Here is an image that often gets thrown around to illustrate the consequences of IT automation in the workforce. The map shows the most prevalent job in each state of the USA. You can clearly see that there are a lot of truck drivers. A recent estimate puts the total number of truck drivers in the US at 3.5 million.
Now think about all the companies developing self-driving cars: Apple, Uber, Tesla, Daimler… what are all these truck drivers going to do when self-driving cars hit the road? What is going to happen to roadside cafeterias and motels? How are garages and gas stations going to stay in business?
It’s hard to estimate the total number of jobs lost to self-driving trucks, but some estimates predict that 6 million jobs will be lost in the US in the next 15 years.
Oh well, you might be thinking. Those are all blue-collar jobs. I have a cushy white-color middle management job so that means I’m safe for the moment.
Think again. You’re next.
So what does a manager do all day?
All large companies have a corporate ladder, with entry level recruits at the bottom of the ladder, and the executive team at the top. The employees at the bottom of the ladder perform clearly defined low-level tasks, and the executive team at the top of the ladder does high-level work: creating a strategy, introducing new business processes, defining the corporate brand, talking to investors, etc.
In the middle of the ladder are the managers. These people run the day to day operations of the company by executing well-defined business processes. A business process is nothing more than a collection of linked tasks which result in the delivery of a service or product to a client. Every business process accomplishes an explicit organizational goal.
There’s nothing mysterious about business processes, companies have been using and documenting them for years. In fact, if you want to become ISO-certified, you need to have all your business processes documented and stored somewhere for reference.
A typical process looks something like this:
Companies use software to manage business processes, but the software is not very advanced. As of today, it can only do this:
- Visually present the business process and store it for reference.
- Change the user interface of an application, based on where we are in the business process.
- Move packets of data between business systems, based on where we are in the business process.
Managers, aided by Business Process Modelling Software, do all the heavy lifting: hire new recruits, assemble teams to execute the process, monitor progress, adjust planning & remove impediments, gather deliverables, and create the final reports.
Some of these tasks are extremely hard to automate with AI. For example, recruitment is almost impossible to automate and will remain a manual task for many years to come.
So are managers off the hook for now?
Let’s take a step back.
An unorthodox solution to the inventory problem
Eleven years ago I was involved in a project to design the supermarket of the future. Our goal was to build a piece of software that kept track of inventory levels in real-time and knew exactly where each product was in the supermarket. We experimented with RFID readers and tags and built a pretty good proof of concept. Our shelf of the future had embedded RFID readers and two attached displays to show narrowcasting commercials when someone picked up a product.
But we never cracked the inventory problem. Fitting a warehouse with floor-to-ceiling RFID readers is prohibitively expensive, tracking accuracy was way too low to be effective, and product margins were so small that nobody could afford to attach the RFID tags to individual products in the first place. We had to estimate inventory based on daily transactions, just like everyone else was doing.
But there’s a much easier way to track inventory, and I can kick myself for not seeing it back then. Marshall Brain brilliantly describes the method in his 2003 novel Manna. In his story, an AI system is running a fast-food restaurant. All employees carry a headset and can talk to the system, and it talks back to them. Here’s how the AI checks the inventory level for hamburgers:
AI (talking to an employee): “John, I need you to go to the storeroom. Tell me when you get there”
John: “I’m in the storeroom”
AI: “Okay. John, how many boxes of hamburgers do you see in the storeroom?”
John: “I see 24 boxes”
AI: “Okay, that’s 24 boxes. Thank you, John”
And that’s it. A simple conversational interface, with an employee who does the checking. No RFID required, no readers, no expensive tags. Just a computer asking an employee to count boxes.
We often picture AI as a crew of robots taking over a shop or factory, but the Manna software achieves the same results in a much simpler way: any task it cannot do by itself (i.e. counting boxes in the storeroom, mopping the floor, assembling a hamburger) it assigns to an employee. That just leaves tracking daily tasks, assigning tasks to employees, monitoring progress, and adjusting the planning where needed.
Sound familiar? That’s exactly what a manager does.
I read Manna three years ago, and what struck me is how easy it would be to write the Manna software. The AI is basically running and monitoring a business process. All the tasks are assigned to people, who report back directly to the system. The only complicated piece is the conversational interface, which requires high-quality speech synthesis and recognition.
Could we build something like Manna today?
The rise of the iCEO
Devin Fidler heads the Workable Futures Initiative at the Institute for the Future, an independent, non-profit research organization based in Silicon Valley. And he asked himself this same question: can we build a system like Manna today?
Last year he announced that his team has been working on a prototype system, dubbed iCEO. The prototype automates complex work by dividing it into small micro-tasks and assigns them to workers using oDesk, Elance, Amazon Mechanical Turk, Uber, and plain email/text messaging.
Devin tested the system by giving it a tough task: create a 124-page research report for a Fortune 50 company.
iCEO got to work and asked workers on Amazon’s Mechanical Turk to curate a list of articles on the topic. The list of articles was then passed on to a pool of technical analysts from oDesk, who extracted and arranged the articles’ key insights. A cohort of Elance writers then turned these into a coherent text, which went to another pool of subject matter experts for review, passing them on to a sequence of oDesk editors, proofreaders, and fact checkers.
iCEO routed tasks across 23 people from all over the world. Devin and his team rarely had to intervene, and instead stood back and watched iCEO execute the project autonomously. iCEO even ran recruitment and assembled the team of 23 all by itself.
The final report was created roughly 5 times faster than a human team could, at the same level of quality. This is the first time someone managed to build a system like Manna in real life.
So what does this all mean?
iCEO is just a prototype. A proof of concept to demonstrate that modern software can run a business process autonomously, eliminating any need for executive management. But business owners are taking notice.
Last week Bridgewater Associates, the largest hedge fund in the world, announced that it is building a piece of software to automate all day-to-day management of the firm, including hiring, firing and other strategic decision-making. The development team is headed by David Ferrucci, who earned his chops leading IBM’s development of Watson, the supercomputer that beat humans at Jeopardy! in 2011.
The fund’s billionaire founder Ray Dalio wants to use this software to ensure the company runs according to his vision. The goal is to build a new business operating system called PriOS, that will make three-quarters of all management decisions within five years. The kinds of decisions PriOS will make include finding the right staff for particular job openings and ranking opposing perspectives from multiple team members when there’s a disagreement about how to proceed. The machine will make all decisions, according to a set of principles laid out by Dalio about the company vision.
There’s no doubt Ray will succeed. What he’s attempting is very ambitious but not impossible, and just within reach of today’s technology. He will have his operating system in 5 years.
Now think about what that means. In five years we will have a template for an AI that automates executive management. Other companies can simply copy what Bridgewater did, plug in their own company vision, and let the software loose within their organizations.
This will completely blow out the middle of the corporate ladder. And the shareholders will love it because all the wages in the middle of the ladder will disappear. By eliminating executive management, the shareholders massively reduce cost and push profits through the roof. They will be laughing all the way to the bank.
Millions of executive managers could be out of a job in 10 years, joining the truck drivers at the unemployment office.
A bleak or bright new future?
It’s not all doom and gloom, however. This automation, rapid as it may be approaching, is simply turning corporations into software constructs. There’s still lots of room at the top and bottom of the corporate ladder. Here’s what you can do:
You can focus on the bottom of the corporate ladder and join the gig economy. Register with ODesk, Elance, or Uber, and pick up all the work that’s not being done by robots yet. Things like HR, QA, research, fact-finding, assessments, etc. But keep in mind that this may well become the new blue-collar work of the future.
A better strategy is to move up the corporate ladder. Become a business analyst and program the AI’s with company visions. Someone needs to load the business processes into the AI’s before they’re switched on, and this will be a human job for a long time. Your job will be safe for at least a decade.
But here’s a thought: you could also become a company owner.
Think about it. In my vision of the future, corporations are nothing more than software. If you have access to this software you can be a CEO too. All you need is a good idea, chop it up into a business process, load it into the AI, and flip the switch. You could be the CEO of your own virtual company with a pool of 50-100 employees from all over the world.
In this future ideas will become the dominant currency, and the means to execute ideas will be abundant, accessible, and cheap.
Which ideas will you have? And what will you do with them?
How will you change the world?
Also published on Medium.