Automating sales outreach with a voice agent comes down to six setup decisions: the lead list, the script, the qualification criteria, the routing rules, the calendar connection, and the launch window. AI Central's How to Automate Sales Outreach walks through each one using ElevenAgents, ElevenLabs' voice agent product. The agent places the calls, screens against your criteria and books meetings live, so reps only ever speak to leads that already qualified.
Reviewed August 2026.
Why outbound breaks at the top of the funnel
AI Central opens with an arithmetic problem rather than a motivation problem. Most sales teams, the guide argues, are wasting 60 percent of their time on cold calls that go nowhere, on leads that go cold before anyone follows up, and on hours of manual outreach with unpredictable results.
The numbers underneath that claim are the ones worth holding on to. A human rep makes 40 to 60 calls a day. In the before picture AI Central sketches, reps spend three hours daily on cold outreach and 80 percent of those calls go unanswered or turn out to be unqualified. Weekends are dead air, and by Monday the leads have cooled.
That is not a coaching problem. Dialling harder does not change the hit rate, it just pushes the same ratio through more hours.
What the agent takes over
The proposal AI Central makes is narrow and specific. ElevenAgents runs the outbound layer: it contacts leads, qualifies them against criteria you define, and books calls with warm leads only. The framing used for what that leaves behind is blunt.
Your reps talk only to pre-qualified prospects
Read that as a division of labour rather than a replacement. The agent absorbs the volume, the dead numbers and the disqualifications. The rep keeps the conversations where judgment and pricing actually matter. AI Central poses it as a question worth sitting with, what if your agent made the first 200 calls.
Step one, build a lead list the agent can work with
Setup starts with data. The instruction from AI Central is to export from a CRM or import a CSV into ElevenAgents, with four fields as the baseline: name, company, phone and context. Salesforce and HubSpot can be connected directly for auto-sync, which keeps the list live rather than a snapshot taken on the day you launched.
That fourth field is the one teams skip, and the guide is emphatic about it.
More context = more personal outreach
The quality ceiling of every call is set right here. An agent holding a name and a number can only run a generic pitch. An agent that knows the account, the trigger and the reason for the call has something specific to open with, which is what earns the next thirty seconds.
Step two, write for a conversation, not a pitch
AI Central sets two constraints on the script. Keep it under two minutes, and shorter is better. And keep the register human.
Keep it conversational, not salesy
The structure prescribed has four parts, an opening, a value statement, qualifying questions and the ask. The order carries weight. Qualifying questions sit ahead of the ask, so the agent has already learned whether the person is worth booking by the time it proposes a meeting.
Step three, define what qualified actually means
This is the step that decides whether automation helps or hurts, and AI Central treats it as core configuration rather than a detail. You tell the agent what a qualified lead looks like, along four dimensions.
- Budget range.
- Decision-maker authority.
- Timeline.
- Company size.
Written down, those look obvious. In practice most teams carry them informally in a rep's head, which is why two reps working the same list produce two different definitions of a good meeting. Handing the criteria to an agent forces the argument into the open, and the answer stays consistent across every call because the same logic runs each time.
Step four, route the outcome automatically
Once a call ends the lead has to go somewhere, and the routing rules AI Central lays out cover the four possible answers. A hot lead is transferred to a rep or booked into a meeting live. A warm lead gets a follow-up auto-scheduled three days out. A not now goes into a 30-day nurture sequence. A lead that is not a fit gets marked in the CRM and dropped.
Zero manual sorting
The nurture and follow-up branches are the quiet win. In a manual process the leads that say not now are exactly the ones that fall through, because nobody is paid to remember them. A routing rule remembers them by default.
Step five, let it book the meeting mid-call
The calendar connection is a one-time step, Google Calendar, Outlook or Cal.com. The point AI Central makes is about timing rather than plumbing: when a lead qualifies, the agent books the meeting immediately, in the middle of the conversation.
Every hop between interest and a confirmed slot leaks. Send a link, wait for a reply, propose times, and the intent that existed during the call has evaporated. Booking inside the call is the difference between a lead and a calendar entry.
Step six, launch inside the lead's day
The final setup decision is call volume and timing window. AI Central notes that the agent runs during business hours in the lead's timezone, automatically, and handles dozens of simultaneous calls at consistent quality.
That is where the arithmetic flips. A rep does 40 to 60 calls a day, one after another, and quality sags as the day wears on. The agent runs hundreds at once, and call two hundred sounds like call one. Outreach also keeps going in the evenings and at weekends, which is precisely the window the guide identifies as where pipeline used to go cold.
What the after picture looks like
The change AI Central claims comes down to three lines. Reps only call warm, pre-qualified leads. Outreach runs around the clock, evenings and weekends included. And the state of the pipeline on the first working day of the week is different.
The pipeline is full every Monday morning
Notice what has not changed. Nobody claimed a higher close rate or a better rep. The whole gain sits in what reaches the rep, a filtered list instead of a raw one, so their hours land on conversations that can actually convert.
Where this goes wrong
Two failure modes are visible in the workflow AI Central lays out, and both are self-inflicted.
The first is loose qualification criteria. The agent enforces exactly what you specify, and at hundreds of simultaneous calls a sloppy definition of what counts as qualified does not produce a few bad meetings, it produces a calendar full of them. Tighten the criteria before you raise the volume.
The second is a thin lead list. Since context is what makes outreach personal, a file with names and phone numbers and nothing else caps the quality of every call the agent will ever make, however good the script is. Fix the data before you blame the agent.
What to do first
If you have never run a voice agent, do not start with the whole list. Take the segment you understand best, write the qualification criteria for it in plain language, and run a small campaign so you can hear what the agent actually says to a stranger.
If you already run outbound at volume, the fastest gain is not the calling, it is the routing. Wire the not now branch to a 30-day nurture and the warm branch to a three-day follow-up before you touch anything else, because that is the pipeline you are currently losing by hand.
Whichever camp you are in, spend the setup time on the context column of your lead list. It is the cheapest lever in the whole workflow.
Do I need a CRM to use this?
No. AI Central gives two paths into the lead list, export from a CRM or import a CSV. Connecting Salesforce or HubSpot buys you auto-sync so the list stays current, but a spreadsheet carrying name, company, phone and context is enough to launch.
How long should the outreach script be?
Under two minutes, and AI Central is explicit that shorter is better. Four parts fit inside that window, an opening, a value statement, qualifying questions and the ask.
What happens to a lead that says not now?
It drops into a 30-day nurture sequence automatically under the routing rules. Warm leads get a follow-up scheduled three days out, hot leads are transferred or booked on the spot, and leads that are not a fit are marked in the CRM and left alone.
Can the agent really book the meeting, or does it just send a link?
It books. With Google Calendar, Outlook or Cal.com connected, the agent schedules the meeting mid-conversation the moment the lead qualifies, instead of handing off a link and hoping.
How many calls can it handle at once?
AI Central puts a human rep at 40 to 60 calls a day and ElevenAgents at hundreds simultaneously, at consistent quality. The agent also works inside each lead's local business hours, so the volume does not come at the cost of ringing people at the wrong time.