How sad Anastasia beat everyone
We ran a test comparing 7 voice synthesis variants for a support bot. From upbeat-and-cheerful all the way to flat-and-exhausted. The gloomiest one won. The team named it "Sad Anastasia."
Why? Pretty obvious once you think about it: when a bot answers in a bright, perky voice, the user immediately knows it's a robot. Because real call center agents don't sound like that. Real call center agents sound tired.
This fits neatly into a narrative I've heard my whole life: "Russia is just a sad place, you wouldn't get it." Convenient, a little insulting, and I don't buy it.
Let's look at it differently.
Sticks, scripts, and penalties for going off-template reliably produce defensive behavior and poor mental health in the people running them. The people arriving aren't broken. The system breaks them. And this isn't a Russia problem. US call center churn rates run around 35%, roughly double the average for other operational roles. In some states it exceeds 100% annually.
"Sad Anastasia" isn't a cultural insight. It's a product design problem.
Zappos built something completely different.
The online shoe retailer inverted every standard call center practice from day one:
After onboarding, before their first shift, every new hire gets a choice: take $4,000 and leave, or start working. No pressure — if you feel this isn't for you, just take the money. About 3% do.
No scripts. No KPIs on call duration. No call monitoring or quality scoring. The entire policy is one sentence: solve the customer's problem, in whatever way you think is right.
They also don't call their reps "agents." They're customer loyalty managers.
The results: 11-20% annual turnover versus 30-45% industry average. 75% of revenue from repeat customers. Revenue growth from $8M to $1B in 8 years. Amazon bought them in 2009 for $1.2B and copied the motivation model for their fulfillment centers.
At T-Bank we're genuinely obsessed with proving that great service drives customer LTV and retention. We're actively running experiments with positive motivation systems for support operators.
In a couple of years, the same voice synthesis experiment will give a completely different result.
Stay tuned
🥷🥷🥷