change: the agent never finalizes a booking, a rate, or a ratecon on its own. The mode only changes how much of the drafting and sending it handles before a decision reaches you.
The three modes
Manual
You're in the driver's seat. The agent helps you find, rank, and draft, but you compose and send the emails yourself. There's no auto-send here — nothing goes to a broker without you clicking send. This is the simplest mental model and the lightest-touch mode.
Auto-Supervised (semi-auto)
The default. The agent drafts and can send replies, but every outbound message passes through the send classifier first — a gate that judges whether a draft is safe to send or needs your sign-off. Anything that makes a commitment, or leans on a remembered number about your driver or lanes, comes back to you for approval. You're supervising, not typing.
Fully Autonomous
The agent works a batch of loads on its own — polling for matches, drafting, and sending under the same send gate, without checking in on every step. It's built for volume. Even here, the human stays in the loop for the things that matter: booking commitments still require your explicit go-ahead. See Autonomous mode and personas for how batches and safe-send guardrails work.
When the send gate kicks in
The difference between modes is really about that send classifier:
- In Manual, the classifier is out of the path entirely — you're the gate. That also means Manual is the fastest per email, with no extra round-trip.
- In Auto-Supervised and Fully Autonomous, drafts run through the classifier before they can send.
- One exception: once you've hit Book It on a load, that load's replies stay on the classifier path even in Manual — because a booking is in flight and the agent needs the guardrail.
Each agent also gets mode-specific rules for the mode you're in, so a reply drafted in Fully Autonomous follows different instructions than the same reply in Manual.
How to switch modes
Mode is set per tab, so different search contexts can run at different automation levels at the same time. The AI-mode selector shows up in more than one place — the main panel and the email rail — and changes apply to the active tab.
The per-tab AI-mode selector A product screenshot will be added here.
One thing to expect: the selector is disabled while a run is in flight. If you try to change mode mid-run you'll see the tooltip "Wait for the current run to finish before changing mode." Let the current run wrap up, then switch. This prevents the agent from changing its rules halfway through a job.
Good to know
- New tabs default to Auto-Supervised if you don't pick a mode. It's the balance most dispatchers want — the agent does the work, you approve the commitments.
- Driver long-term memory is off in Fully Autonomous for the send gate. In the supervised modes, any broker-facing claim grounded in remembered facts routes through your approval first.
- No mode removes you from booking. The agent never sends a booking commitment on its own — that's always Book It, and voice booking actions require a spoken confirmation.
Related
- Autonomous mode and personas — polling, batches, playbooks, and safe-send guardrails.
- How negotiation works — the rules every draft follows.
- Book It and the booking flow — the explicit commit step.
Need help? Email support@numeo.ai or use Report a problem in the app.