ENTERPRISE AI PLATFORM · VOICE & TELEPHONY

AIVI

Your AI Communication Platform

AIVI answers a new lead within seconds, qualifies them over voice, SMS or email, and hands the live call to a human advisor with the full customer record already on screen. It is built for lending, insurance and financial advisory teams, where being first to respond decides who wins the deal.

Voice & telephonyMulti-channel outreachGoverned AI agentsCRM syncBilling & paymentsPolyglot monorepo

Discuss a similar project Visit AIVI ↗

< 3 secSpeed to Lead

+25–30%Contact Rate Lift

50,000+Leads Re-engaged

Three things the platform does

Everything in AIVI serves one of these. They are the product's own framing, shown to every user at sign-in.

AI Lead Engagement

Automated calls and conversations that reach a new lead in seconds and carry the conversation far enough to qualify them.

Intelligent Enrichment

Every lead scored, profiled and prioritised automatically, so advisors work the list in the right order instead of top to bottom.

Conversation Intelligence

Sentiment, intent and performance analytics drawn from the calls themselves, not from what anyone remembers to log.

HOW A CALL ACTUALLY WORKS

From new lead to live advisor

The hard part of a voice platform is not making a call. It is knowing when to stop and fetch a person, and arriving at that moment with the context intact.

A lead arrives

The agent is triggered the moment a record lands, rather than waiting for someone to work a queue.

The agent opens the conversation

Voice, SMS or email, in the agent's configured language, following the instructions set for it.

Identity gets verified

Until the caller is verified the agent can only use its public-information skills. Verification is what unlocks anything further.

It qualifies and enriches

Answers are scored and written back to the CRM as the conversation happens, so no one re-keys it later.

Warm transfer to a human

When a person is needed, the call moves across with the record and live transcript already in front of the advisor.

The call is analysed

Outcome, sentiment and scorecard results feed the dashboard and the agent's own learning window.

GOVERNANCE

Agent Foundry

Turn real conversations into governed, reusable agent capabilities.

The riskiest thing about a voice agent is that it can say or do something it should not. Agent Foundry answers that with a runtime hierarchy: skills are grouped into scopes, and a scope only opens once the caller has met its requirement. Public Information is always open; Authenticated Customer unlocks only after Verify Customer Identity passes; Employee Operations only after Verify Employee. Sensitive Transactions and Manager Approval & Accounting sit behind their own gates.

Scope map showing exactly which skills become available at each stage of a call
Skills carry a review state, from draft through review to live, so nothing untested reaches a customer
Per-agent learning controls: minimum replay threshold, calls required, and a rolling learning window
A live evidence panel where call moments stream in as conversations are analysed

Agent Foundry runtime hierarchy showing gated skill scopes

CONFIGURATION

Building an agent

Configure how this agent talks, listens, and routes calls.

Agents are set up by the team that owns them, not by an engineer. The builder walks through identity, instructions, the channels it may use, transfer rules, callable functions and the knowledge it can draw on. A draft agent cannot take real calls, which makes it safe to experiment.

Basics, Instructions, Communications, Transfer, Functions and Knowledge as separate steps
Draft, active and inactive states, with drafts blocked from live traffic
Outbound-and-inbound or inbound-only, with assigned phone numbers
A built-in phone simulator for test calls across voice, SMS, email and video before going live

AI agent configuration screen with built-in phone simulator

VISIBILITY

What the team sees each morning

Here is how your outreach is performing.

Managers get one screen that answers whether outreach is working: how many transfers happened, how many callbacks are booked, how long calls run and what share succeed, each compared against the previous period. Below that sit call volume over time, the distribution of call outcomes, transfer success, an agent leaderboard and customer sentiment drawn from analysed calls.

Transfers, scheduled callbacks, average call duration and call success rate, each with period-on-period movement
Call volume and transfer charts across a selectable date range
Agent leaderboard and customer sentiment derived from analysed calls
Workspace switching and an editable layout, so each team arranges its own view

COMMERCIAL

Subscriptions, funds and metered usage

Manage your subscription, payment methods, and billing history.

A voice platform cannot bill on seats alone, because every call and message costs real money the moment it happens. So billing here runs on two tracks at once: a subscription that sets the plan ceiling, and a prepaid balance that metered usage draws down. I implemented the payment side, including the plan selection flow, stored payment methods, top-ups and the usage accounting behind the balance.

Three tiers: Organization, Agency and Custom, each billed monthly or annually with a 20% annual discount
Plan limits covering users, phone numbers, AI agents and contacts
A prepaid funds balance with top-ups, low-balance warnings and billing history
Pay-as-you-go metering for SMS, voice calls and AI agent usage, accounted separately from the subscription

AIVI billing screen showing subscription plans and a prepaid funds balance

My role

What I worked on

I worked on the platform's web surfaces and the runtime behind the Agent Foundry hierarchy, alongside the core engineering team and product leadership.