How LeadPedlar's AI Workflow Runs Unattended Calling Campaigns

An unattended AI calling campaign is one you set up once and leave to run: the AI agent dials the list, talks to the people who answer, leaves voicemails on the beep, books callbacks and hands hot leads to a person — with nobody sitting at a dialer. This is how that works in LeadPedlar, screen by screen, so you know exactly what you are switching on.

The case for it is arithmetic. Gong Labs' analysis of 300 million cold calls put the average connect rate at about 5.4%, so roughly nineteen dials in twenty end in a voicemail, a silent line or no answer. A human team pays for each of those dials in salary and attention; an autonomous AI dialer pays in minutes, and handles the voicemail properly instead of hanging up on it.

We cover the campaign context, the review with the AI, the handling mode, hand-off rules and SIP extensions, calling hours, the start, what the AI does on each kind of answer, and how you watch the campaign and redial.

What “unattended” means in an AI calling campaign

In older dialer software an unattended or agentless campaign plays a recorded message or a keypad menu, and nobody on the line can reach anyone. An AI calling campaign is different in kind: the agent holds a conversation, answers questions from the campaign's brief, and can pass the call to a person or book a callback. Unattended means nobody has to be watching — not that nobody can be reached.

LeadPedlar offers four ways to handle a campaign's calls, and the unattended one is the end of a spectrum. AI start to finish: the AI agent handles every call itself and never passes it to a person. AI first, people when asked: the AI talks first and passes the call to a free person when the caller asks; if nobody is free, it books a callback. People first, AI steps in and People only keep humans on the calls, with the AI covering answering machines or the moments when nobody is free.

Responsible use is the same in every mode. The list you load should be one you are entitled to call; LeadPedlar then keeps to the campaign's calling hours in each lead's time zone, skips do-not-call numbers, honours a do-not-call request made on the call itself, and has the AI introduce itself as an AI assistant.

• ~5.4%Average cold-call connect rate across 300M calls — most dials reach a machine, a silent line or nobodyGong Labs, 300M cold calls (2024)
• 64%Customers who would prefer companies didn't use AI for service (survey of 5,728) — why a hand-off path mattersGartner customer survey (Jul 2024)
• ~7×More likely to qualify a lead when contacted within an hour than an hour later (audit of 2,241 firms)Harvard Business Review (2011)
• 28%Share of the week sales reps actually spend selling (survey of 7,700+) — dialing is where the rest goesSalesforce, State of Sales

Step 1 — Write the campaign's AI context

Everything the AI says on a campaign's calls comes from the campaign's AI context: one free-text box the team leader or an admin writes in their own words. The AI agent you pick separately only sets the voice and behaviour; the context sets the substance. Each campaign has its own, and a blended campaign has one for outbound calls and another for inbound.

The box's placeholder shows the shape that works: the business and what it offers; the goal (place an order, or book a callback); what to say first; facts and prices; the questions to ask; when to pass to a person; what never to promise; and the tone. A separate field, The AI opens with, holds the first line — left empty, the AI opens with “Hi, this is <agent> from <business>. Am I speaking with <first name>?”

Write it like a briefing for a new hire rather than a script. The AI reads phone numbers digit-group by digit-group, keeps its turns short and stops when interrupted, so you need the facts, the boundaries and the goal rather than stage directions. A campaign cannot be turned on without a context of a minimum length.

Step 2 — Review the context with the AI

Under the box sits Review with AI: a short chat in which the AI reads your context and asks what it still needs to run the calls well. You answer in plain language — opening hours, what counts as a qualified lead, which objections come up — and it may propose an updated text. Nothing changes unless you press Apply; Keep mine leaves your version untouched.

If the box is empty, the same chat offers to help you write it from scratch. The reviewer runs on the campaign's AI agent, so pick the agent first. Use the review to put in the hand-off conditions and the never-say list; those lines keep an unattended campaign safe when nobody is listening.

Step 3 — Choose how the calls are handled

The Call handling step asks “Who talks to the people you call?” For an unattended campaign the answer is AI start to finish. For a safety valve, AI first, people when asked keeps the AI in charge but lets a caller reach a person when one is free, and books a callback when nobody is.

Next comes When an answering machine picks up. Leave a personalized voicemail: written for that lead from your voicemail instructions or from the context. Hand to a person: the call rings your people like a hand-off so someone can leave the message or decide — not offered when the AI handles every call start to finish, since that mode never passes a call to a person. Hang up: nothing is said, and the lead is retried later like a no-answer.

The optional voicemail instructions take the reason for the call, a number to call back and the tone; without them the AI writes the message from the context. The AI will only leave a callback number that is in those instructions or in the context — it never improvises one. For why a message on the beep is worth leaving, see answering machine detection and AI voicemail drop.

Step 4 — Hand-off rules and SIP extensions

A campaign in AI first, people when asked mode needs to know who to ring. The setting is titled When the AI hands the call to a person, hand it to: People on the team (whoever the hand-off rule names); These extensions (only the chosen SIP extensions ring, offline ones are skipped, and if nobody answers the AI takes a message); or People and extensions (both ring at once; the first to answer takes the call).

Hand-off rules decide the order. A rule has up to three steps; each names who rings — whoever started the call, everyone working the list, a team, particular people, the list's team leaders, admins, or SIP extensions — plus how (all at once, one at a time in order, round robin or longest idle) and for how many seconds. The built-in default rings whoever started the call and everyone working the list, all at once.

The result of every hand-off is recorded on the call: connected to a person or an extension, nobody available, all agents busy, or no answer — in which case the caller goes back to the AI, which apologises and books a callback. SIP extensions let a teammate take hand-offs on a softphone such as Zoiper, X-Lite or Bria, behind an IP allowlist, so an unattended campaign can still land a call on a desk phone.

Step 5 — Calling hours, attempts and turning it on

The calling window is a set of weekdays, a start and end time and a time zone; an end before the start runs overnight. The recommended Judge the window in each lead's own time zone option uses the zone known from the number, so a list that spans a country is called at each lead's local time; leads without a known zone use the campaign's zone. The AI dials only inside the window and waits outside it.

Attempts per lead sets how many times each lead is dialed in total (1 means never redial), and the retry wait sets how long the campaign waits after an unanswered call. Then you turn the campaign on. No call goes out unless the AI agent is ready and the campaign has its context, and the AI can work the list one call at a time or several in parallel.

1. 1

   Step 1: Write the context

   Business, offer, goal, opening line, facts and prices, questions to ask, when to pass to a person, what never to say, tone.

2. 2

   Step 2: Review it with the AI

   Answer what the reviewer asks; Apply its suggested text or Keep mine.

3. 3

   Step 3: Pick the handling

   AI start to finish for fully unattended; AI first, people when asked for a hand-off path. Set what happens when an answering machine picks up.

4. 4

   Step 4: Set the hand-off

   People on the team, these extensions, or both — and the hand-off rule that orders them.

5. 5

   Step 5: Set the hours and attempts

   Days, window and time zone, judged per lead; attempts per lead and the retry wait.

6. 6

   Step 6: Turn it on and watch

   Listen to the first calls live, check the stats, star anything worth reviewing, and refine the context.

What the AI does on each kind of answer

This is the part of how AI calling works that most vendor pages skip. When a call is answered, the LeadPedlar agent listens first instead of talking over the person and classifies the answer: a person, an answering machine, a silent line or a carrier intercept. Each gets its own handling, and each is recorded as an outcome on the lead.

A person answers

The AI says its opening line, confirms who it is speaking to and works toward the campaign's goals — transfer to a live person first and otherwise book a callback, or callbacks only, as the context says. It keeps its turns short, stops when interrupted, respects “not interested” and do-not-call requests, and when it hands off it passes a summary of who the lead is and what was said.

An answering machine — the message goes on the beep

If the machine action is a personalised voicemail, the AI waits for the beep, says the message written for that lead, and hangs up. If a person picks up mid-message, the message stops and the conversation starts. Gong's data is why the default is to leave one: in their study, a voicemail lifted later email reply rates from 2.73% to 5.87%.

Silence, and carrier intercepts

When a line is answered and nobody speaks — a silent line — the AI hangs up and the call is logged as no answer, so the lead can be retried within the campaign's attempts. A carrier intercept recording (number not in service, number changed) is classified as a failed or bad number and grouped separately, which tells you how clean the list was before you blame the brief.

Monitoring: live listening, stats, transfer outcomes and starred calls

An unattended campaign is not an unobserved one. The Live calls page shows every call from the first ring; on AI calls an admin can listen in, guide the AI or take the call over. Live listening is how you catch a context mistake in the first ten calls rather than the first thousand.

The campaign page's stats are plain counts: Calls today, Total dialed, Connected (calls that reached a person), Voicemails left, Transfers (calls the AI handed to a person), Callbacks booked, Sales and Sales amount, and Leads remaining (never dialed, plus retries that are due). Each chip opens the leads behind it, and each lead offers Call with AI — one AI call with the campaign's agent and context — or Call myself in the WebPhone.

The activity card has four tabs: Live, Calls, Starred and Leads. The Calls list shows every call with its hang-up reason and outcome chips to filter by — a person, voicemail, silent line, transferred, transfer failed, callback, not interested, failed. Anyone who can see a call can star it with a short note, and the Starred tab switches between Mine and Everyone; that is how a leader flags the calls worth replaying against the brief.

Results and redial

Every call leaves a transcript, a recording, a summary and an outcome on the lead's timeline, next to the CRM stage, notes, callbacks and dispositions — including the sale amount. The recorded hand-off results tell you whether the people side of a blended campaign is keeping up with the AI; a rising Transfer failed count means the rule names too few people for the hours the campaign runs.

When the list is done, Redial leads puts groups back in the queue by what happened last: the entire list, never dialed, AI left a voicemail, voicemail with no message left, no answer or busy, silent line, hung up, transferred to a person, transfer failed, callback booked, not interested, or failed and bad numbers. Leads on the do-not-call list are never put back. Redialing a running campaign restarts it with the same agent, context and calling hours, and Duplicate copies a finished campaign's settings for a new list.

How LeadPedlar combines sourcing, CRM, dialing and AI — with the autonomy you choose

The campaign above runs inside one platform rather than across five. Leads come from the lead marketplace — verified lists bought with escrow protection — or your own uploads; the CRM holds lists, pipeline stages, notes, callbacks and dispositions; the cloud dialer and WebPhone call from the browser through any SIP provider; the AI agents run outbound, inbound and blended campaigns; and teams, roles and permissions sit over all of it, so the leader who writes the context owns the list and the hand-off rule. The full list is on features; AI calling hours are priced on pricing.

The autonomy is a dial, not a switch. Fully AI-run — AI start to finish — is what this article walked through: the fewest human interventions. Human-only — People only, manual or autodial — keeps the AI off the calls entirely. Blended is the middle: AI first, people when asked for outbound lists; People first, AI steps in when nobody is free or an answering machine picks up; and inbound campaigns where the AI answers when everyone is on a call, nobody is signed in, after a set number of seconds, or outside the answering hours. A campaign's handling can be changed while it is off without losing its leads, team, window or contexts.

For the economics, see is an AI call center really cheaper?; for the qualification design an unattended first touch should follow, agentic AI lead qualification; for how an AI dialer compares with predictive and power dialers, AI dialer vs predictive dialer; and for the all-in-one argument, one platform for lead generation, closing and service.