During the process of rapidly growing TaskUs to over 100 employees, Jed jokingly said “I’m starting to feel like a cyborg!” I passed it off as a funny, over dramatic joke. But lately, I have been hearing that word a lot in articles covering the Service-on-Demand market. Technology has advanced to such an extent that many things previously preformed by people in high paid professions are now being preformed by computers. But some of the most advanced processes still lack 100% accuracy when done by machine so a human must intervene at various stages in the process. The current Service-on-Demand is neither a labor market, nor a technology market. It exists at the intersection of the two. Simply put its a Cyborg market.
The three biggest Cyborg services are translation, transcription and sentiment scoring. Each of these services can be completed by sophisticated algorithms, algorithims that are improving daily, but are far from 100% accurate. On one side of the spectrum rest Pure Technology Solutions. Think Google Translate, Google Voice or TweetFeel. Each of these solutions is 100% automated and therefore (a lot) less than 100% correct. Have you ever read a webpage translated by Google, or seen a voicemail transcribed by their automated solution. Check out these examples:
One would think that 140 character tweets would be easy for a computer to score as positive or negative, but a close look shows that context is essential. If I say, “Inglorious Basterds was sick!” I mean it was a fantastic film (POSITIVE). If my mum says the same thing she means it was gratuitously violent and difficult to watch (NEGATIVE).
At the other end of the market are Pure Service Solutions. These consist of online translation services that charge $0.10 to $0.25 per word, depending on the amount of checking and editing required. Firms like Trusted Translations and Applied Lanaguages are good examples. In the transcriptions industry firms like VerbaLink and GMR offer 100% human transcripts for $1.50 to $4.00 per audio minute depending on the number of speakers and clarity of the audio. Finally, high-end sentiment solutions are availible from Neilsen BuzzMetrics, Visible Technologies and Wird Sentiment all of which charge $5,000 to $10,000 per month for comprehensive sentiment scoring for a particular brand.
In between is the avant-garde of Service-on-Demand – Cyborg Solutions. At their best these are symbiotic solutions in which technology improves the efficiency of labor, and labor improves the accuracy of technology. My friend Jamie‘s company Phonetag is a great example. Their voice-to-text algorithm is more accurate than Google’s! Yet, unlike Google, they charge their customers for voicemail transcription so they have to have humans go in and clean up the results assuring (almost) perfect accuracy. Humans have also been used to improve the results of automated webpage translation. There are paid services like Smartling. But more interestingly their are armies of translator volunteers helping to translate huge sites like the Economist and Facebook. Again these people piggy back off of algorithmic translation to improve efficiencies. Finally most reputable auto-sentiment solutions like TweetFeel Pro and Radian6 offer the ability for humans to rescore results. A few of these companies are also thinking of offering a hybrid model in which automated sentiment scores are cleaned up by an army of humans.
For now those of you wanting to launch your own cyborg’s can turn to the solutions like Amazon’s Mechanical Turk – crowdsourcing in which freelancers around the world offer to rescore tweets, provide translation, even clean up your Google Voice messages all for as low as one penny per activity. Mechanical Turk is an amazing idea. Who wouldn’t want to crowdsource their tedious data entry for rates that amount to $2.50 per hour? But Mechanical Turk has yet to really take off because of two major flaws – reliability and security.
The former issue has been addressed by startups like CrowdFlower and CloudCrowd. These programs tap into Mechanical Turk’s labor pool and use another algorithm to predict the accuracy of individual workers. Customers can then submit tasks and request levels of accuracy – the more accurate the more expensive. Rates range from $3.00 to $12.00 per hour.
I’ve been very impressed by all of these innovations in the Cyborg Market but I don’t think anyone has got it yet. Service-on-Demand is a new industry that has caught the eye of individuals because it sounds sexy – Everyone can have their own army of assistants! But the real consumers of Service-on-Demand are not individuals, they are business. Dollar for dollar businesses will outspend individuals by 100:1 in this industry over the next decade. And if there is anything businesses demand as much as service its reliability and security. Brand matters in this area and you can’t build a brand off the reputation of 100,000 freelancers who disappear frequently, work while they are watching TV (if in the US) or fighting off a Typhoon to hold on to internet and power connection (if in the Philippines).
Security is even more important for many corporate clients. Can you imagine any Fortune 500 company ever asking 100,000 people to score the sentiment around their brand?! Imagine all the negative reinforcement that they would lay themselves open to. Or how about a lawyer or doctor (the two largest consumers of transcription services) asking people chosen at random to transcribe their notes and depositions? Sure enough someone would end up transcribing the results of their neighbors mammogram.
So I think that there is room – a wide open gap, in fact – for a reputable, secure Cyborg Corporation, set up to utilize the ever changing advances in technology to deliver reliable results, while securing the entire process from the human variable.
This is the company that I want to build.