Continuing Silicon Valley’s long and storied history of misreading dystopian satires as instruction manuals, a startup has created a tech product that makes call center workers' voices sound white.
The startup, called Sanas and founded by three former Stanford students, was first reported on by Joshua Bote for SFGate on Monday. On Sanas' website, you can “hear the magic”: a simulated conversation between a call center worker with an Indian accent that can be modified with a slider that applies Sanas’s accent translation. After Sanas is applied, the voice sounds more like a text-to-voice reader than another human being, but it does sound typically American and white.
In a demonstration for Motherboard, company President Massih Sarimand and COO Sharath Keysheva Narayana called an employee in India who talked about his background and work. Then the Sanasa filter was applied. It removed the employee’s accent and created a passable white and American-sounding voice, albeit a bit robotic.
Sanas describes its approach as "accent matching," and advertises on its website that it can "improve understanding by 31% and customer satisfaction by 21%." Apparently, the software can offer multiple accents at the touch of a button—although its demo only features an Indian accent being turned into typically white and American—and the company frames its technology as "empowering" workers. According to materials offered by the company to SFGate, the company claims to have garnered about $132 million worth of funding thus far.
Ironically, given its focus on empowerment, Sanas' software to turn call center workers' voices into white American voices mirrors the plot of Boots Riley's 2018 dystopian satire Sorry to Bother You. In the film, the ability to put on a "white" voice on the phone allows the film's Black protagonist to rise up in the company, but introduces tension in the workplace that undercuts a union drive and eventually pits him against his former co-workers.
In an interview with Motherboard, Sarim and Narayana sketched out their business strategy and explained why call centers were their first choice as clients.
"Call centers have a very specific speech pattern. You don't have people laughing, crying, singing—those nuances of speech are not there so it is easier to build it. The next use case is on enterprise communications. As we started going live with enterprise call centers, they said 'Hey we have teams in Asia, we have teams in Africa,'" Naryana told Motherboard. "We want to leverage a tool like this so that we can give them a choice and we want everybody to be heard. We want to build a very inclusive work culture and we think this could be an extremely great product and technology to actually bring people closer."
Call centers are heavily surveilled workplaces—dominated by “employee monitoring” which some are eager to argue is somehow beneficial for the workers. These workers suffer the brunt of a customer’s anger when something goes wrong but lack the autonomy to go beyond a script or narrow guidelines laid out by management. Viewed as disposable, closely surveilled, experiencing little if any autonomy, and forced to deal with angry or racist customers, call center workers typically burn out in a few months—when they don’t, their mental health suffers. The core question here, then, is what deploying Sanas will actually do to working conditions for call center workers.
It’s not hard to imagine scenarios where the introduction of Sanas results in companies demanding more of their workers because they now have “accent matching" that is supposed to increase their performance with customers—a typical outcome when workplaces with minimal autonomy implement performance-boosting software. Narayana said increased performance would be a side-effect of Sanas' software, but it mainly has the potential to improve every aspect of this industry: call center labor conditions, the mental health of these workers, and the experience of customers on the phone.
"I don't care about metrics, I think about mental health, employee satisfaction, retention, and overall employee happiness. All of those metrics are checked out for me," Narayana told Motherboard. "I strongly believe once you keep your employees happy, all the business metrics will get better. So that's a secondary result of what we're trying to achieve. But the primary result for me is actually improving the lives of all these agents."
Sanas’s product doesn’t address the structural issues with call center work nor racism from callers, which its product implicitly side steps. Sanas acknowledges the potential for misuse of its software and says nobody will be "forced" to use it because workers themselves activate it with a button—however, this doesn't acknowledge the possibility of being forced to use it by default due to performance quotas. Sanas also says it has a "code of ethics" with three values: individual choice (it's activated by the worker), personal control (effectively the same point), and flexibility (Sanas offers multiple accents).
The desire to communicate clearly and seamlessly with one another is understandable—as Sarim and Narayana reiterated to Motherboard multiple times, and as the website says, 80 percent of this company’s workers were immigrants. Sarim and Narayana have both worked call center jobs where they dealt with racism. The two insist this informs their end goal: not to simply have an accent translation engine that turns anything into white, American English (“many-to-one”) but one day to develop a translator for any accent to any accent (“many-to-many”).
Too often, technology is deployed to address—and profit from—an issue far removed from the core problem. Is the real problem to be solved that call center workers are misunderstood because of their accents? Or is it that we are content with growing an industry rife with surveillance, racism, and intolerable working conditions?