PulseDial
DemoBook a demo
A contact center at dusk, with human agents and humanoid figures standing in for AI voice agents at their desks

The bot opens the call. The person closes.

PulseDial is the software where your AI agent and your human team work the same campaign: outbound dialing, intelligence on every conversation and the cost of every call in view.

You connect your AI voice agent, your CRM and the carriers you already use.

Humans, AI, or both on the same call.

It isn't a dialer with a bot bolted on the side. It's telephony and dialing where the AI agent and the person are the same kind of agent.

Contact center agent in a headset, smiling in front of their monitor

Human agents

Predictive dialer, inbound queues and a live wallboard. Your team works the way it always has, with the cost and the recording of every call in view.

Humanoid figure in a headset taking a call at a contact center station

AI voice agents

You connect your Retell, Vapi, ElevenLabs or 008 Agent account and the agent dials, answers and takes queue calls like any other. Around the clock, without taking up a seat.

Human agent and humanoid figure shaking hands over a desk with the platform on the monitors

Both, together

The bot opens the call and hands it to the person with the context already on screen. The customer repeats nothing and the agent comes in knowing what happened.

The dialer was born with AI inside.

This is not a dialer with channels bolted on. Channels, AI agents and SIP trunks come and go without rebuilding the operation.

Cards for two call attempts, a WhatsApp thread answered by AI and a confirmation SMS

AI First Omni Dialer

Voice is one channel among others, not the only one. Phone, WhatsApp, SMS and Telegram over the same contact and the same record, and every channel takes an AI agent or a person.

  • Four channels over one contact and its history
  • Each channel runs an AI agent or a person
  • The attempt jumps channel when nobody picks up
  • Voice
  • WhatsApp
  • SMS
  • Telegram
Telegram thread where the bot returns campaign status and confirms an agent was paused

Native MCP, not one more report

Plug in your LLM or your agent and run the contact center by talking. Not just status: pause an agent, create a campaign or move a queue from wherever you already work.

  • Live status: queues, ASR, cost and agents
  • Real actions, not dashboard reading only
  • The MCP sees no more than its own user

/status campaign

/pause agent 204

/new campaign

Panel of SIP trunks with their contact rate and automatic selection turned on

Carrier agnostic, AI agnostic

Bring your own SIP trunks and your own AI agents (Retell, Vapi, ElevenLabs, 008) and the platform uses them. No exclusivity, no captive minutes, no mandatory voice engine.

  • Several carriers at once, one per country if needed
  • The engine picks the trunk with the best contact rate
  • Switch provider without rebuilding your campaigns
Available trunksEngine on
  • Troncal MX-0198.2%
  • Troncal MX-0295.1%
  • Troncal MX-0387.5%
  • Troncal CO-0172.3%

Three things break at once.

Contact rate decays on its own

Numbers wear out, voicemail eats the attempts and nobody knows which of the contact's eight phones to call.

  • Quality per number, attempt by attempt
  • Caller ID rotation
  • Answering machine detection before connecting

The cost shows up at month end

The carrier invoice arrives once the campaign is over, and by then the margin was either there or it wasn't.

  • Per-prefix rate, with zone and group
  • Cost visible on every call
  • Retroactive re-rating

Compliance is proven too late

When the request or the audit arrives, reconstructing how someone was spoken to takes days of manual work.

  • Consent history that can't be edited
  • Suppression lists checked before every call
  • Audit trail on every live supervision
Contact center agent in a headset, focused on their monitor

Five different operations, one platform.

Each client, its campaign, its cost

You operate for several clients with different SLAs and you need each campaign's margin without waiting for the carrier invoice.

  • A-Z per-prefix rates, with retroactive re-rating
  • Capacity planning with Erlang C, per interval
  • Roles and permissions per team, plus campaign templates

Six things it does, with the proof.

Dialing strategy panel in preview mode with a grace period control

Pacing adjusts without stopping the campaign

Preview, progressive and predictive. You change the pacing preset or the pre-dialing while it runs and the dialer readjusts on its own.

  • Grace period configurable from 13,000 to 30,000 ms
  • The panel shows why the last automatic adjustment happened
  • It also changes by API, without opening the interface
Contact record with eight phones, two discarded, next to the contact's local time

The best phone, at the right time

Each contact holds up to eight phones, each with its own quality. On every attempt the dialer picks the best one still alive and drops the dead ones.

  • It dials within the contact's time window, not yours
  • Caller ID rotation: LRU, round-robin, random or fixed per contact
  • When attempts run out, it enrolls them in WhatsApp or texts the best number
Voice agent call in progress, its transcript and a quality score of 87 out of 100

AI that talks, and AI that listens

Your voice agent dials and answers like any other agent. Separately, conversation intelligence transcribes, summarizes and suggests a disposition, and an automatic judge scores quality.

  • The AI agent uses your Retell, Vapi, ElevenLabs or 008 Agent account
  • Conversation intelligence ships turned off, feature by feature, with a daily spend cap
  • A human can override the judge, and that score stands
  • Answering machine detection adds two to four seconds before connecting
Per-prefix rate table and the breakdown of one call's charge

You know the cost today, not next month

Per-prefix rates with a connection fee, per-minute rate and billing mode. Every call stores the cost, the zone and the group it was rated with.

  • An unanswered call costs zero, with no connection fee
  • Ringing and talk time broken out in the call detail
  • Retroactive re-rating over a full date range
Wallboard with five agents and their status, the Erlang C forecast and an SLA alert

You see the operation and can step in

Real-time wallboard of agents and queues, with wall mode and an audible alarm. You listen, whisper or barge in on any call without cutting it.

  • Every listening session is audited: who, which call and in which mode
  • Agent forecasting with Erlang C, interval by interval
  • Alerts to Slack or email for a trunk down, low SLA or high abandonment
Compliance panel with the consent history and a preview of a data purge

Compliance is written down, not promised

Suppression lists checked before every call, lawful basis per contact and a consent history that only appends entries and never edits them.

  • Access requests resolved from one screen, searching by any of the contact's phones
  • Deletion previews what will be erased, table by table
  • Platform-wide audit trail and mandatory second factor on sensitive profiles

It isn't only a dialer. It answers too.

Inbound queue screen with the skills it requires, the routing strategy and the agents who match

Queues that really enforce the skills

An agent needs every skill you mark to take the call. Five routing strategies, priority from one to ten and a boost for VIP customers.

  • Whisper to the agent before connecting, announcing the queue
  • Last-agent routing, automatic or by pressing one
  • Overflow if no agents are free or several are already waiting
Waiting queue with a caller who hung up keeping their place and their callback already scheduled

Nobody loses their place by hanging up

The caller hangs up keeping their place and the platform calls back when their turn comes.

  • Configurable SLA, eighty percent in twenty seconds by default
  • Saturation alert to the supervisor
  • Rescue by WhatsApp or SMS to the best usable phone

The platform, from the inside.

This is the console with the operation running. Pick a capability and it opens just as your supervisors use it.

You pick the dialing mode, the number pool, concurrency and the outbound trunk. The prompt comes from pre-recorded audio or text to speech, with each contact's variables.

New campaign screen with the number pool, the maximum concurrency and the voice provider selected

How it gets going.

Six steps, in this order. Each one leaves something working before you move to the next.

  1. 1

    You connect your SIP trunk

    You register the trunk and test it for real, with inbound and outbound rates assigned.

  2. 2

    You register the agents

    Users, skills and profiles. Second factor stays mandatory on the sensitive profiles.

  3. 3

    You upload the contacts

    Up to eight phones per person, with their lawful basis and their suppression lists loaded.

  4. 4

    You prepare audio and scripts

    Pre-recorded prompts, text to speech or your AI voice agent connected, and the call script the person sees on screen.

  5. 5

    You launch the first campaign

    You pick the dialing mode, the pacing strategy and the caller ID pool, and start.

  6. 6

    You plug in your integrations

    CRM in both directions, signed webhooks and your own storage for recordings.

Questions.

Does it work with my current SIP provider?

Yes. You register your own SIP trunks, test them before using them and assign inbound and outbound rates to each one separately.

Can I keep the recordings on my own infrastructure?

Yes. Automatic daily archiving to Amazon S3, Wasabi, Azure Blob or your FTP or SFTP, in copy or move mode. Credentials stay encrypted at rest.

How do I handle an access or deletion request?

From a single screen, searching by any of the person's phones. You generate the data package or run the deletion, and the operation is audited with a hash of the number, not the number.

Can I use my own AI voice agent?

Yes. You connect your Retell, Vapi, ElevenLabs or 008 Agent account and the agent joins the campaign or the queue like any other agent, with its recording and its cost per call.

Is the AI mandatory?

Conversation intelligence comes turned off and turns on feature by feature, with a daily spend cap. The voice agent is separate: you connect it if you want, and if not, your team runs with people only.

How is abandonment controlled in predictive?

With pacing presets and four parameters you adjust live, plus a panel that shows abandonment today, abandonment over the last five minutes and why the last automatic adjustment happened.

Does it integrate with my CRM?

Salesforce, HubSpot, Zoho and Zendesk in both directions, each one switchable on its own. For everything else, a generic signed webhook and an API with permissions per resource.

What happens with someone who never answers?

When the attempts run out you can enroll them automatically in a WhatsApp sequence or send them an SMS, always to the best usable phone they have left.

Can I know what a campaign cost me?

Yes, call by call, with the zone and the billing group it was rated with. And if prices change later, you apply a retroactive re-rating over the full date range.

We show it using your own operation.

Thirty minutes, no generic deck. Bring a real campaign and we set it up in the demo.

Book a demo

We show you the platform with your case on the table. Thirty minutes.

Who you are
Your operation
optional
optional

By submitting your details you accept our privacy notice.