That weird thing Jeff Bezos said 17 years ago
How conversational interfaces are narrowing the gap between “I want it” and “I got it”
17 years ago, Wired magazine ran a story about Jeff Bezos. It was 1999, the year Amazon expanded into selling toys and electronics (and other things besides just books). From all the insights Mr. Bezos shares, one caught my eye in particular. I thought it was weird. Weird enough that I still remember it 17 years later:
Asked, however, to name the one missing technology that, if it existed, would dramatically improve Amazon.com’s business prospects, he says simply, “Windows instant on” — meaning a personal computer that boots up as quickly as a TV or a PalmPilot. “At home it’s a real pain,” he says, “because in the 90 seconds or two minutes that it takes, I’ve forgotten what I was going to do!”
He was worried about the time passing between wanting something, and doing something about it. You see, back in the old days of 1999, when you ran out of milk it would take you two minutes to boot up your computer, browse to kozmo.com (Amazon didn’t have groceries until much later), and stock up on more milk. Jeff was right. By the time your desktop appears, there’s a very good chance that you’d forget why you switched your computer on, and just start checking your emails.
And today? Well, we all have our instant-on wonders: our laptops and tablets and smartphones, ready to light up in a second. But you’d still need to browse to Amazon.com (or open the Amazon app) find the milk and order it. Unless, well, unless you have an Echo. If you own one of these bundles of joy, you’ll just speak into the air:
Alexa, reorder milk
Wow. That was, what? 3 seconds? That’s 2.5% of the time it used to take! Well played, Amazon. There is no way I’m going to forget I needed more milk (or laundry detergent, or a new iPhone).
Why does saying it works so much better than browsing for it?
Achieving the shortest path between intent and action
So I want my fridge full. Oh, and I want a taxi later. I want to know how busy I am tomorrow. I want to know when I should leave to be on time for my downtown dinner party. Is it going to rain? What wine should I get? Okay, I want that wine too. Have it delivered with the milk today.
All these good intentions (I just learned that Google calls them micro-moments), need to be translated into actions or be lost forever (or until I open the fridge again). With so many things competing for our attention, what’s the quickest, with least effort, way to handle this translation?
The quick answer: we need an assistant.
Longer answer: well, it actually breaks down into three capabilities.
Unified context. A conversation
Groceries, taxis, weather. There is an app for that, right? Actually, these are three different apps to switch between! It’s too slow and too long.
Consider instead, a conversation:
Me: What time will I arrive at the hotel on Thursday?
Assistant: Your flight to London lands at 4pm London time. With average traffic, you can expect to arrive at around 5:30pm
Me: What was the name of the nice Greek place we went to last time?
Assistant: 10 Greek street, on Soho. Would you like to make reservations?
Me: I'm not sure. What's the weather going to be like?
Assistant: Cold and rainy, with a chance of Thunderstorms
Me: In that case, find a place that serves soup, walking distance from the hotel.
What just happened? In one go I engaged my flight, weather, traffic and restaurant info. It works, because the conversation unifies all these services under one single interface: words. Moreover, my assistant established that we’re talking about Me, in London, in Thursday, and this information was carried over to all my queries.
Here lies the power of a conversation as an interface: a context is easily defined and kept. I can then proceed to seamlessly engage multiple services without mentioning the context again or having to switch to a different app.
You heard Mr. Bezos before. I can’t be bothered to wait for the thing to turn itself on. That’s why the Echo is always on. Always listening. It’s plugged into a wall outlet, so it doesn’t need to be recharged or turned off. And it has enough microphones to be present everywhere in the house (yes, there too).
That said, communicating by voice has a few disadvantages. First, while it’s okay for me to yell out commands when I’m at home, it can be a little awkward in the office, or on the train. And more important, talking to someone (or something) is a huge cognitive load. You need to pay attention and listen to the answers, and you can’t be interrupted mid sentence.
Enter instant messaging. While not “Always on”, it’s the next best thing: it’s in our pocket and on the desktop, usually already open and running throughout the day. We use it to talk to everyone anyway, so incorporating an assistant into the conversation comes naturally.
Better than voice, it’s very good for browsing the conversation’s history, and for multitasking. The conversation can pause and restart at any time.
A Good assistant knows you very well. He knows where you are right now, and where you plan to be later. If it’s morning and you’re asking “how is traffic”, you’re actually asking to compare today’s projected commute time to the average time it would usually take. If I want a table at “that Greek place”, he should scan my history for a Greek restaurant around my location.
He should remember some of my details, so that he won’t bother me with asking for them every time. Things like credit card number and my apartment floor and number. These don’t have to be present on first use. If it’s the first time I’m ordering Pizza, ask for my credit card and home address. Next time, just confirm:
I will pay for this with your credit card ****3814. Delivery address is 122 Allen street apt 14.
Say “Go” to order, or “Edit” to make changes.
But not just that. If I have loyalty points with Delta, remember that on my next flight search. Don’t book any meetings for me on Wednesday afternoons, that’s my time with the kids.
Next, predict the future
Here we are with the perfect assistant. Always there, in touch with every service we can think of, knowing everything about us. Takes us 3 seconds to get a tailor-fit solution for all our needs and wants.
But why settle for 3 seconds? What if my assistant knows what I want before I say it? Before I even realize this is what I want? This is not far-fetched. Armed with all the capabilities we discussed, an assistant is perfectly capable of proactively suggesting ideas:
I found one of your contacts, Dan Taylor, is going to the convention too. He was just promoted to the position of Purchase Manager at Acme Co. Should I try to reach out and arrange a meeting?
In fact, at Meekan we have found that our assistant’s users are 5 times more likely to engage if approached proactively (as opposed to waiting for them to initiate a request).
Meekan’s assistant robot is alerting you about an upcoming problem in your schedule, and immediately offers 2 possible solutions right there.
- Bring everything into one, highly-available platform (Slack, Echo, Facebook messenger).
- Use a unified interface (words) to operate everything, and smoothly switch between different services.
- Collect knowledge as you go, so you don’t have to ask for it again next time.
This is a part of Meekan’s journey to create the digital office manager. Read more stories here