Zia Chishti: A Pakistani-American Startup Legend

Even as I sit here talking to you, it might seem interesting, but there’s probably a kernel of doubt: “Maybe he is just taking an optimistic view.” Or “he’s half Pakistani, so he’s got some other motive.” Going there is important.

Afiniti “pairs people in an enterprise context.” How, exactly?

When people hear that line about pairing, they often say, “That sounds like Tinder or Match.com,” and that’s actually good insight. Except that our application of that matching process exists in the enterprise domain, and solving a different problem. If you’re on Tinder, you can swipe … which way is it? Is right good? [Laughs]

I think so.

Whatever is the good way to swipe, you can swipe that way 100 times on those apps and have 100 potential matches. In the enterprise context, you have five calls and five agents – you can’t assign one to more than one.

This one-to-one matching has multiple incarnations for an enterprise, by the way. The way people are most familiar with is how we optimize calls. But if you think about it, there are so many other ways that an enterprise interacts with its customers. You can have a human sales force, for example, who are assigned based on geography: west of the Mississippi, east of the Mississippi. Or seniority, “I’m the CEO, I talk to the CEO, my counterpart. You’re the junior sales guy, you talk to procurement.” That’s a good start. But aren’t we missing a critical component of that? What if we examine the behavior of the sales person, examine the behavior of the customer, and then pair also based on that?

Or say you call a cable company because you have a technical issue. They then have to send a truck to your home. Which truck do they send? Normally it’s the shortest possible geographic route. But if you could analyze the behavior of the truck technician and analyze the behavior of the individual that you’re going to see, couldn’t you pair better? So you send a truck that’s slightly further away, but hit upon interpersonal pairings that are more efficient.

Was Afiniti inspired by a poor experience you had on a call?

No. My PE firm [TRG] owned a call-center outsourcer. These are hard businesses to run. They’re very close to marginal costs, highly commoditized. So we sat down and said, “How can we optimize this business so that we actually make some money on it?” We examined how you win clients, how you serve clients, how you train and hire agents, how you compensate. We concluded that there were some optimizations to be had. But the most obvious one was how calls flowed. Because nobody had challenged the orthodoxy for 40 years.

For 40 years calls flowed in the order received. If you were the first person who called, you got the first available agent. And I said, “That just can’t be right. There’s gotta be more information that we can bring to bear that can improve that outcome.”

It took us four years before we had an algorithm that worked at all. And I remember exactly the moment we cracked it. I was going up a set of stairs – actually here in Manhattan – to see a friend. And I was mid-way up the staircase, thinking about this problem. And I was like, “Oh, I know what I need to do.”

That makes me feel badly about what I think about when I’m climbing a set of stairs. Did you see your friend? Or leave to write code?

I got to my friend’s place and picked up the phone and called our chief scientist. He said, “that’ll never work.” But I told him, “Just try it. Try it.” The intuition was around perfect squares. I won’t spend too much time, but I’m a math geek so I’m going to have to throw in my two cents.

If you think about behaviors, your intuition may be to pair like with like – if I have an outgoing personality, pair me with an outgoing personality. That’s not actually correct. But if you can characterize behaviors in a certain way, then you can fit them together where larger impact behaviors are paired, as are the smaller impact behaviors. The closer you can align those, the bigger result you get.

Now, the best use of the circumference of a four-sided object is a square – that grants you the
greatest possible area. If you have a two-by-two square, that’s an area of four; whereas if you have a three by one rectangle, that’s an area of three. So the closer you can get to a perfect square is the closest you can get to optimizing for area. And I realized, what if we could classify behaviors for two people in a way such that you can line them up in a perfect square?

What’s your response to leaders who say, “We’ve heard how AI is going to transform our business. It hasn’t. Why are you any different?”

I’m in that camp myself. We’re big AI skeptics at Afiniti. We think it’s all kind of a bubble and a mania. It just doesn’t do anywhere close to the things that people think it does. It’s just a bunch of algorithms used to find some patterns. And it’s clearly blown way out of proportion. Another in that vein, by the way: blockchain. This is unfortunately part of a generalized hype engine that Silicon Valley pumps out.

But what I tell leaders is that we’re not “AI.” I walk them through how we’ll measurably add millions – or depending on their size, billions – of dollars in revenue. And how we’ll do it in a way that leaves their customers and employees happier.

People tend to have such an implicit, immediate negative reaction to algorithms shaping human behavior. How do you overcome that?

I think time is in our favor. I recall a conversation I had seven or eight years ago with a large credit card company. We walked a leader through our pitch and, with this dripping look of disdain, he said something to the effect of, “We would never vary how we interact with our customers. They’re all valuable and important to us.”

And he kind of smirked and basically told us to get out of his office.

I tried to re-purpose that. I said, “We’re making the experience for everybody better.” But he couldn’t get over the hang-up that treating people differently was inherently wrong.

Fast forward six years and I’m meeting the same executive. And he said, “We’re really into customizing the customer experience and micro-targeting these days. We understand that you guys can help us in this journey.” [Laughs] I try not to let the irony of those two statements rile me too much.

There’s growing recognition that the information you have about people leads to better outcomes for those people. And if you manage your technology well and thoughtfully, everybody can be better off. This is a parade of optimal outcomes.

That is the dominant narrative now, or at least getting to be the dominant one. We’re headed in the right direction.

That’s quite a change – from “treating people differently” to “customizing customer experience.”

Exactly. “Micro-targeting.” [Laughs] The difference between 2010 and 2018.

What’s next? Billion-dollar business No. 4?

This is my last real run at business. I’d like to turn to something more philanthropic. I’ve always been interested in education. And I’ve been wanting to build a peculiar kind of school for a long time.

Peculiar how?

In countries like Pakistan, it’s insufficient to have a school that’s free. Government schools are free, but there still remarkably low levels of literacy. The reason is that a lot of Pakistanis rely on subsistence farming. They can’t feed everybody if they send the kids to school – so they don’t send the kids to school.

The insight is that you could pay the families to send their kids to school. If you look at a country like Pakistan, I’d say that’s a potential target population of ten million kids that we could send. I suspect we’d get a few million applications if we gave out $1,000/month, which in Pakistan is a serious amount of money.

Out of those applications, I’m confident I could find 20 or 30 absolute stars and give them 15 or 20 years of education. That’s the idea, my “Professor Xavier’s school for the gifted” sort of thing. So that might be what’s next. 

Will Rasmussen is a Director specializing in cybersecurity and cross-border issues.
Edward Stephens is Deputy Editor of the Brunswick Review. Both are based in New York.

Photograph portrait: Tasia Keetman

Photograph: Courtesy of James Ahmed/Afiniti

Graphic: Peter Hoey