Technology Morphology
Domo Arigato, Mr. Roboto
S. G. Lacey
​
Mexico City, Mexico: AMZN
Boxes of seemingly infinite variety whir past: big and small, brown and white, sealed and open. These parcels move on all manner of conveyor mechanisms which are almost as random and diverse as the contents inside. Rube Goldberg would be impressed.
As each package travels on its amusement park-esque roller coaster ride, a variety of analytics data is taken. Despite the seemingly haphazard and circuitous path each item takes, in reality this maze is an efficient sorting and filtering device. Like a giant pinball machine, each bumper, rail, and lever change the course of the object, all funneling towards a specific end goal. In this case fortunately, it’s not a ball return, but instead a purposeful location within the warehouse.
Throughout the journey, laser bar codes are scanned and monitored, adjusting the path of each package. Bar codes are vital throughout this facility, being used for product identification, warehouse location, and throughput optimization, and address verification, along with many other “ations”.
The site for this Amazon fulfillment center, the 4th and most advanced of its kind within Mexico, is perfect. Just minutes from the Queretaro International Airport, a cargo transportation hub, and only an hour by truck from downtown Mexico City, the 50-acre plot offers an ideal balance of cheap land cost and high population density accessibly. Recent USMCA legislation provides additional motivation to store products within Mexico, rather than at one of the nearly 200 warehouses Amazon currently operates within the United States borders.
Online sales are growing rapidly here in Mexico, already surpassing $250 billion pesos annually, which represents over 5% of total retail sales revenue in the country. Resultingly, we’re constantly inundated with both incoming and outgoing products to sort through.
The scale of this fulfillment center is truly massive. The warehouse building itself is expansive in the most literal sense the word: 1 million square feet which is the equivalent of 12 soccer pitches at the famous Estadio Azteca just down the road. In fact, that entire stadium, which holds over a 100,000 people, along with all the associated parking lots, could easily fit within our footprint.
Viewed from above, the warehouse floor is a complex maze. Linear pathways with sharp turns, and lots of dead ends. The floor plan could be straight out of a puzzle book.
When I started out here, I spend days simply exploring the vast space, running into no-exit points down long, narrow isles of product, and tripping up on metal shielding located along the ground at intersections to protect the lower stored items from damage. However, over time, I recorded and memorized the entire building layout. Every dark alley, every hidden turn, every product placement.
Complex algorithms documenting shipment patterns are constantly being utilized to reorganize the warehouse inventory by frequency of selection. Amazon houses over 2 million unique SKUs under this roof alone. While pretty much everything leaves the facility in tidy brown cardboard box with the signature smiley face arrow on it, inside each of these non-descript parcels in a menagerie of random products. Designer blue jeans, a nerf gun set, multi-packs of AAA batteries, and even heavy-duty garbage cans; there’s an electronic position for each product based on the unit codes, with associated GPS coordinates.
At this facility, we separate our operators into 2 key functions. “Stowers” are responsible for stocking shelves with incoming inventory, while “pickers” execute the order pulling and shipping to customers.
The key element of bulk item transportation here is the “pod”, a 3-foot square by 8-foot tall plastic shelf. Most of these racks, like pretty much all the product handling bins inside this Amazon warehouse, are bright yellow. They must have found a sale on dandelion paint and goldenrod injection pellets.
Robot usage has become ubiquitous at Amazon warehouses these days, with over a quarter million units deployed across our entire fleet. Granted, most of these mechanical beasts are simple operatives, only capable of basic pod lifting, or mundane box sorting duties. With our various robotic mobility operations, it’s important to keep consistency throughout the transportation process.
Up ahead, one such machine, a low, squatty item that resembles an oversized Roomba, moves towards its next target. Powered by electricity, these devices are free from the bulk, and pollutive gases, associated with an internal combustion engine. Machine models are selected for various responsibilities based on their carrying capacity, much like humans; smaller, nimbler units can move faster, and execute more rapid tasks, provided they remain unburdened by heavy lifting.
The flat robot scooting around in front of me has a sleek metal body with rounded corners, and a front panel of dark plastic behind which the proximity sensors used for navigation are housed. While aerodynamic, this vehicle’s primary function is not speed, but carrying capacity. Weighing in at only 300 pounds, they are capable of lifting and moving pods up to 5 times their weight; most insects would be proud of this ratio.
Sliding smoothly under the targeted pod, the air-powered pistons engage, raising the entire tall shelving unit just a few inches off the ground to free it up for movement. There are very specific protocols regarding how products are arranged on the pods, with heavier, unstable items explicitly kept off the top shelves. At the slow speed and short distance these units are traveling, there’s little risk of toppling over.
Working as a team, dozens of similar robots move pods back and forth in unison. The entire operation resembles a dynamic Tetris puzzle; efficient, linear movements focused on geometric optimization, with the goal being to shrink the pile of inventory through shipments faster than it grows via deliveries. These autonomous beauties are at the core of this facility’s inventory management process.
The minions seem have this simple pod movement well under control. Time to head over to the loading dock for our hourly deliveries. Arriving there 3 minutes later by means of an optimized route, I find the usual chaotic scene.
Along the seemingly endless external wall of the building are 100 consecutive semi-truck bays, nearly all of them occupied. My scheduled unloading task is at Bay #37. I cruise down the loading dock line, avoiding other heavy machinery, piles of discarded plastic wrap, and various boxes which have been dislodged from their stack during off-loading.
I safely reach my desired spot just as the back of the tractor trailer door is rolling up. Perfectly timed.
I eye the full palleted load in front of me, analyzing the quickest way to empty the container shipment. Standard 2-foot cubed carboard offerings. The scanned code reveals the contents. Kitty litter. Denser that most materials, good thing I checked.
Running the quick math, 2 boxes deep and wide, stacked 4 rows high, provides a rough estimated weight. We’re looking at 300 pounds per box, about 2.5 tons per pallet. Easy. With my machinery lifting capacity and hoist range, I can definitely move 2 stacked pallets at once.
Moving forward, I maneuver into position at the bottom of the load, deftly slotting the forklift tines into the gap between the metal truck bed and the wooden pallet slats. We’ll have this rig unloaded in no time.
As I empty the truck, pulling packaged blocks off individually, then stacking them two high into 16-foot tall towers on the loading dock for longer distance transit, my mind wanders.
These wooden pallets seem so archaic, with their crooked, warped wood, and rusty nails sticking out at odd angles. With no standardization in material, shape, or construction, other than that a lift must be able to fit its forks underneath. These packaging devices of yesteryear continue to hinder operations even at this advanced facility.
As the industry standard of tractor trailer transport globally, we still receive over half of our incoming inventory on these splintery timber rafts. Product delivered in this manner is immediately off-loaded and transferred to more universally compatible containers. Hence, the massive stack of wooden pallets at the far end of the loading dock. Time for a bonfire.
On my way back to central processing with my last load of kitty litter, I pass by one of our newest machinery investments related to packaging innovation which is already paying dividends from both a cost and throughput standpoint, two critical metrics here at Amazon.
At one end, various products to be shipped out are registered into position, typically with smaller items of an individual order stacked on top of a flat object like a folded dress shirt, or holiday calendar, which acts as the transfer platform down the conveyor belt. As the organized pile flows downstream, the key metrics, like weight, via inline scales, and dimensions, via laser sensors, are determined.
By the time the shipment’s contents have reached the heart of the operation, the correct box size has already been selected, and turned from a flat cardboard panel into a rectangle packaging vessel.
Sophisticated 3D imaging software compares the contents to the box’s volume, then deposits bubble wrap off a roll, shredded filler paper, or corn starch beads from a hopper, based on space modelling algorithms, plus the fragility of the products. The box is then taped up with precision alignment, using the now ubiquitous blue Amazon “Prime Day” tape.
Here as well, packages are tracked by barcode, information transferred from internal product item codes to the external box label, with key delivery details and invoice paperwork printed on stickers and applied in real time.
This self-sufficient apparatus can process up to 750 boxes per hour, over 5 times as many as human counterparts executing the same task. Not to mention the reduction in returns from damaged items, or those shipped to the wrong address. Granted, it is helpful to have a human overseer around in case the hot glue runs out, or a stack of products topples over and gets jammed during boxing.
Other than the food packaging industry, there’s few manufacturing operations that provide this level of automation, while also yielding a finished good which is headed directly to the end user.
One of these complex mechanized machines costs $1 million USD, but the investment is recouped in under 2 years via labor reduction and increased throughput. We’re already running 2 such lines full out, and the capital outlay request for a 3rd unit has been submitted to the finance department. Efficiency and cost savings are core tenants that drive all Amazon’s business decisions.
Speaking of efficiency, time to check my travel log. I’ve covered 24 miles already, just short of an official marathon. Not bad before lunch. And I’m not even sweating yet.
Often on my journeys through the expansive Amazon warehouse, I encounter random items on the ground. The item-du-jour is a golf driver is a glossy plastic sleeve, which I almost run over coming around a blind turn a little faster that recommended by safety protocols.
It’s an object that presents a handling challenge for three reasons. Slippery, check. Odd shaped, check. Unwieldy balance point, check. Also, I need to figure out where to put this goofy shaped piece of hardware so I can transport it back to the packaging area. There’s certainly more opportunity to improve our protocols here in Mexico.
One of the challenges that still needs to be tackled in our warehouse is working with these uniquely shaped products that are difficult for basic robotic handling. The real skill lies in having the dexterity to handle intricate items, that due to shape, texture, or balance are difficult to maneuver.
Imagine asking your 5-year old child to carry a 6-foot bronze lamp, with delicate stained-glass shade, around the playground. Or try picking up a cold, bulky, slippery glass jar of pickles, weighting over 3 pounds, using one hand, but without opposable thumbs. Solving this technology skill set at the fulfillment centers is very important for future throughput gains.
The Amazon technology team has been working for years to unravel this human dexterity challenge with robots. They’ve sponsored several global engineering contests specifically on this topic, and made investments in numerous start-up companies in the robotics space. Still, there’s lots of development work to be done before mechatronics engineering can achieve the complex range of motions a humanoid hand does almost instinctively.
The most interesting attempt I’ve seen to date is a massive, table-mounted, robotic arm. The “hands” of this monstrosity were suction cups, made from a tacky elastomer material like on an octopus tentacle or gecko foot. However, since this mechanical beast was stationary, it provided little benefit for product picking operations throughout the warehouse, which are the most time-consuming activity in the order fulfillment process. Not to mention, the film of slime that the suction cup paws left on any product it touched.
This innovation was quickly abandoned in favor of smaller, more functional prototypes. A few of the more useless components that weren’t cannibalized for other robotics projects are still stashed in the back recesses of the warehouse. Sometimes it’s fun to swing by that spot and reminisce.
Suddenly, a change in my onboard infrared temperature sensors pulls me from this idle musing. It must be regular warehouse hours, since the ambient temperature always kicks up a few degrees when the overhead lights get turned on. All night, aside from the faint red glow of barcode scanning, and the occasional picture flash as identification markings on certain packages are documented by camera, a mouse would have trouble navigating around the floor of this facility.
Too bad, I function at my best in a completely dark warehouse. Anxious to find my next lifting project, I steer the seatless forklift deftly down the long aisle of products ahead of me.
Manila, Philippines: JAL
This is not how one would picture a typical call center based on the 1970’s era “Jerry Lewis” telethon fundraisers. No bank of phones on desks awaiting their summons, no lively performance acts, no row of identically dressed ladies prepped and ready to record incoming pledge values. In fact, there’s not a single audible ringing noise in the entire building.
Our facility focuses on incoming calls, or “inbound processing” as the industry calls it, for Japan Airlines. Partnering with both IBM and Accenture on artificial intelligence technology in recent years has completely changed the call center dynamics at JAL.
This industry is big business for the Pilipino economy, with nearly a million people employed in various BPO, or business process outsourcing, tasks. The average worker is in their mid-20’s and well educated, with over half of the staff being female. This occupation offers many young women the chance to escape the class system and gender discrimination that is so prevalent in many emerging market Asian cultures.
Call center operations vary in size significantly here in Manila. At the low end are budget shops, with just a few operators sitting behind dented and rusty metal desks in humid, smoke-filled rooms. Not the kind of people you want to pass your credit card information along to. On the other extreme are full-service tech firms, with hundreds of service personnel, relying on extensive computing power housed on seemly endless server racks in air-conditioned closets, engulfed with the relentless hum of technology. Here at JAL, we are the later operation.
The goal, the therefore format, of the consumer call center has completely changed through the use of electronic automation. Historically viewed by companies as a required, but unfortunate, money sink, with new computerized techniques, the customer service wing of JAL is now a valuable revenue generator.
Fielding calls in the traditional manner costs money, roughly 50 yen per transaction for these types of airplane travel logistics inquiries, so speed and efficiency are placed at a premium.
Industry standard is under 5 minutes for each interaction, including the associated paperwork to be filled out, equating to just under 50k transactions in a calendar year per human call center operator’s seat. Our facility, through a variety of robotic process optimization techniques, handles 10 times that volume per work station, with greater accuracy and better customer service.
An impeccable and attentive customer experience yields repeat business, especially in the Far East where culture and tradition are important. Also, it helps that we usually know more about what the client’s needs than they do. Reviewing background information, itinerary details, and purchasing history in advance, it’s pretty easy to make predictions about incoming travel related inquiries.
At our routing facility here in Manilla, there are several different branches of the customer service organization. Electronic communication (e-mail and text, which is especially popular with the younger crowd, for flight status and other notifications), e-commerce (all things sales related including flights, meals, in-flight internet, and mileage redemptions), security (facial recognition, passports, visas, and other travel documents), and phone calls (last-minute rebookings, special VIP seat upgrades, and anything else which is so screwed up that a few strokes of the keyboard can’t easily rectify it). The last business unit is my department. All these interconnected activities flow through this JAL Virtual Customer Assistant center.
All online booking is input automatically, along with corporate business travel, which requires another layer of special attention. Large companies pay big bucks for real time support, so there’s an element of prioritization on these bookings that needs to be tactfully facilitated.
Our accomplishments over the past 18 months since implementing this AI-based customer service system have been staggering. Our primary measure of success, identified as “CX” for the customer experience, encompasses several traditional variables like post-call surveys, 3rd party polling, and in-flight comment cards, which a surprising number of travelers still take the time to fill out on airplane tray tables right alongside their country entry documents.
However, more complex analytics tell the real story. Unbiased positive feedback is revealed through voice analysis of the caller’s tone at the end of the discussion. Also, monitoring client’s phone or internet activity in the 15 minutes after they hang up, secretly of course, shows very low frequency of itinerary changes. Needless to say, Japan Airlines corporate strategy in Tokyo is already reviewing investments for the next phase of this computing and software revolution.
These days, incoming callers are blissfully unaware of how much information we have already gathered on them before the phone connection is made. Inbound data via cell phone #, IP address, or god-forbid, land lines are immediately cross-referenced against the current JAL database.
This doesn’t just include those individuals with current or future flight bookings on Japan Airlines or its affiliates, but anyone who has accessed the website on their device in the past 2 years, or clicked on one of our partner booking websites, where they often already have a user account with even more background information. This data isn’t cheap to acquire, but well worth it for enabling predictions on these customer service inquiries.
We’ve given people the option to verbally or tactile provide the number inputs for the initial automated phone questionnaire. In this era of almost universal texting adoption, nearly all users opt for the touch screen approach. Understandable, since voice recognition technology 5 years ago was terrible. However, with our current native language processing software, there’s more likelihood your fat fingers press the wrong button, than us not being able to interpret speech between the numbers “2”, “5”, and “8”.
These self-sufficient algorithms handle the most menial of incoming calls, about 75% of all volume. Most basic inquiries inevitably come from older customer’s that have not fully grasped the power of the internet yet. Or the fact that they can update their seating assignment, fix any social security # typos, or reserve a vegetarian meal online. The nice part is that the naivety of this clientele group makes them blissfully unaware that their entire phone conversation is with a robotic personality, and a dumb one at that.
Incredibly efficient and cheap to run, these computer programs manage hundreds of calls simultaneously, while updating the relevant airline reservations in real time on the JAL computer database. Granted, the cognitive function and mental speed of the person on the other end of the line is usually pretty paltry.
For these basic AI voice bots, beyond a set of core air travel specific discussion topics, responses are limited and simple. “How can I help you?”, “Will that be all?”, and “Thanks for your business.” are the extent of their more personal, social interaction coding.
The remaining 25% of more complex incoming call inquiries is where the real customer service skills are needed. To this end, our team started out at the call center with extensive language, cultural, and industry specific training.
Still, on initial calls in the beta testing phase, I was hesitant on how to interact with callers, and each situation was a new experience. Initially, all of us trainees started with a basic script, that we essentially followed step by step unwaveringly. However, over time I have gained confidence, and learned to ways to promptly and efficiently deal with a wide variety of customer service situations.
My linguistic skills have always been strong, I’m fluent in 5 languages and a variety of other local dialects. However, being able to speak in a customer’s native tongue is just a start of the communication. It’s important to be able to relate to the caller, who is often distressed, and calm their anxiety to ensure that they are providing accurate information on their problem. Cultural cues and reassuring anecdotes can go a long way on this front.
I have a very good memory, and try to take notes on each successful engagement, so the learnings can be used in the future. No two situations are identical, but little tricks related to slowing down the discussion, offering menial perks, and reassuring travel plan feasibility, can go a long way.
I also try to leverage insights from my coworkers; we keep a shared document with feedback on specific situations that can be referenced at any time. As a group, we are constantly learning, and improving, with each customer interaction. The network effect in real time.
Essentially, by the time a live caller has gotten to me, they’ve been routed through all the other phases of the standard hierarchy. This must be a special airline travel case that requires a more delicate personal touch. It’s go time.
“How can I help you?” I offer in my most generic English, seeing the incoming “+001” country code, and other more subtle details which hint at the caller’s origin.
The rapid-fire onslaught of words which follows catches even my honed linguistic skills off guard. It’s like a dam has just broken, releasing a tidal wave of commentary that has been held back with increasingly thinning patience. Apparently, that’s what 10 minutes on hold will do.
I churn through my training protocols, trying to pull clues from the conversation, while simultaneously acknowledging and inputting the personal key details.
Female . . . strong East Coast US accent . . . distressed . . .
In parallel, the flight itinerary comes together slowly through this rushed and muddled commentary. Apparently, the letter “r” is not in this woman’s vocabulary.
“Boston Logan . . . Tokyo Narita . . . delayed . . .”
The relevant data, which has already been uploaded from the initial prescreening is fortunately available to provide double check confirmation. I’ve got 5 tickets showing up under one booking code. All the last names are Sullivan, with the oldest age 71, and the youngest age 5. There’s also a note about an infant in arms for a Mrs. Sally Sullivan, age 34. That must be the lovely Irish lady currently addressing me on the phone. My dialect training is paying off already.
Slowly, the pieces of the puzzle start coming together. Apparently, 3 generations of the Sullivan family are travelling together, the first time to Asia for everyone. The goal was to watch some baseball and eat some sushi, but who knows if they’ll ever get there now. Currently, the grandparents are fatigued, and the baby needs to a diaper change. No one is at the JAL customer service desk in Boston.
Same story, different day. But you don’t get this kind of detailed color commentary on an automated computer printout.
Trying my best to calm the increasingly hysteric and emotional voice on the other end of the line with verbal soothing, I simultaneously examine the various travel options on the computer.
My powerful mind, highly skilled at multitasking, kicks into high gear. There’s one remaining feasible itinerary leaving BOS later tonight, but it’s got a layover in LAX, rather than their original direct booking. I toggle over to the weather forecast. Not looking good. A blanket of white, blue, pink, and purple covers the Doppler radar for the entire Northeast. Nothing else is getting out of Boston tonight without a plowing and deicing miracle.
There is a direct flight at the same originally planned 1:30 PM departure tomorrow, with enough seats to at least get 3 pairs together. Middle row offerings, but better than nothing. And the weather looks more promising. Maybe we can even put the old folks in first class. I enter a note for an automatic upgrade, pending availability during boarding. A nice little surprise for their old backs on the 14 hours across the Pacific.
I’m back on the line in 3 seconds.
“We’re going to put you up in a hotel near Logan Airport tonight, the shuttle will be waiting at the curb in 15 minutes. There are 2 connected rooms under your name Mrs. Sullivan, and room service in on the Japan Airlines tab. Your flight departs for Tokyo at the same time tomorrow. Is the email on file with this reservation still good for the confirmation documents?”
There’s a long pause on the other end of the line. This is the first time Mrs. Sullivan has not been talking since we connected at the beginning of the call.
“Yes, that’s perfect.” I finally hear in a hushed tone. Then, there’s more crying.
But this is different than the hysteria earlier in the call. My ear, trained for tone and emotional assessment, senses happiness rather than fatigued stress in this most recent round of liquid release. The woman from Boston is at peace, at least for now. These mental states are always difficult for me to interpret and understand.
“Glad to help, enjoy your vacation.” I reply. This last part was not in the original script, but something I have added over time. I’ve found it important to remind people they are on vacation, or going to be, if all their travel plans work out. It’s amazing how depressed humans get while on holiday.
Another Japan Airlines travel crisis averted.
I take a few seconds to scan my project status log before accepting the next incoming line. A few yellow status items to adjust based on a maintenance issue on a 787 in Honolulu, and typical “government mandated” runway restrictions in Shanghai. Plus, one VIP Japanese business client just decided he needs to go to Singapore later today on a fully booked flight. We’ll have to kick someone out of first class. Easy enough, I can monitor and adjust this trivial logistics issues in the background during my next call.
Overall, my success rate for the current session is 97% green. This status means answered, resourced, serviced, confirmed, and closed tickets. Happy travelers. I take extreme pride in my efficiency, but even more in customer satisfaction.
I check my internal clock. It’s 4 AM here in Manila, exactly 12 hours different than in Boston. In another hour or two, the group travel industry calls will start rolling in from New Zealand and Australia; these are two of JAL largest tourist destinations. No rest for the overworked.
I’ve been going for 53 hours straight, but feel great and am ready to handle the next rush of incoming calls. My record without a reboot is 3 full days, hopefully I can beat that this time.
Zurich, Switzerland: UBS
I gaze through the immaculately polished rectangular glass window, trimmed in dark mahogany that was ornately carved sometime in the distant past. Window is a modest term, since the opening is 10 feet wide and 6 feet tall, affording a perfect view into the lobby of the bank.
Granted, from the other side, the flat space on the wall is simply a digital display screen projecting key market ticker pricing and other financial data to patrons as they make their banking transactions at technologically advanced kiosks.
This facility is semi-private, reserved exclusively for the high net worth clients that the Union Bank of Switzerland has catered to, throughout various mergers and name changes, for over a century. We manage money for over half of the world’s billionaires, as measured in either Swiss francs or American dollars, which are roughly at parity currently. This portfolio of clients makes us the largest wealth management firm in the world, and precisely the reason why discrete banking facilities like this downtown Zurich office exist.
Finding and accessing this space is not easy. The only external corporate marking is small brass placard, with “UBS” in the traditionally presented red block font, and the interlocking 3 key logo in white. Mounted at waist level next to the industrial knob on a heavy, black metal door, this sign is all that identifies the office as a banking and investment center.
Entry into the main lobby yields a comfortable space, broken out into 4 semi-enclosed cubicles, each with a large touch-sensitive computer screen. This portion of the bank is open 24 hours per day. The security cameras and finger print scanning systems make it easy for ID verification around the clock, as our clients are often active in a variety of financial markets throughout the world.
Upon admission to the facility, a client’s identity is confirmed automatically, as either existing or perspective. For those individuals already in the system, account features including credit history, net worth, employment record, and other personal details are instantly populated on my front display screen.
Right now, the view through my generous window reveals that we only have two customers, which matches my computer terminal displayed user profiles. Both are banking self-sufficiently at the electronic terminals. These units are much more advanced than traditional ATMs, and allow for document scanning, credit card printing, passport application, as well as the full suite of day trading analysis tools that a Bloomberg terminal user would be jealous of.
Inside my private office, the walls are lined with soundproof paneling, not the geometric triangular foam found in a traditional recording studio, but instead dark blue cloth tacked down to create 12” squares which gives the effect of a puffy quilt hanging on a wall. This room, and my services, are for clients who need more privacy regarding their financial matters, which typically involve substantial sums of wealth.
Behind this classy office space lies the actual brains of the operation. Here is housed a bank of servers, rather than the traditional bank vault contents of cash, gold, and jewelry. In addition to the computing power needed to run the artificial intelligence systems which drive client interactions, there are innumerable accounting and investment transactions that need to be monitored in real time.
I review the background banking details on my next incoming client. $5 million investment portfolio, well diversified with $250k cash split between Euros and Swiss francs. Pretty standard all around.
Also listed is private vault storage of 50 traditionally sized 400 troy-ounce London gold bars. Checking the current gold spot price, and doing some quick math, reveals a value of roughly $35 million USD. That’s more interesting.
It’s amazing that such capital can be stored within a few empty cases of wine; granted the cardboard would not hold up to the weight. Another odd fact is that location of this gold is undisclosed. During my 3 years in this current role, I’ve only seen this notation one other time; but have a good idea what it means. Now I’m intrigued.
Right on time, the woman who has sparked this excitement enters the room. Visually, she appears to be hovering right around 40 years of age, tall and trim with short, straight blonde hair, but I know from her paperwork she’s 51. The cream-colored business suit she has on fits immaculately over her well-kept muscular figure.
Changing a toggle on my control panel, the large glass window transitions from clear to smoked, opaque grey. I gesture for her to take a seat in one of the soft leather chairs, which flank the small, nondescript metal table painted white. It’s a seemingly out-of-place and simple piece, considering the refined décor in the rest of the space, but every element of this room has a purpose. This subtle table houses more pieces of technology than an airport security checkpoint.
“A drink, perhaps?” I offer congenially.
Though I know the woman’s name from the bank’s bio, I plan to not use it. Anonymity is best in these kinds of transactions; it gives the client more confidence in our security protocols. However, it’s still important to convey an image of civility and be a gracious host.
For now, we’re just two middle age women having a casual conversation. She will have to reveal more than her name soon enough to access the banking funds in question. Here, we don’t use the numbered Swiss bank accounts of movie lore, but instead have other ways to protect bank-client privilege.
“A kirsch and a water, please.”
Her direct request does not surprise me. Sipping on some quality booze is not uncommon during these types of life-altering transactions, and her German roots narrowed the field on preferred local liqueur.
Almost before the words have left her mouth, an opening appears in the blue paneled wall, and a small metal tray rotates out, stopping just shy of the left armrest on the chair she is seated in. On the tray sit two delicate crystal rocks glasses with nearly identical clear liquids, aside from the tiny bubbles which drift slowly to the top of the sparking water.
In reality, we only have about 20 spirits, along with a variety of sodas, juices, and hot drinks to choose from. I’m sensing it’s not this woman’s first time in our secure banking facility.
The first words out of her mouth, drink order or otherwise, allow my team to start the authentication process. Voice recognition, while imperfect, is a good baseline screen to verify identity, and can be executed without inconveniencing the client, or even having them know. I also have a deftly trained ear and am cross-referencing this woman’s speech volume, cadence, tone back to the voice reference samples on file. She seems legitimate thus far.
Digital security is one of our primary considerations here at UBS. Each year we spend more on firewall procedures and other malware deterrents than the governments of most countries. With our ultra-rich clientele and broad basket of financial holdings, electronic hacking and blackmail are always front of mind, requiring constant cyberspace vigilance and technology innovation.
One ever-present concern is “keylogging”, which involves covert recording of keystrokes on a computer or phone. This can be done remotely through software hacks, and is a way for criminals to acquire account #’s, usernames, PIN codes, passwords, e-mail contacts, and other sensitive information from potential marks. This is one reason why other physical user verification systems are needed.
However, at UBS we have been able to turn this technology around for beneficial purposes as well. Keystroke dynamics can verify specific user IDs based on how a person types on a keyboard, or enters their password on a cell phone. Security is not just about what password someone types, but also the cadence at which they enter the details.
The previously boring and non-descript white table now becomes relevant. The tablet monitor in the center of the table turns black, revealing a touchscreen layout of a traditional computer keyboard on its smooth surface.
“Password please,” I offer gently, but our energetic guest is already typing away.
When we open each new account, we require the user to generate a password. There are no character or format restrictions other than it must be repeatable. We record the keystrokes for timing, pressure, and of course accuracy, during the initial interview on this same flat keyboard format. That same deft touch must now be replicated.
Voice and password out of the way, we can get down to business.
“How can I help you?”
“I would like to move 5% of my total gold holdings into Bitcoin,” she replies calmly, before taking a delicate sip of her kirsch.
“Would you like to diversity across a basket of cryptocurrencies perhaps?” Then I add, “In fact, UBS has our own blockchain technology for inter-bank transactions. We are well-versed in the space.” My internship with the sales team has never left me apparently.
“No thank you. Just Bitcoin please. I would like to average into the position over the next 3 months. Weekly purchases with strategic timing on buy points based on technical market trends. I trust your trading team can handle this?”
A woman who know what she wants. I appreciate that. Sure, our team can execute this order, but what is the goal and reasoning behind this seemingly well thought out transaction.
I nod silently to her. I don’t get paid to ask questions. Especially not about changes to asset allocation. My function is to provide precise and timely banking functionality.
“With your amount of assets, this will require full security authorization. I hope you understand.”
It’s now my guests turn to nod silently. No doubt she knew the hoops she would have to jump through when she set up the meeting.
Noiselessly, the center of the table raises and rotates 90°, aligning at eye level with the German woman. She holds her head stoically for the required 15 seconds as a band of infrared light flits across her iris’s twice.
In real time, a camera on the same pedestal is capturing how the IR light reflects off the eye. The human iris has 266 unique patterns, while fingerprints only have about 40 discerning characteristics, so the optical analysis provides a much more refined identity assessment.
In parallel, and unbeknownst to her, or anyone not on the UBS security technology team, 3 other cameras in the room have taken video which allows for triangulation of a digital facial map. Once you have someone who isn’t moving, may as well capture as much data as possible.
Currently, we don’t have this 3D facial data on all of our clients, since the iris capture was originally done using more traditional goggle style methods, so this is simply a chance for us to enhance our security protocols for the future.
Our software team is even working on complex aging algorithms, since in the high net worth space much of the wealth transfer is generational, and we may not see a client for 20 years or more.
Content with the ocular findings, the tablet collapses and stows back in its tabletop recess with the same silent motion.
As we go through these security screening procedures, I’m carefully monitoring our client’s body language. Heat sensors provide insight on if the individual is sweating, which implies nervousness, while impulsive twitches like a jittery leg or a knit brow are telltale indicators of distress. Our lady displays none of these signs of agitation as she methodically executes the battery of authentication protocols.
Time for our most stringent security check, the palm scan. In over 1 billion palm scan checks done to date globally, the error rate has been 0.00001%, with not a single false positive measurement. If executed correctly, this verification technique is essential flawless.
Placing her right-hand palm down and flat about an inch off the table top, the white surface suddenly comes to life with a dim glow. The biometric scanner uses near-infrared light to map and record the vein pattern in an individual’s hand. Deoxygenized hemoglobin in the veins absorbs more near-infrared light, resulting in black lines on the reflected image captured by the camera.
A further benefit is that these palm vein details will not change noticeably over an adult’s lifetime, and are not affected by skin issues like cuts, burns, or callouses, making this technique ideal for banking client identification over long periods of time. Also, though morbid, palm scans require blood to be flowing through the subject’s body, while finger print ridge profiles do not. We must consider all potential circumstances for our security protocols.
Two hand outlines appear on the database processing screen, which I have visibility to. Wriggling black vein lines crisscross through the middle of each image. The computer cross-references the original and recently taken palm scan, checking a multitude of data points between the image maps.
The 4-stage security authentication process is complete. Our lady is all clear for the requested banking transfer.
Moving on to financial portion of the process, the cogs of my finely tuned mathematical mind start churning; liquidity, valuation, diversification, tax implications . . .
Taxes. Since these specific secretive gold holdings don’t need to be reported to either the Swiss or German government annually, we may have some flexibility there.
“As you know, your gold is stored in a secure vault of undisclosed location in the Swiss Alps. As such, it will take us time to secure and liquidate the 5% position you requested. However, based on your excellent financial standing at UBS, and your family’s history with our bank, we can extend a line of credit to begin the Bitcoin purchase transactions this week. Do you want to proceed?”
“Yes, very much so.” Another controlled response, followed by a final swig of cherry liqueur.
“Transactions will be initiated immediately. Please look for the electronic money transfers and trade confirmation details via the usual discrete channels.
Silently, the woman rises from the fancy leather chair, her cream business suit still crisp even after 30 minutes being seated. She moves forward slowly, passing through my shimmering holographic image, briefly blocking the light beams from the ceiling projectors above and reflective floor surface below, and heads for the black metal door which leads back to the real world.
Beijing, China: KFC
Three flashing red numerals identify the quantity of patrons currently pending service. In reality, this value is an extrapolation of the incoming order frequency via digital channels, as well as the touch screens located just inside the door of our restaurant. Granted, behind the scenes GPS tracking for anyone who has downloaded the KFC app allows for pinpoint location after an order, and heat maps in the dining space can take an accurate count of seated persons in the physical eatery.
Still, it’s important for us to stay one step ahead of these static counts, adjusting order timing if someone on route for a food pick-up gets stuck in traffic, or if an especially larger catering order is pending based on scraping business invoices from the past few weeks.
This Kentucky Fried Chicken location in downtown Beijing is the largest in the country. The Chinese patrons love their fried chicken, especially the younger, millennial generation. This restaurant, with seating for over 500 people, offers a convenient spot for social meetups over an affordable meal. Combine that with the steady rush of scooter-delivered take-out orders, particularly around lunch time when the tourism and corporate demands coalesce, and the kitchen is typically running flat out.
I check the temperature sensors throughout the facility, an operation I execute methodically every 15 minutes. It’s important to monitor the internal conditions, since the pressure fryers can put out a lot of heat. The outside environment is less controllable, but equally important. Weather can have a significant influence on our customers’ ordering patterns.
The kitchen zone is running at 28°C and 65% relative humidity, with the dining space monitors averaging out 3°C and 10% lower. A little warmer and steamy than desired, but no reason to adjust any cooking processes or HVAC units at this time. The weather outside is quite different. 5°C at best, with humidity pushing 80%. A cold rain is pending. Time to fire up more lotus soup mix, and a get additional carafes of black tea steeping.
I switch my attention to the live cameras. Digital temperature readouts are only one piece of the overall restaurant ambiance puzzle. Clearly visible is a row of 10 austere white oversized tablet displays on telescoping vertical poles, which sense and adjust to the user’s height. The integrated camera and scanner systems allow for automatic payment, while documenting several other key visual details of each customer for future analysis of course. We can already predict what a repeat customer will order when they walk through the door with 85% accuracy. No algorithms to account for spontaneity yet unfortunately.
Currently, only four of these stations are occupied, not because the restaurant is empty, but because most patrons favor ordering with their own cellphone.
The real activity happens behind the counter, in the kitchen of sorts.
One of the best elements of our modern KFC kitchen layout is keeping human interaction with the food to an absolute minimum. There’s rarely a physical worker in the space, but all the orders flow through with precision accuracy and timeliness. This eliminates many of the concerns associated with traditional fast food preparation: poor sanitation, inconsistent cooking, and a general lack of attention to detail.
The layout of the kitchen is based on the high efficiency and low mobility traits of these food service “workers”. Food moves along conveyor belts with various mechanical arms providing additive inputs to each order. Turn tables, vertical shelves, and pressurized dispensers all help to deliver the diversity of potential operations required on each subsequent dish.
At the far end of the serving assembly line, I can see the other key offerings that will provide a balance meal for our patrons. The side dishes have been streamlined; large vats of each menu item ready for distribution.
On cue, one of our signature red and white striped paper tubs drops from the dispenser and moves along the black rubber conveyor belt. It stops in the desired position, directly under a hopper of piping hot mashed potatoes. Servo motors controlling the trap door at the bottom of the chute open and close automatically, depositing the perfect amount of potatoes into the tub.
Still steaming, this bowl of fluffy starch moves further down the line, where a directed stream of brown gravy covers the top of the mound. Lastly, a suction cup plunger picks up and places a clear plastic lid on the tub, then executes a 360° spinning motion during which pressure is applied locally to the lid via a tapered arm which resembles a reverse can opener.
The result is a perfectly portioned 150 milliliter bowl of mashed potatoes; brown gravy smeared across the inside of the lid without a single drop spilling. Automatic food dispensing perfection at work.
Enough distracted gloating, back to the task at hand.
To my right, a few simple mechanical devices are executing the menial chicken preparation tasks before my critical cooking operation. The chicken arrives in 25 kg tubs, already butchered and processed into the standard individual Kentucky Fried Chicken offerings: legs, thighs, breasts, and wings.
This chicken is never frozen throughout the transportation and storage process. Each piece of chicken gets the exact same treatment. This is the traditional KFC preparation upon which the company has built its heritage. On the production line, we call it Program Secret 7.
Chicken moves in 1.5 kg batches, which is roughly 15 pieces. This amount of meat is maintained throughout the staging process, and represents coverage for one rack in our proprietary pressure frying process. Each batch of chicken parts is dunked in the brining liquid for precisely 7 seconds, then dried for the same 7 second duration, aided by vertical convection air from above.
Next, the meat is dumped into a tray of the Colonel Sanders’ unique dry breading mix. In addition to the well-known ingredients of flour, salt, milk powder, and egg powder, are a proprietary blend of herbs and spices.
During employee orientation, we learned that KFC’s 11-ingredient aromatic blend is an industry secret, and only 2 humans alive know the recipe. Bags of this premixed dry rub arrive in palleted bags directly from the United States, and are kept closely guarded at each restaurant facility, until used to make our famous fried chicken here in Beijing, or anywhere else around the globe.
After all the chicken pieces are sufficiently coated via agitation from the mechanized shaker table located underneath the dry rub tub, mechanical grips extract the basket of chicken from its powdery bath, and provide a seesaw tumble shaking operation, you guessed it, 7 times. This basket of perfectly breaded chicken, just enough meat to fill one rack in the pressure fryer, is now in my capable digits.
One of the most critical tasks in the Kentucky Fried Chicken cooking process is arranging the pieces on the rack. As such, I’m very proud to be in charge of this element of the operation, taking great pride in accuracy and repeatability.
It’s important to fit as many pieces as possible on each rack to improve efficiency, but if the meat is overcrowded then it will not have enough surface contact with the oil, which leads to undercooking.
During training, we spent countless hours learning how to optimize this placement for different sizes of legs, breast, thighs, and wings, as well as accounting for the different anatomical shape that each of these appendages can come in. Another trick is to tuck the tips of the wings underneath; the wings are the thinnest part of any chicken component, and therefore are prone to overcooking.
Scanning the batch I’m currently working on, I mentally run through the various configuration options based on the distribution of pieces in the metal breading basket next to me. For simplicity, I break these organic chunks of meat into simplified geometric shapes which closely represent the surface area each one takes up on the rack.
Thighs are essentially squares, breasts can be rectangular or trapezoidal depending on their size, Legs are elongated triangles, and wings are small, easily manipulated, rectangles, that can be used to fill the negative space.
Content with my positioning, I lift and load this rack of raw chicken, the 4th and final of this cooking round, into place. It’s very important to close and lock the pressure frying device. This is a mandatory step that has been drilled into my mind by managers, and I always double check before initiating the cooking process. Safety is paramount at KFC.
The pressure frying process takes just under 13 minutes; 12 minutes and 40 seconds to be exact. I am obsessed with perfection. Apparently, despite all his wisdom and superstition, Colonel Sanders couldn’t come up with a cooking process that takes 7 minutes, or an even multiple thereof. Pressure frying allows for faster cooking times, with more retained moisture in the meat, and improved oil longevity. All good results when running a high throughput fast food establishment.
I can arrange each tray of chicken in 1.5 minutes, even with an unfavorable mix of odd shapes that requires a few repositions. Managing two fryers, located back to back with identical preparation lines, leaves me with a just over a minute between each cooking changeover cycle.
After arranging each batch, I make sure to use the extra time to take the necessary cleaning precautions which avoid cross-contamination between raw and cooked chicken. I dip my joints and appendages into the sanitizing liquid, agitating them vigorously, and staying submerged for the required 20 seconds to ensure complete sterilization. We take pride in our cleanliness here.
Even from my secluded location in the back of the kitchen, I can see most of the other preparation stations. Also, I have visibility to all the orders that come though the digital point of sale system. Because of our policy to not serve chicken any longer that 90 minutes after it’s been fried, managing throughput on the pressure fryer is another one of my important tasks.
Looking at the order lists, I’m always amazed by how many desserts we sell. The egg tarts seem to be especially popular. These beauties are incredibly easy to produce with no human interaction. The flaky crust is mixed, with the key wet and dry ingredients being fed via gravity from long, cylindrical tubes at the required baking precision, then mixed in what could be considered a glorified oversized food processor.
In the most complex step of the egg tart procession, a robot with a modified ice cream scoop device extracts the desired amount of dough from the mixing bowl, then drops the lump into one cavity of the continuous baking tray belt which is slowly drifting by underneath. Once the deposition process is completed 4 times across each row, a matching mandrel comes down and simultaneously forms this quartet of dough balls into the desired concave pie cups they will inevitably become.
On their way to the heating element, 4 identical, aligned pastry bags of premixed egg custard are squeezed via an airbag sleeve. Fine flowmeters monitor the dispensing of this viscous paste and adjust the backpressure accordingly. Continuing down the line into the induction heating elements, these tasty morsels only spend 5 minutes cooking.
The clever part of this silicone conveyor belt system is that it’s flexible. As the belt rolls over the bar at the end of the line, toasted pies being dispensed onto a sheet pan for final packaging and serving, it goes inverted for the return journey.
Underneath the baking zone, in a moisture sealed chamber, the silicone pie cavities are streamed, rinsed, and dried via a traveling dish washer system. Right before popping back into their natural concave position at the start of the assembly line, a gentle bristled brush applies a fine layer of canola oil to the inside of the cake trays. This lubricant serves as a natural mold release, while also providing diners with the golden-brown color and crisp, but flakey, crust they expect on their tarts.
I have a great appreciation for technology, and enjoy networking with the other food preparation devices in our KFC restaurant kitchen to collaborate on new ideas.
The pressure fryer alarm dings, and I turn back to the task at hand. The top of this device rises automatically like an elevator car, the exposed metal racks of meat spewing scalding hot steam. Reaching in with no gloves, I extract a scorching rack of chicken pieces and place it on the drip tray. If only I could tell if these crunchy, bronze morsels of chicken smell as good as they look.
As I prepare the pressure fryer for the next round of raw chicken parts. I notice one of my favorite songs coming on through the restaurant’s sound system. When ordering via the on-site electronic kiosks, patrons are allowed to select a song to put in the queue while they wait for their food. Not as engaging as the karaoke bars which dominate the social life in China, but a touching gesture.
Over the well distributed speakers lining this vast KFC dining space, the familiar Styx chorus erupts in hybridized Japanese.
“Domo Arigato, Mr. Roboto.”
