File 001 · Part IITagent
For Andy Lee

Forty stories on tape. One on your feed.

Our first file found forty stories in your interviews and none on your LinkedIn. Since then, one has made it to your feed: the year you left Alorica to be Gateway's CIO. It got 302 likes, more than any other post except the CXP launch.

40 stories on tape1 posted · the Gateway year
Your last 26 posts

Your audience already voted.

Your last 26 posts on LinkedIn, sorted by what each one is about. Each dot is one post. Further right means more likes.

Three of your four most-liked posts are about you. The most common kind of post, 10 of the 26, is about someone else: other companies' new CX leaders, Meta's support hires, Medicare plans, USC alumni, Target's job ads. Those get a median of 44 likes. Posts about you get 196.

"Was there a formula here, or did we just get lucky?"Your question for every win · your own post

A formula. Of the 26, the Gateway post is the only one with a moment only you lived through and a choice that cost you something. Your interviews hold forty more like it, and we've found more since.

What's on tape

Eight years of you, mapped.

Every point is 30 seconds of an interview you gave between 2018 and this September: 676 moments, about 63,000 words. Lines join moments that share a person, a company or an idea. The grey strand off to one side is your feed. Pick a story to see where it comes from.

On tape Your feed This month's news Six stories
Loading the tape
Hover to read · drag to explore · scroll to zoom
What we'd talk about next

Six stories for the next month.

Each one is a moment from your past meeting something in this month's news, and each one ends at Crescendo. Every fact about you is from your own interviews, with the minute you said it.

01

The caller who never gives up

How it could open"For decades, a lot of customer service counted on one thing: the customer giving up."

Then the 90s: five companies, nobody responsible Now Bain on customers' personal AI agents

Then · the 1990s

You watched people buy a PC, a monitor, a printer and AOL, then call five companies to make them work together, and hear from each one that it was somebody else's problem. "These companies are not taking accountability," you said, and that's why you started Alorica. Later the industry learned to manage cost with deflection: phone trees, self-help, slower pickup, "all to kind of manage to a budget."

Extrology, Sep 2026 · 9:33 · TechSurge, Nov 2024 · 5:20
Now · November 2026

Bain wrote this month that customers' personal AI agents are already contacting companies to dispute bills, chase refunds and cancel subscriptions. The service leaders it spoke to couldn't say how many of their contacts came from one.

Bain & Company, "When Your Customer Sends an Agent"

A personal agent doesn't get tired. It sits through the phone tree, takes the third transfer and calls all five companies at 3 a.m. Deflection was a budget decision that worked on people who give up, and the new caller never does. The only strategy left is the one Alorica started on in 1999: take the call and fix the problem.

For CrescendoThe case for resolving instead of deflecting, with Crescendo arriving only at the end.

02

Your agent and theirs

How it could open"A personal AI agent needs instructions. A company's AI agents need a culture."

Then 100,000 people and one reason Now personal agents, and a company agent switched off

Then · 100,000 people

You woke up one day running about 100,000 people in 17 or 18 countries, and couldn't control what each of them did on every call. What you could do was give them one reason the company exists, "to make lives better," and the principles to decide with. You posted those principles this fall: "I've stopped trying to control what my teams do."

Extrology, Sep 2026 · 16:50 · your post on the nine principles
Now · this year

Personal agents went mainstream: Google launched Gemini Spark in May, an agent you email like a colleague. On the company side, in July a subscription brand switched off its customer-service AI agent after it wrongly confirmed to a customer that their price was about to more than double.

Google I/O, May 2026 · CX Foundation, Jul 2026
Your personal agentA company's agent
Works forone personevery customer who calls
Runs onyour instructionsthe company's principles
Remembersyour inboxevery conversation, on every channel
When it's wrongyou fix itthe company owns it

When your own agent gets something wrong, you notice and fix it. When a company's agent gets something wrong, a lot of customers meet the mistake before anyone inside notices, and the company owns every word. You learned with people, long before AI, that you can't write a script for every conversation. A company's agents need what the 100,000 needed: one reason, shared principles, and a memory of every conversation.

For CrescendoYour own CXP launch post describes agents that "share context from every interaction and pass what they learn to each other."

03

Most returned computers weren't broken

How it could open"In the 90s, a lot of the computers people sent back weren't broken."

Then returned PCs that worked fine Now Adobe's AI-assisted holiday season

Then · age 21, then 1999

For a PC maker, a return was the most expensive thing that could happen, and a phone call was cheap by comparison. Products came back, got tested, and "there was actually nothing wrong with it." So your bet was to spend 10, 15, even 20 minutes on the phone helping the customer, because that still cost less than the return. You'd learned to hunt for that pattern at 21, putting sensors on your father's factory floor to find out why returns spiked at certain times of the year.

Extrology, Sep 2026 · 2:48 · 11:14
Now · holiday season

Adobe expects a record $275.1 billion in US online holiday sales this year, with traffic from AI assistants to retail sites up about 130%. In Adobe's research, 69% of shoppers said they're less likely to return something they bought with an AI assistant's help.

Adobe Digital Insights, holiday forecast, Sep 2026

The conversation that stops a return used to happen after the box came back, if it happened at all. Now it can happen before anyone clicks buy, for every shopper instead of the few who call. It's the idea Alorica started on, with the timing finally fixed.

For CrescendoOn tape you say that when Crescendo's AI is in the conversation, clients see baskets 20 to 30% bigger and online conversion up 15 to 25%. Best posted on Cyber Monday, or the week after Christmas.

04

The $2 billion company that didn't change the industry

How it could open"I built a $2 billion company and still failed to change my industry."

Then $10,000, alone, billed by the hour Now software priced per resolution

Then · 1999 to 2024

You saw that the whole business ran on labor hours, while clients wanted to buy an outcome for their customers. You started with $10,000, never raised outside money, and did it alone. Alorica grew to $2 billion, and the industry still billed by the hour. When you asked yourself why you'd failed to change it, your answer was that changing an industry takes "the power of many."

TechSurge, Nov 2024 · 4:10 · 35:16
Now · 2026

This year the biggest customer-service software companies started charging per resolved conversation instead of per seat, from about 50 cents to $2 a resolution.

Futurum · TechRadar, 2026

Billing by the hour pays the provider for slow work. Billing by the outcome pays for problems fixed. It took AI to make every conversation measurable enough to price that way, 25 years after you wanted to. This time you aren't doing it alone: Crescendo was started with General Catalyst and people who built some of the best-known contact-center software.

For CrescendoExplains outcome pricing and why you co-founded Crescendo, from the one person who can call it a second attempt.

05

The customer has the better memory now

How it could open"For the first time, customers can have a better memory than the companies they call."

Then 20 forms at the doctor's office Now agents that remember everything

Then · always

You've always asked why you fill out 20 forms at a doctor's office, write your name 20 times, answer the same question ten times, and where all that information goes. Your feed said the business version this fall: support teams run on six or more systems that were never built to talk to each other, so the customer ends up repeating themselves.

Extrology, Sep 2026 · 1:07 · your post on why CX software is fragmented
Now · this year

Personal agents now carry a person's whole history. Gemini Spark, launched in May, works from your inbox around the clock.

Google I/O, May 2026

When the customer's agent remembers every order, every promise and every earlier call, a company that makes them start over looks careless. In your own words: "As humans, our memories are not infinite," while AI "has infinite memory. It never forgets." The question for every company is which side of the call remembers more.

For CrescendoFrom your launch post: "a customer moves from chat to voice without starting over."

06

An AI roll-up before it had a name

How it could open"Nine months after we started Crescendo, we bought a company with 3,000 people."

Then a hundred companies merged Now the AI roll-up

Then · Crescendo, month nine

Crescendo bought PartnerHero, a business of about $80 to $85 million a year and more than 3,000 people, to use as a sandbox: a company already serving the customers you wanted, where the model could be proved. You'd done this many times at Alorica (counting the companies those companies had bought, about a hundred), and you spent the first 30 days on one thing: getting everyone behind one mission.

TechSurge, Nov 2024 · 30:45 · Lakshmi Rebecca, Mar 2026 · 1:57:19
Now · the new playbook

Venture firms now have a name for it, the AI roll-up: buy a services business and rebuild it around AI. General Catalyst set aside $1.5 billion of its latest fund for the strategy, and its own write-up points to PartnerHero.

TechCrunch · General Catalyst, "The future of services"

The model is new to most investors. Merging companies isn't new to you. On tape you say integration is where it hurts, and that you're now "very, very clear on what is required to ensure a really clean integration." That's the part of a roll-up nobody can buy.

For CrescendoPuts Crescendo first in a trend investors are now chasing, told by the person who has done the hard part most often.

"It's such a noisy place. Everybody's talking about AI. Right now we're really working on getting our story out."

You, on Extrology · September 2026 · 39:58

We think the fastest way to get Crescendo's story out is to tell yours, starting with these six.

The story of the attorney who came home one day and quit her firm stays yours and Joyce's, until you both decide otherwise.

Write to Nitesh
  • One conversation every two weeks. Sixty to ninety minutes. We record it and listen for the stories.
  • Two posts a week, built from your tape and that week's news, in your voice.
  • Nothing goes out without your yes.