Autonomy

Let Corty close the loop.

Start by reviewing every draft. Corty drafts until its review score earns L3 for a client; then it sends routine replies and follow-ups on its own, and steps back automatically if quality slips.

14-day free trial · 150 Neural Credits · 1 client, up to 2 campaigns · no credit card

Autonomy levels

Autonomy it has to earn.

What if Corty sends something wrong? It starts at L0, where every draft is reviewed. It only moves up when its reviewed replies score well, and it moves back down on its own when they don't.

L0 Default start

Review everything

Every draft goes to the review queue. A person approves, edits or skips it.

L1

Reps: routine goes straight out

For reps and connected agents: routine replies send directly. Hot leads, opt-outs, legal requests and uncertain replies go to review. Corty still drafts for review at this level.

L2

Reps: send, with spot checks

For reps and connected agents: replies send directly and one in ten is pulled for a spot check. Corty still drafts for review at this level.

L3 Earned

Earned autonomy

Corty sends on its own once its review score is 85 or higher over at least 5 reviewed replies. A connected AI agent gets here when an owner or admin raises it on its review record. People go up to L2. Corty never sends legal and GDPR requests on its own.

If the review score drops below 85, Corty steps down a level automatically and tells you. You can pull any client back to L0, or block sending for a client entirely, at any moment.

Why it matters

Speed books meetings. Autonomy makes speed possible.

More revenue from the same leads comes from answering every positive reply fast, well, and without dropping the ball. At some volume that means letting Corty act on its own.

You shouldn't have to take that on faith. Every rep, every agent and Corty itself starts in review, earns autonomy only on a proven review score, is spot-checked once it sends on its own, and is pulled back automatically the moment quality slips.

Review score

Earned on reviewed replies, per client.

  • Threshold 85, over at least 5 reviewed replies in the last 30 days
  • Set per client, per rep and per connected AI agent
  • Steps down one level on its own when the score drops
  • Corty never sends legal and GDPR requests on its own
Review score and autonomy eligibility per client

Spot checks

Once it sends alone, one in ten gets a second look.

At L2 and L3 a random 10% of sent replies lands in the spot check tab. A reviewer marks it OK or corrects it, and that feeds the review score. Nobody reviews their own sends.

Spot check of a reply Corty sent on its own

Decision Trail

Every send, on the record.

Per reply: who or what sent it (rep, client via portal, Corty, an AI agent), at which level, and which playbooks and guidelines the draft was built on. The record you can show a client when they ask what went out under their name.

Decision Trail of a reply Corty sent autonomously

Safety

When in doubt, it doesn't send.

  • Only the autonomy gate can issue a send permit. If the gate can't be reached, nothing goes out.
  • A blocked client stays blocked: the check runs again right before delivery.
  • Review queues older than 4 hours are escalated to the owner.
  • An account that falls behind on payment goes read-only after 48 hours: no AI, no sends.

Bring your own agent? Connect your AI to the same rules.

Autonomy you can trust: earned, spot-checked, reversible, and on the record.

Questions about autonomy

What is the review score?

When a person reviews a reply Corty drafted or sent, the outcome counts toward a score per client, rep or agent over the last 30 days. Corty can only send on its own (L3) with a score of 85 or higher over at least 5 reviewed replies.

What happens if quality drops?

Once there are enough reviews, a score below 85 moves that client, rep or agent down one level automatically, and the owner is told.

Can I stop all sending for a client?

Yes. The "Turn Off Sending" switch blocks every send for that client, from the dashboard, the portal, Corty and the API. Only owners and admins can change it.

Will Corty make things up?

Corty works under hard rules that override every playbook. It only states facts about your client that are written in the client's documents: prices, features, certifications, customers, dates. If a lead asks something those documents don't answer, Corty doesn't guess. It writes a neutral holding line and flags it on the draft as an "Open question" for a person to answer. It never invents meeting times; it offers the booking link instead. Legal and GDPR requests are never sent automatically and always go to a person.

Does this apply to my own AI agent too?

Yes. An agent connected through the API sends through the same gate, starts in review and earns its level the same way. See Connect your AI.

Prefer your client to send? See the client portal. Works with Instantly and Smartlead.

Same leads. More revenue.

Connect Smartlead or Instantly and see which replies Corty turns into meetings this week.

14-day free trial · 150 Neural Credits · 1 client, up to 2 campaigns · no credit card