Since a high percentage of services will likely end in a transaction, there's a huge opportunity to become the Amazon-scale retailer of services. I find I don't even google products I want to buy anymore, I just immediately search Amazon. Amazon is my retail UI. If Magic can do the same thing in this space—become the service UI—I'd imagine it won't be long before more revenue comes form partnerships than directly from consumers.
If you know what you're looking for, it's fine, but as soon as you try to comparison shop it falls apart fairly quickly.
I've gotten into the habit of looking for experts on other sites that do hands-on comparisons for the thing I want to buy (some recent examples: rain jackets and fitness meters). Amazon is honestly a terrible UI for this sort of thing.
I know that they're trying, but at this point they've lost my trust that what they show in search results is relevant unless I'm being extremely precise.
Haven't people been dismissing them over their UI since the beginning of time? And yet it works and is successful?
As an added bonus, I don't have to relearn how to use it every 6 months when the design team gets bored and decides to redo the whole thing for no good reason.
There are a few additions I'd like to see (PadMapper style apartment search being the big one), but all things considered I like Craigslist a lot more than most websites.
Compare it to Amazon where half of the items are miscategorized, sort by price doesn't actually sort by price, and sort by average customer rating is way too literal. In what world do I want to see something with a single 5 star review over the one with 191 reviews, 94% of which are 5 star, and another 4% are 4 star? But that's tucked away on page 7 of the results after all the items with nothing but a single good review. Amazon's got issues to work out.
Need an on demand helicopter in NYC? Great, Magic knows just who to call for that because they already did it for Justin Kan. Need to rent the Exploratorium? Great, they can do that too.
For more common requests that fit neatly into other on-demand services offering, they could ostensibly partner directly with those folks and integrate with them. That way, Magic could have their own master dashboard to purchase services with minimal effort.
As someone who works in that space, helping people do similar things with greater accurately and quickness delivers a surprising amount of value... when you're talking about deploying that across 250+ desks.
The target isn't to throw technology at 99% of all requests, but to deploy it on the smaller portion of the process it's suited to. The only thing faster than 100% machine is 99% machine + 1% human.
Excuse me if this is a bit naive, but I've always had the impression the whole point of Magic is to build the AI side. Maybe not communicating back to the user... perhaps that part's always done by a human, but a human with machine-aided decision making. I would love to see the internal interface that Magic's personal assistants are using today.
Doing every task by hand with people probably doesn't scale. Maybe for $100/hr it does, but that could just be their way of targeting tasks for executives because they have a higher percentage of profit. I expect they're using the data from requests today as training to recognize narrow patterns to be automated tomorrow.
Edit: I found this in their FAQ.
> Q: Do you use AI to fulfill my requests?
> Magic+ is fulfilled through a combination of top-tier executive assistants and concierges, paired with intelligent software. Our unique combination of humans and software is what enables us to consistently deliver an unparalleled level of service.
They don't need to compete with Google or Apple on the front of AI as a polished product. Not now, anyways. All they need to do is make something that works. They can get away with building something that wouldn't even be beta-worthy for big tech companies to release. To Magic, AI could simply mean more efficient infrastructure, keeping human interaction in place of UX.
The more efficient they get, the more they can lower the price. The more they can lower the price, the more users they'll have. The more users, the more revenue, the more cash to spend on improvement.
AI doesn't have to roll out of MIT looking like Jarvis just to be useful.
I'm not the OP, but...
I guess, but again, where's the software? You can entirely reasonably solve that problem with a well-indexed medium-sized binder and a Rolodex.
We may not yet have the technology to create a Jarvis, but we can make a person a lot more efficient at their job with some clever tools.
For example, in the website there's a request that mentions getting a private helicopter booked to LAX in 3 hours. A personal assistant would have to look up private aviation companies that fly out of a base nearby the GPS location of the user, guesstimate the closest one and most likely call to book as the online bookings might not be automatically processed and it's a short notice.
With today's technology, you can parse that request so that the software automatically picks up "Helicopter booking", from: GPS to: LAX. If it's right, the operator accepts it and it creates a checklist: 1) Call "Bob's helicopter co." 2) Enter exact time of booking: [ ] 3) Click to book car for booking_time - travel_time at [current_location|office|home] 4) send nice reply to customer.
Once you've done this a couple times, you write new software that automates the call step by filling in a form online, or sending an email to the receptionist, or both.
More importantly, since you log every action the operators take, you take decisions on what to automate based on data. Are our employees spending 24% of their time on google maps? let's add a database of businesses so they can simply choose from a list.
Note: I just drove from San Jose to Raleigh with two young kids, including a day from Salina, KS to St Louis. We found the most amazing place to stop, and I highly recommend it to anyone driving E-W or W-E on I70. The Moon Marble Company (https://www.moonmarble.com/) is not just a handmade marble shop. It has the best collection of retro toys I've seen anywhere in my life. We spent 3 hours browsing!
Have you ever worked with a really great Grand Budapest Hotel style concierge? They really can do pretty interesting things, if you tip well and are willing to pay. But, you gotta go to that hotel and talk to that one guy. The software scales that person.
Google, Facebook, many other companies (perhaps not Apple) publish many of their research findings and a lot of innovation is coming straight from academia.
To answer your question more directly, the hardware, software and data required to run an effective machine learning platform is getting cheaper and more powerful every year, which should drive down price if they can replace human tasks with AI solutions.
As evil as it seems, that's probably a sane strategy - as actual automation eats the job market, the number of unemployed, or those seeking multiple/alternate employment, will rise. These will more and more comprise displaced white collar workers. This will make the employment market a buyers market (moreso than it is already) and will result in decreased labour costs.
So - call me cynical, but if that's their plan, it's robust - if potentially evil - at least it'll be providing employment.
Look, it may not be Tesla, but there are a couple of unicorns that seem tech savvy now, that weren't so much for the first little while. AirBnB comes to mind - their analytics now are probably amazing, but they started as a website with a database. Doesn't make them any less of a technology company.
Leveraging technology to break down barriers (immediate, constant, on-demand availability and scalability of a personal assistant) is a technology solution to a problem.
Call me skeptical, but where exactly is the tech involved? Currently this is just text messages going to people + people doing the things.
One possibility would be through AI (a bit of a weasel word most of the time, honestly), but Google and Apple are both miles ahead on NLP, so by the time you can parse "get me a peperonni pizza", Google will have been doing it for free for the past year , and include a 10% off coupon for the pizza!
Considering your value is based almost entirely on getting things done through an assistant that understands English, it looks like costs are basically how much you pay the assistant themselves... Maybe you can get away with paying them $20/hour (and equipping them with good tools to do the common stuff fast).
The Tesla strategy works (well, in theory) because so much of the cost of the good is based in things that are continuously getting cheaper (batteries).
But here... What's driving down the cost of the assistants? Seems like the solutions are much more about HR-style "innovations" , business partnerships, and the like than anything coming out of MIT.
I sincerely hope the strategy isn't "find the cheapest human labor possible"