Draco · Skyline Blog

What Is Copilot Mode in a Sales AI?

Fabián Alvarez5 min read

TL;DR

  • Copilot mode is when the AI proposes a reply and a person approves or corrects it before it's sent.
  • It's the recommended starting point for an agency that has never used AI in sales.
  • Draco learns from every correction the human team makes.
  • Once the team trusts the replies, it switches to autonomous mode, and can revert at any time.

Copilot mode is a mode where the AI drafts the reply and a person approves or corrects it before sending. In Draco, Skyline's AI CRM, it's every agency's starting point: it lets the team validate tone and content before giving the AI full control, and switches to autonomous once there's trust.

What Is Copilot Mode in a Sales AI?

What exactly is copilot mode

Copilot mode is the mode where the AI drafts the reply and a person approves or corrects it before sending. That is the definition, plain and simple. It is not a chatbot that answers on its own, nor an assistant that fires off messages automatically: it is a draft that waits for a green light.

In Draco, Skyline's AI CRM, copilot mode is the starting state of the account. Draco connects WhatsApp, Instagram, Messenger and TikTok in one inbox, and that is where a team of five AI agents works. When an agency activates Draco, it starts in copilot mode: the AI proposes, a person approves or corrects, and it only moves to autonomous when the team trusts it. That switch can be reverted at any time, without losing any of the setup.

For an agency fielding dozens of messages a day between quotes, itinerary questions and follow-ups, copilot solves a concrete problem: nobody on the team has to write every reply from scratch, but full control isn't handed over to a system that doesn't yet know the business.

Why copilot is the right starting point

Assuming a sales AI must be "ready for anything" from day one is the most common mistake when adopting these tools. An agency in Cartagena fielding questions about boat tours, beach plans and transfers doesn't have the same catalog or tone as an adventure tourism agency in the coffee region. The AI needs to calibrate to that specific context before it talks directly to a customer.

Copilot mode gives an agency three things it wouldn't have starting in automatic:

  • Control over brand tone. The team sees every proposed reply before it goes out and can adjust the style until it sounds like the agency, not a generic AI.
  • Review of catalog and pricing. Since Draco is connected to the real Skyline Partner catalog, proposals already start from experiences and prices that actually exist, but the human team still validates that the recommendation makes sense for that specific customer.
  • Progressive trust. Nobody at the agency has to "trust blindly" from the first message. Trust builds by reviewing real proposals, one at a time.

This is different from what a rule-based chatbot offers, where replies are fixed and there is no room for fine correction: either the flow is programmed correctly or it fails. Copilot, instead, assumes the first version doesn't need to be perfect, only reviewable.

How the AI learns during copilot mode

Here's the point many agencies overlook: copilot mode isn't just a safety layer, it is also the training mechanism. Draco learns from every correction: the Approve and Correct buttons feed the AI. Every time someone on the team adjusts a reply proposed by Vega, the agent handling attention and closing, that adjustment gets recorded and sharpens the next proposals.

This changes how a team should think about its daily work. Approving or correcting isn't just another administrative task: it is how the agency teaches its own AI how to sell. An agency running diving tours in Santa Marta, for example, might notice that Vega's first proposals are correct on price but generic in tone. By correcting those replies during the first weeks, the team isn't just managing today's conversation: it is shaping how Draco will respond tomorrow.

What happens to each agent during copilot

Copilot mode applies mainly to Vega, who is the one talking directly with the customer, prequalifying, quoting, building packages and carrying the sale through to payment. The other agents on the team work in parallel: Kai handles lead scoring, Axel analyzes every night how the team handled conversations, Leo manages proactive follow-ups and Iris reviews quiet conversations on the board. The work of approving or correcting concentrates on what Vega proposes, because that's the direct interaction with the customer.

When and how to move to autonomous mode

There is no fixed date for leaving copilot mode. The switch happens when the team trusts that the AI's proposals are consistent: the tone is right, prices match the catalog and replies don't need frequent correction. At that point, the agency can move to autonomous mode and let Vega reply directly, without waiting for a person's approval on every message.

What matters is that this step isn't permanent or irreversible. The agency can go back to copilot mode at any time, for example if it launches a new product that isn't well calibrated yet, or if it changes season and wants to closely review how the AI responds to new traveler questions.

A table to decide which mode to work in

Agency situationRecommended mode
Just activated DracoCopilot
Launching a new funnel or productCopilot
Team notices consistent, error-free proposalsAutonomous
High season with message volume overwhelming the teamAutonomous, with periodic review
Brand tone change or special campaignTemporarily back to copilot

What to do next

If your agency is evaluating a sales AI for WhatsApp, Instagram or TikTok, the most useful criterion isn't how "autonomous" it is from day one, but whether it has a real copilot mode where your team can approve or correct before the AI learns your business. Check how that mode works in Draco before deciding when to move to autonomous.

Frequently asked questions

Does copilot mode slow down replies to customers?

The team reviews the proposal before sending it, so speed depends on how fast someone approves or corrects. Many agencies use it in the first weeks and then move to autonomous precisely to gain speed.

Can you use copilot in some funnels and autonomous in others?

Draco allows multiple funnels per business, each with its own knowledge and mission, so it's reasonable to adjust the trust level funnel by funnel based on how mature each one is.

Which Draco agent works in copilot mode?

Copilot mode applies mainly to Vega, the agent that handles attention, prequalification and closing by talking directly with the customer. The other agents on the five-agent team work on parallel scoring, analysis and follow-up tasks.

How does the team know it's ready to switch to autonomous mode?

There's no fixed automatic signal: the sign is that the AI's proposals stop needing frequent correction in tone, price or content. It's a team decision based on trust built by reviewing replies.

Does correcting a reply take a lot of the team's time?

It's a quick action on an already-drafted message, not writing from scratch. That correction also feeds the AI, so the time invested reduces future corrections.

Draco

See Draco in action

The AI CRM for agencies and tour operators: WhatsApp, Instagram, Messenger and TikTok in one inbox, handled by five AI agents.

Try Draco
draco.skyln.co
Draco

Leer este artículo en español →