Leopards Eating Faces

5 mins

I never thought leopards would eat MY face, sobbed the woman who voted for the Leopards Eating Faces Party.

There are variations on this meme that take the form of cartoons, for example Slugs for Salt or the one with two members of a flock of sheep, who pass a billboard showing a picture of a wolf with the tagline I am going to eat you, with one sheep remarking He tells it like it is.

This is not a political post. It’s about the survival of your startup.

Yesterday, Amazon announced yet another new service - Amazon Transcribe Medical. This is an extension of the existing Amazon Transcribe service which, in Amazon’s words, is:

...a new speech recognition capability of Amazon Transcribe, designed to convert clinician and patient speech to text. Amazon Transcribe Medical makes it easy for developers to integrate medical transcription into applications that help physicians do clinical documentation efficiently. It can automatically and accurately transcribe physicians’ dictations, as well as their conversations with patients, into text. Moreover, the service enables automatic punctuation and capitalization, allowing physicians to speak naturally when transcribing voice notes. At launch, Amazon Transcribe Medical is HIPAA eligible, and offers an easy-to-use API that can integrate with voice-enabled applications and any device with a microphone. Output transcripts will support word-level time stamps and confidence scores. Users can call the API to open a secure connection over WebSocket protocol and start passing a stream of audio to the service. In return, users receive a stream of text in real time. The raw text can be pumped into downstream text analytics services, Amazon Comprehend Medical, to extract valuable medical insights.

The graphic below is one I created after researching this over the past summer. The data comes from perhaps twenty different sources and doesn’t (afaict) exist in a nice aggregate image like this anywhere else.

It’s a huge, multi-billion-dollar market. At first blush, this sounds great for any startup that is digging into the medical transcription market.

But is it really?

A startup needs to capture & retain market. Capturing market share requires a method or mechanism by which you incentivize a change in behavior for customers currently in a market or - and this is where breakouts can occur - you incentivize completely new behaviors.


Medical transcriptions - as a market category - isn’t new. There’s an existing customer set with observable behavior. To grow revenue in this market, a company has to do one or more of these things:

  • Obtain new entrants, i.e. entities that aren't currently participating in the market
  • Get existing customers to spend more, on existing & new services - or -
  • Get existing customers of competitors to switch to your offering

Let me expand on this.

With regard to transcriptions generally - and medical transcriptons specifically - Amazon up til this announcement was (at least publicly) only providing the transcription engine. This is a hard problem to solve (I developed my own over the summer) so having a well-performing, highly-available, accurate transcription service available via an API leveled the field for a lot of startups to innovate on services that did something with those transcriptions. Tincup Voice is such a service.

But now Amazon is walking up the value chain that’s enabled by ubiquity of a transcription to directly innovate on top of their own service. This is a virtual torpedo to any - read: every - startup that is or was developing a medical transcription service on top of AWS Transcribe.

There’s another level to this that makes it particularly insidious: at least one major health services & data company, Cerner, has apparently been working with Amazon on this since this past summer (if I’m reading the press release correctly). Realistically, they were disclosed on this new service a long time ago & got a huge head start on it.

Are you using AWS Transcribe in your startup’s offering? Were you predisclosed on this? Do you have a new service that leverages this new AWS functionality on day one of its GA?

Yeah, me neither.

In a single announcement, Amazon has essentially guaranteed that a not-insignificant portion of the medical transcription market’s lucrative revenue will be (has been) captured by… Amazon or its big partners.

If you’re a startup like Robin, with an Alexa-style smart microphone device for medical offices, a whole lot of your market differentiation just got commodotized. I don’t have any special knowledge of Robin, but imagine how awkward it would be for their team if they were using AWS infrastructure to power their service. Ouch.

This is just one example but it makes my point for me brilliantly.

Back when I was doing legacy server migrations, I developed a point-and-click utility that would clone an existing - and still running! - on-premise server and deploy it into EC2. Then AWS announced their migration services, commoditizing my secret sauce.

If a customer were to use my services instead of AWS’, I’d be forced to compete on cost. It’s pretty quickly apparent that you can’t do that. I analyzed this exact issue in my Build Systems as a Service post here.

Which brings me to my point:

If you’re a startup that’s delivering a service & you’re using AWS infrastructure, what’s your Plan B for when Amazon decides to eat your face?

The race to zero is over fast & the only winner is Amazon.