Pravar
Contact Download CV
Case Study 02·2025–Present·AI Product & Human-Centered AI

Lisa

AI as a teammate. How call intelligence became a digital receptionist that makes sure no small business misses an opportunity.

Scroll to begin

Act 01Context

Small business owners miss opportunities every day. Not because they don't care. Because they are busy.

Running a business often means being simultaneously the owner, the operator, the sales representative, the receptionist and customer support, all at once. The phone rings while they're driving, while they're working, while they're on another call, while they're asleep.

Every missed call is potentially a missed customer.

The question was never "how do we build an AI receptionist?" It was: how do we help small businesses stay available when they can't be?

Chapter theme

Every earlier lesson about trust and confidence, converging into one teammate.

The goal was never to build an AI receptionist. It was to make sure a small business never missed an opportunity simply because nobody picked up the phone.

Executive Summary

A digital front desk, not another feature.

Challenge

Business communication stopped when the owner became unavailable. Traditional IVR systems created friction, and hiring a dedicated receptionist wasn't realistic for most small businesses.

Breakthrough

Instead of building another AI feature, the team designed an AI teammate: a digital front desk that could answer questions, capture leads, transfer calls, and represent the business.

Outcome

Lisa evolved from a missed-call assistant into a full AI receptionist, eventually becoming a superset of Auto Attendant itself, and the first AI product LinkedPhone fully adopted.

The evolution

ConfigurationTraditional settings and forms
AssistantA tool that answers questions
EmployeeSomething you onboard, not configure
LisaAn AI teammate, trusted with real calls
Timeline
2024–25Present
Platforms
iOSAndroid · Web · Voice
Category
NewBusiness communication
Discipline
AI ProductHuman-centered
Legacy
SynthesisOf trust and AI

Where We Were

A long chain of work, already happening behind the scenes.

Before Lisa existed, the journey had already moved from call recordings, to transcripts, to summaries, to action extraction. The call stopped being a recording. It became structured information.

Even after all of that, one problem remained: communication still depended on humans.

If nobody answered, nothing happened. The opportunity was lost. For large companies, staff solves this. For most small businesses, it doesn't.

Concept · A business owner, surrounded

Phone, laptop, invoices, tasks, ringing

The everyday reality behind every missed call.

Act 02Problem

The Friction

"I'm busy." "I'm driving." "I missed the call."

Users repeatedly faced the same situations. For large companies, this problem is solved with staff. For small businesses, it often isn't.

Investigation

"Can AI answer calls?" became "should it?"

At first the goal wasn't building an AI employee. The team was experimenting with voice AI infrastructure out of curiosity: what if AI could participate in calls, answer simple questions, collect information? The more it was tested, the more potential appeared, and the question evolved.

Artifact · Early voice AI experiments

Curiosity, before conviction

The earliest tests were about what was technically possible, not what should ship.

Insights

Nobody wakes up wanting AI. They want their time back.

Insight 01

Nobody wakes up wanting AI.

Observation
Customers never asked for voice models, agents, or automation by name.
Meaning
They wanted more time, more availability, less stress.
Implication
The interface should sell the outcome, not the technology.

Insight 02

Businesses already have the knowledge. It's scattered.

Observation
Websites, PDFs, brochures, social media all contained the same answers customers needed.
Meaning
Onboarding didn't need to start from zero.
Implication
Let Lisa learn from what already exists, instead of building it manually.

Insight 03

Settings create anxiety.

Observation
Traditional configuration screens felt technical and cold in testing.
Meaning
People don't want to configure AI. They want to onboard a teammate.
Implication
Design the entire interface around a profile, not a settings page.

Insight 04

Trust builds gradually, or not at all.

Observation
Businesses weren't ready to hand over every conversation immediately.
Meaning
Confidence needed a safe, low-stakes place to start.
Implication
Start with missed calls only, expand from there.

Act 03 · Turning Point

We thought users wanted AI.

They wanted a teammate.

The challenge became: how do we make AI feel useful before it feels intelligent? That question shaped everything from the interface to the character itself.

Exploration Lab

Configuration, then a tool, then finally a person.

Direction 1

Traditional Configuration Screens

Failed
Why it looked promising
  • Familiar enterprise pattern
Why it failed
  • Felt technical, cold, complicated
Approachable
Trustworthy

Direction 2

AI Tool

Partial
Why it looked promising
  • Feature-centric, functionality-first
Why it fell short
  • Users saw technology, not value
Approachable
Trustworthy

Direction 3 · The winner

Digital Employee

Succeeded
Why it looked promising
  • Profile-centric, personality-first
Why it succeeded
  • Users instantly understood it
  • This direction survived every test
Approachable
Trustworthy

The percentages above are illustrative evaluation scores from the exploration process, not measured data.

Breakthrough

"Here's my website." "I've learned about your business."

The biggest realization came from onboarding. Most competitors ask users to configure, define, train. LinkedPhone discovered something different: most businesses already have the information, it's simply scattered.

So instead of building Lisa manually, users could talk to her, upload content, share links. Lisa would build herself.

Scattered knowledgeWebsite, PDFs, social media
A conversationShare it, instead of typing it in
Lisa, self-builtOnboarded, not configured

We thought users wanted AI. They wanted a teammate.

Act 04System

System Architecture

Knowledge, conversation, routing, automation, together.

Answer questions

Capability

Capture leads

Capability

Transfer calls

Capability

Schedule conversations

Capability

Handle missed calls

Capability

Represent the business

Capability

Humanizing AI

The character came after the architecture, not before.

As testing continued, the team noticed users naturally referred to the assistant like a person. That inspired the mascot, the identity, the name. Design goals: friendly, capable, approachable, intelligent, slightly nerdy, trustworthy. The final avatar: a white blazer, a blue shirt, blue goggles, subtle LinkedPhone branding.

Because businesses hire people, not features. When users saw Lisa, they immediately understood: this person helps answer my calls, without reading documentation.

Product Showcase

A profile, not a settings page.

Screen · Lisa's dashboard

Designed like a teammate profile

Avatar, confidence indicators, capability cards, an onboarding flow. Never a settings page.

Screen · Knowledge onboarding

"Here's my website." "I've learned about your business."

Onboarding by conversation, not by form.

Screen · Call handling

Missed calls first, then all calls

Trust extended gradually, the same way it's built with a new hire.

Interaction & Motion

Why Lisa looks human.

Not to create emotional attachment. To create instant understanding. A receptionist, an assistant, a teammate, a coordinator: familiar roles that need no explanation.

Act 05Impact

Validation

Missed calls first. Then, all of them.

The first rollout only handled missed calls, not every conversation. Trust, not technology, set the pace. Adoption was strong: users actively enabled it, used it, returned to it, and expanded usage.

Confidence increased over timeBusinesses that started with missed calls expanded on their own schedule.
Lisa moved from missed calls to all incoming callsA milestone that only happens once trust is genuinely earned.
Auto Attendant's usage patterns shiftedLisa began absorbing the role Auto Attendant used to play alone.

Ripple Effects

Lisa changed more than AI. It changed the company's philosophy.

Lisa Auto Attendant Knowledge Systems Onboarding Website Strategy

Lisa reshaped Auto Attendant, knowledge systems, onboarding and website strategy. Most importantly, it changed how the company understood AI itself.

Archive

The evidence vault.

Character exploration wall

Figma

Onboarding prototypes

Figma

Voice AI experiment recordings

Video

Naming and brand explorations

Doc

Adoption research

Doc

Founder reviews

Notes

Lessons

What we believed. What we learned.

Lesson 01

AI is not the product. The outcome is.

Nobody adopts intelligence for its own sake.

Lesson 02

People trust teammates faster than technology.

Framing changes adoption more than capability does.

Lesson 03

The best AI interfaces don't feel like AI interfaces.

They feel like the thing they replaced, just more available.

Lesson 04

Automation should remove effort, not add configuration.

If onboarding an AI feels like a chore, it's designed backwards.

Legacy

A voice layer became a way of thinking.

Lisa began as an experiment layered on top of phone calls. It became a digital employee, a front desk, a receptionist, a knowledge system, a communication layer.

And ultimately, the first time LinkedPhone stopped asking "how do we manage communication?" and started asking "how do we represent a business when nobody is available?" Auto Attendant taught the company to organize complexity. The Work Model taught it to organize outcomes. 10DLC taught it to organize operations. Lisa is where all of it became a teammate.

Closing

The goal was never to build an AI receptionist. It was to make sure a small business never missed an opportunity simply because nobody picked up the phone.