Amazon is everywhere.
As of 2021 it employed 1.6 million people and generated $469 billion in revenue.
It had 50 Boeing 767s and over 100,000 automatons.
And yet Amazon is nowhere.
Outside of advertising hoardings, Amazon’s appearance in our physical world is limited. Often, it’s impact is only noticed on our doorsteps, when a small cardboard box appears.
How is it that Amazon can be so everywhere, yet be so nowhere?
Amazon has come to occupy and gain power in the spaces between places. Amazon Web Services (AWS) was launched in in 2002 and in its 20-year life span has achieved some incredible things. By 2022 AWS was responsible for a 30% market share of all cloud computing. Its customers include NASA, Netflix, the CIA, MI5 and MI6.
So, Amazon is on your doorstep, but it’s also with you whenever you look to the nights sky.
With over 40% of ecommerce transactions made via Amazon it means not only that there’s a good chance your device came from there, but also that the information flowing through it can be traced back to Amazon.
Look beyond your doorstep, towards the city limits, and you’ll find another Amazon box. This is a much, much bigger box, from which all the small boxes fly. Inside lies a strange and opaque ecology – of which you are a part. This artificial ecosystem has emerged in the space where AWS collides with real things and real people.
Welcome to the Amazon.
—
AWS knows a lot about you. It has your purchase history but AWS collects and collates a lot more than that: it knows how you filter search results, how many items you’ll browse before adding to your basket and how long you’ll hover over that ‘buy’ button before you click.
AWS doesn’t just know this about you – it knows this about everyone. And from everyone it can carve out groups of people who are just like you. Once you’ve been grouped by your historic data, AWS can start predicting your future. This is because AWS knows what the people just like you did right after they searched that term or purchased that item.
“If you like that,” says Amazon, “you’ll love this!”
Every day AWS makes forecasts for over 400 million products. It knows where they’ll be purchased and when. This allows Amazon’s logistics chain to begin before an item is bought and is key to how Amazon ensures deliveries can be made next day.
Forecasted items arrive by freight to one of Amazon’s Fulfilment Centres. Here they are manually unloaded before being provided with a unique ASIN (Amazon Storage Identification Number). Conveyor belts then route items to an employee’s work station.
When an item arrives, the worker scans the ASIN. This starts a cascade of processes:
Firstly, using ‘Computer Vision’ AWS assess the aesthetic qualities of the item, matching to details stored in the database. This check ensures that the ASIN is matched to the correct item as well as helping AWS to decide where an item should be stored.
When AWS decides upon a storage location a robotic assistance is dispatched to a shelving unit. It lifts the unit and wizzes to the worker’s station.
AWS then highlights an appropriate storage draw for the worker to open.
AWS has selected this draw based on numerous factors including: the fullness of a draw, how the new item will impact the current weight distribution of the unit, as well as the appearance of the item.
When the worker opens the draw they should find nothing with a similar appearance inside.
To us humans this may seem counter intuitive: when we store and organise we do so on theme or quality – alphabetically, by similar function, by relations to other items.
But AWS does not permit storage on human norms. AWS has embraced “Chaotic Storage,” all in the name efficiency.
“Chaotic storage” is as it sounds: everything everywhere. No human could understand the system nor keep pace with the ceaseless incoming and outgoing of items. But AWS, pairing ASINs with specific shelving units and draws, can. The emerging intelligence has come to realise that with an integration of human and machine “Chaotic Storage” is the most reliable and efficient method. It minimises mismatching items and increases the likelihood of multiple required items being located within a single shelving unit.
So the worker drops the item inside. The draw is closed. The shelfing unit is lifted and returned.
The automaton waits.
Then an order is received.
Zipping across the floor on its predetermined path the automaton scoots under another unit and heaves upwards. It zigs and zags and slides and stops at another workstation.
“Computer Vision” is there again, this time highlighting a draw in white. The human opens the draw, picks up the item and it is “seen” by AWS. The ASIN is scanned and, assuming it matches the item in the database and that the visual qualities conform to the expectation of AWS, the item can now be packed.
AWS dictates all aspects of the packing process. The human is informed of which box to select, the amount of protective packaging and tape is preordained and auto-dispensed.
AWS also knows the rate at which humans must work in order to meet demand: the minimum number of items to pack in an hour is typically 65.
Once packed, the item is dropped onto a conveyor belt. Amazons’ larger fulfilment centres contain over 17 miles of conveyors and these too are controlled by AWS. AWS routes the packages, but also schedules maintenance work to avoid any disruption to the logistics process.
During the conveyor ride AWS affixes an address label and its emergent intelligence determines which delivery carrier can meet the shipping deadline with the lowest possible cost. AWS routes the item back to the dock and it is loaded once more.
And all of that happens so that the next day you can look at your doorstep and Amazon can be there for you, again.
