Why the future of work is work from home

Rev is a company that transcribes audio files, captions and subtitles video, and translates documents. None of these are in and of themselves revolutionary business ideas, so you can forgive Founder and CEO Jason Chicola if he gushes over the company rolling out a new automated product, Temi. From the outside, it seems like the future is arriving.

“The typical reaction we get from Rev customers trying Temi is that the quality is too low for their use case. There’s a minority of people that try it and say, ‘This is perfect for me.’ For people who speak slowly in rooms with good acoustics, it is dynamite for them.”

And if you thought that was a soft sell, customers received an email announcement that bolded the words “the quality is much lower than Rev.” For Jason, Temi is not a game changer. It is not going to alter the way we have conversations. It is not artificial intelligence. It’s just a considerably worse version of the services the company already offers – albeit faster and at one-tenth of the price.

This was, initially, puzzling. We had reached out to Jason to discuss Temi, and he quickly made it clear that Rev wasn’t about automation at all. In fact, if anything, Rev’s business model is a huge bet against the popular mythos that robots are going to replace workers. To the extent that the company has innovated, it has simply connected human labor with jobs that need to be done.

“Rev’s goal is to create millions of work at home jobs. That’s what we exist for. Everything we do is in furtherance of that mission,” Jason tells us.

“I’m very pro-free market, and I specifically think that meaningful work is the best way for people to achieve happiness and purpose.” It makes sense, then, that he has devoted much of his career to building marketplaces to get freelance workers hired. First, as employee number three at oDesk (now Upwork), the massive freelance marketplace, and now with Rev.

“Both of those companies are about letting people work from home and giving economic opportunity to people. As a practical matter, this means letting people who typically didn’t go to an Ivy League school, or don’t live in New York or San Francisco, find work they really enjoy, and being able to make the most out of their skills and abilities.” In this article, Jason tells us about the philosophy behind Rev, the difference between selling labor and services, and how robots are not, in fact, coming to kill us all.

Rev started with a bet: the number of people working from home will continue to rise. They wanted to facilitate that. From there, it was a question of narrowing focus.

The most important thing to Jason was finding a niche where they would be able to get clean, objective data on whether or not they were getting results — call it the “beauty is in the eye of the beholder” problem.

“First, we looked for a service we could standardize. It couldn’t be something where judgment of quality is subjective, like logo design or software design. We wanted a clear answer as to whether we had delivered on what we’d promised to our customers.,” Jason tells us. With transcription services, either the words are written down correctly or they are not. Turnaround times leave even less room for interpretation.

Next, their service had to adapt to the marketplace.“For any business or marketplace, you need to have a lot of repeat usage. Repeat usage is essential,” Jason tells us. Obviously having people pay you multiple times is better than having them pay you once, but according to Jason, the ancillary benefits are arguably larger.

“Building a marketplace around a once in a lifetime need is not a great strategy because the customers are never going to be engaged enough to give you the feedback you really need to make your service delightful,” Jason says. “We have over 100,000 customers, and a high percentage of them use the service on a very regular basis. They love telling us all about their gripes and how to make Rev better, so we’re constantly improving our products. Without that feedback, we couldn’t survive.”

Rev’s founding strategy was a data strategy. With quantifiable goals and a business based on repeat customers, Rev could know whether or not it was hitting the mark, and if not, pivot to where it could succeed. As it happened, despite translation being their initial offering, Rev quickly learned that their transcription business was more robust, and adjusted to that reality.

Hiring freelancers online can be a mixed bag – one that brings to mind Craigslist grifters and Nathan for You-caliber work. Jason wanted Rev’s users’ experience to be … not that. He wanted it to be as frictionless as possible. “If we wanted remote work to become mainstream, the platform should do the hiring and managing, and not require the customer to have that skillset,” Jason tells us. “We only provide services where we can vouch for the quality of the output. We don’t sell labor; we sell services.”

Thus, the pricing is clear. All a user has to do is upload a file. A few hours later, having received nothing other than a notification email shortly after uploading the file, voila, a transcript. Here’s what’s missing: picking who does the job, negotiation, and literally anything else that happens between Rev receiving your file and you receiving a transcript. All that happens on Rev’s end with zero input from the customer, who is unaware of who did their transcription until they are given the chance to assess them upon receiving the transcript.

But in order to make a silent labor market function well, Rev has to be cognizant of the requirements of both parties: the freelance worker and the customer.

“I think of Rev as a managed marketplace,” Jason says. “A freelance worker on a typical labor marketplace spends a huge amount of time applying for work. A couple of problems occur with that. There’s often a race to the bottom with pricing. It’s very difficult for the customer to tell the difference between high quality and low quality. This means that the workforce spends a lot of the time not just selling itself but also dealing with customers who don’t know what they want.  The freelancer can end up in a situation where if somebody doesn’t know how to ask for what they need, the freelancer might get frustrated with the whole project.”

“When somebody works on Rev, they spend zero time selling themselves. They can work whenever they want. Every job they take, they’re getting paid for it. They don’t negotiate the rates. They take the jobs or they don’t. This results in higher productivity because they only have to work, versus having to market themselves, but it also results in much higher happiness. We fundamentally believe that if you’re really good at what you do, you shouldn’t have to spend time selling yourself.”

Despite using it in his products, Jason Chicola is not seeing the awesome, but job-destroying powers waiting to be unleashed by artificial intelligence: “When Elon Musk talks about how Skynet’s going to kill us, I think he’s trying to sell cars because it’s free media coverage. Now, maybe some people out there have this incredible artificial intelligence. I hope they do. I’m confident it can be used for good instead of evil, but what I see is pattern matching that doesn’t work very reliably.”

This may seem like a somewhat gloomy assessment. After all, The Signal initially reached out to Jason to discuss a new product, Temi,  another transcription service that costs one-tenth as much and delivers transcripts within minutes rather than hours. The catch? Temi is an automated software program, not a human doing the transcribing. And as Jason will happily tell you, that means Temi transcriptions are simply not as good as Rev ones.

Jason does not see the technology in Temi as a replacement of Rev, but rather a supplement. People will record and transcribe things they previously wouldn’t have, and accept that the transcript is less precise. So, why build it then?

According to Jason, even if there isn’t an immediate, obvious way for this technology to make workers more effective, down the line it will be. In this case, it’s easy to imagine a worker who can amplify their output with a product like Temi commanding a larger paycheck. That, broadly, is the future Jason sees for Temi, more so than the machine totally replacing the worker.

“This is counterintuitive. I would say 9 out of 10 people that I might meet in passing in San Francisco would say, ‘Oh, you’re a technology company doing transcription? Well, clearly you must automate everything or automate most things.’ If we thought that’s what customers wanted, we would have done it, but we haven’t because the technology has generally been not good enough and is unlikely to improve to that point anytime soon.”

And as to whether the technology to replace humans in this function is around the corner — again, in speech recognition, which an area that is fairly advanced as far as machine learning goes — Jason is deeply skeptical. “I believe that there’s a massive misconception in the public, both the general public as well as the technology community, about the state of artificial intelligence and how effective some of these things are. We’re perpetually 18 months away from self-driving cars.

“The people that I’ve met that are most optimistic about speech recognition have been undergraduates in computer science programs. The people that I’ve met that are the least optimistic about the technology are the ones that work on it the most.”

The hype is overhyped. “Most of what’s being called artificial intelligence would probably be more accurately termed machine learning, where the machine learns from patterns,” Jason says. “If I say a word like ‘coffee’ 100 times, and you say ‘coffee’ 100 times, and someone else says ‘coffee’ 100 times, and someone else says a word that might or might not be ‘coffee,’ a machine can actually be pretty good at knowing, did he or did he not say ‘coffee?’ If I’m talking about the history of the Civil War, and I’m talking about who won a battle and what happened next, that computer listening to it is not going to compare this to a database of how the Civil War played out, and be able to know that after Battle X, Battle Y happened. It’s not intelligent in the way that a 12-year-old who has some context is.” Jason doesn’t see this contextual failure as a problem machines are going to be able to overcome quickly.

Rev, after all, hires the majority of its transcribers from North America because that’s where their customers are. That way, they have the cultural context gained from going through the same educational system, spelling the word “color” correctly, watching the same TV shows, and everything else that comes with being North American, and that shared context allows the transcribers to fill in gaps when they hear things in a recording. If British people have trouble adapting to the American context, computers certainly will, too. “That filling in of the gaps, it’s a very different way of thinking than, ‘What does that sound mean?’” Jason says.

So count Jason Chicola as an AI skeptic for now.

“As the founder of a company, what’s most important is the velocity of ideas that we’re trying, and when we see the data, that we’re honest,” Jason tells us. It’s a variation on a familiar theme: be lean. Move fast, break things. But it’s a philosophy at Rev. “The key to success is not being right every time, but it’s making a lot of thoughtful bets and being objective about figuring out which ones work, and which ones don’t and not caring about winning the bet.”

And for a company like Rev, a key aspect of being fast is keeping the experiments simple, because, in simplicity, there is clarity.

“As a company, right now we have eight product teams. At any one time, each team has a single metric that they’re trying to move over a three-to-six-month period. The metric is almost always tied to customer or worker happiness. It’s something that allows us to say to customers, ‘We’re trying to get you your transcripts in half the time.’ Keeping it simple works great. When we have something that really succeeds, we can ask why it worked, and if it doesn’t succeed, we can investigate why not. We’re always focused on that process, so we don’t obsess over the failures,” Jason says.

Rev is a company built on a simple idea: more people are going to work from home, and they’re going to facilitate and accelerate that process. Right now that means transcribing, captioning, subtitling, and everything else, but in the future, who knows? “We aspire to be like the Amazon of from-home work, and we think of transcription like they do of books: the first business that they nailed, and the formula for everything came after.” Whether the company can meet these grand ambitions is impossible to say. But in the end, when the story of Rev is told, they’ll be there to transcribe it.