Pre-revenue · Platform live · 3 pilot slots

Every call
becomesa booked job.

It answers, understands the business, qualifies the caller, books the slot and writes it to the CRM. Autonomously. On the first ring.

Live call 00:01:42
Agent
"Is this a repair or an install?"
Caller
"Repair. It's making a noise."
Agent
"I have Tuesday 2:30 or Wednesday morning."
Caller
"Tuesday."
QualifiedBooked Tue 2:30CRM updatedSMS sent
0Voice providers
behind one core
0Answered
no queue, no voicemail
0Customers
pre-revenue, by choice
0Pilot slots
open now
01 Don't read. Listen.

Interrupt it.
Change your mind. Twice.

Anything can be scripted into a clean 40-second video. Only a real agent survives being talked over.

02 The gap

The lead didn't convert badly.
It was never handled.

Every step after "hello" is deterministic work. It just happens to sit behind a conversation nobody had time to have.

CALL ARRIVESNobody picks upNo qualificationNothing bookedNo follow-upNo record it happenedLOST JOB
03 One agent · six actions

Not a chatbot with a phone number.
Every step writes to a real system.

01

Answer

First ring, any hour. No queue.

02

Understand

Answers from the client's own knowledge.

03

Qualify

Their script. Structured answers, scored.

04

Act

Real calendar. Books, routes or escalates.

05

Follow up

SMS and email fired by the outcome.

06

Record

Transcript and next action into the CRM.

04 Past the demo layer

The hard part was never
the first conversation.

It's tenancy, state, integrations and recovery. All of this is running today.

Platform

  • Django 5 + DRF
  • FastAPI voice relay
  • React 18 · TypeScript
  • PostgreSQL · Docker

Multi-tenant

  • Isolated data per client
  • Per-tenant agents + voices
  • Per-tenant knowledge
  • Role-based access

Conversation

  • Tool-calling drives actions
  • Tenant-scoped retrieval
  • Structured qualification
  • Human transfer paths

Integrations

  • Calendar + busy time
  • SMS & email follow-up
  • CRM record writing
  • Inbound + outbound
05 Architecture

Models commoditise.
Workflow doesn't.

Workflow engineours
Orchestrationswappable
AI modelscommodity
Telephonycommodity
01
Workflow engineQualify · Book · CRM · Follow-up · Consent · Audit · Escalate
Ours
02
OrchestrationTwilio · Telnyx · Retell — three paths, one core
Swappable
03
AI modelsImproving for everyone at once
Commodity
04
Telephony networksNo advantage lives down here
Commodity

Everything above the second line stays with us: the context, the actions, the customer data and the business logic between the call and the outcome. A provider change is an operational task, not an existential one.

06 Defensibility, honestly

No customers yet.
So no moat claims.

Five things we believe compound — each stated so it can be proven false, with the test that settles it.

01

Workflow IP

The value is the work behind the call, not the talking.

TEST → 2nd vertical ships reusing >60% of the engine
02

Playbooks

Conversations generalise within an industry, not across all.

TEST → pilots 2 & 3 live with no bespoke build
03

Context

Per-tenant knowledge gets expensive to rebuild elsewhere.

TEST → a pilot declines to migrate after the free period
04

Eval data

Knowing where calls fail beats owning the model.

TEST → task success rises on a fixed model version
05

Abstraction

Staying above infrastructure keeps margin and optionality.

TEST → live tenant moved providers, invisibly
07 Why now

The category question
is already settled.

0M
Vapi · May 2026

Valuation after crossing one billion calls, and being chosen by Amazon Ring over 40 rivals.

0B
Parloa · Jan 2026

Tripled in eight months on a $350M Series D. Enterprise budget is moving, fast.

0
Of them sell to a plumber

Those valuations chase enterprise contact centres. Nobody at that price point serves six vans.

That last number is our honest answer to "why won't Vapi just do this." It isn't technology. It's cost structure and customer size — and a team in Dhaka and Sacramento can serve a market they can't profitably reach.

08 Where we are

Stated plainly,
so you don't have to dig.

Shipped7
  • Multi-tenant platform, isolated data
  • Inbound + outbound call flows
  • Per-tenant knowledge and rules
  • Calendar booking with busy time
  • SMS and email follow-up
  • CRM records from the conversation
  • Three providers behind one core
In progress5
  • Hardened public demo agent
  • Agent evaluation + observability
  • Three pilot deployments
  • Vertical templates for the wedge
  • Product screenshots
Planned5
  • Metered billing, self-serve onboarding
  • Multi-location and franchise
  • Partner and reseller distribution
  • Additional languages
  • Enterprise controls, certifications
09 Compliance posture

Controls, not
a legal guarantee.

The FCC treats AI voices as artificial voices under the TCPA. No software can guarantee compliance — it depends on how a business uses it. What software can do is make the safe path the default and every call auditable.

Consent controlsChecked before any outbound call
Opt-out & DNCHonoured across channels
Call audit historyTranscript, outcome, rules applied
Human handoffOn request or on trigger
Tenant isolationNo data crosses organisations
Data retentionConfigurable; recording optional

This is also why our wedge is inbound. When the customer places the call, most of the consent question does not arise. We hold no SOC 2, HIPAA, PCI or ISO 27001 and do not claim to.

10 How it makes money

Revenue grows with call volume.
Not with headcount.

The same platform serves every tenant. Adding a customer adds usage, not staff.

PlatformMonthly subscription per business. The recurring base.
Subscription
UsageVoice minutes and messaging. Scales with how much work the agent does.
Subscription + usage
ExpansionMore agents, more locations, deeper integrations. Grows inside the account.
Subscription + usage + expansion
$299Starter
1 agent · basic CRM
$799Growth
Multi-agent · integrations
$1,999Pro
High volume · advanced automation
CustomEnterprise
Private deployment · SLA

Planned launch pricing, per month. It is a management assumption until pilot cohorts validate it — which is also the only way to know real gross margin across three voice providers.

11 Where this goes

Start narrow.
Earn the right to widen.

Every stage runs on the same engine. The expansion is conversation types, not a rebuild.

Now · Wedge

Inbound calls

Phone-driven service businesses where a missed call is a lost job. Answer, qualify, book, follow up.

Next · Expansion

Every other conversation

Support, reminders, retention calls, recruitment screening, order status. Same engine, new scripts and actions.

Then · Platform

The record of what was said

Every customer conversation captured, structured and acted on — across phone, SMS and chat, in one place.

12 Who's building it

Built from a failure
we kept watching from the inside.

NA

Nayem Abedin

Founder & CEO

Since 2019 he has built software, automation and growth systems for service businesses in the US and Bangladesh. The same failure appeared in almost every engagement: the client paid to make the phone ring, then lost the lead at the moment it rang. We built the voice system to close that gap for one client. It became clear it was a product, not a project.

Operating since 2019NVIDIA InceptionBootstrapped

A small core team, in-house. Not a marketplace, not rotating contractors — the people who built it are the people who run it.

AI & full-stackAgents, tool-calling, console
Backend & platformTenancy, state, voice relay
IntegrationsCRM, calendar, messaging
QA & product opsCall testing, release gates
DhakaEngineering
SacramentoCustomers & go-to-market
US & UK firstThen GCC & South Asia

Three pilots.
Then numbers, not theories.

Three phone-driven service businesses on real inbound traffic. Free while we learn — in exchange for a measured baseline and permission to publish the result.

Nayem AbedinFounder
0Operating since
NVIDIAInception member
SAC · DACSacramento + Dhaka