Last week, President Trump addressed the nation on the coronavirus pandemic. Countries around the world have been taking aggressive measures to stop its spread. Closing borders, emergency funding programs, and city and countrywide mandated quarantines are just a few of the actions taken to help combat this invisible enemy.
The disruption caused by this pandemic is being felt by many. Various industries are on the verge of shutting down simply because people can no longer do their work. This is because so much of our work still relies on physical, human interaction, which exacerbates the spread of COVID-19. Separating the healthy from the ill has also posed challenges, as people can be infectious without any visible symptoms.
Since the strength of our economy hinges tightly on our ability to continue doing our work and being productive despite the circumstances, the effect of this pandemic on our economy is paramount. This is reflected in the downward spiral trend of our financial market over the past few weeks.
Digital Transformation: The First Step Towards a More Resilient Workforce
I am fortunate to work in the hi-tech software industry. I can already work virtually and remain productive when I need to. Although software companies across the country are instituting work from home policies, part of my work is still disrupted. Many of the conferences that I would be keynoting have either been postponed, canceled or turned virtual.
The virtualization of my work made me wonder about ways artificial intelligence (AI) could help make our workforce more resilient in times of global emergencies like this. Once our work is virtualized, it can be digitized, tracked, turned into data and used to train an AI system to mimic how we work. The bigger question is how will AI change the way we work in the near and far future?
Let’s take a journey into the future and examine this question together.
There is no doubt that AI is going to automate more and more of our digital work in the near future. This will start with the most mundane and repetitive tasks, but AI will eventually automate some creative tasks that are digitized. But, before this can happen, we must shift more of our physical work into digital work that can be tracked digitally. This is already well underway today with many industries embarking on journeys of digital transformation.
Phase 1: The Exchange of Data for Automation
The advent of digital technology has already shifted much of our physical work to digital. Consequently, AI-based technologies are becoming more pervasive in our work. This will change the way we work in the following ways.
First, we will exchange data for automation by providing digital records of our work as training data for the AI to learn and mimic how we do our work. Machine automation offers huge efficiency gains (in both speed and scale), so businesses must adopt it to compete more effectively. For employees, however, it’s simply more convenient and less work for us if we let machines automate the mundane and repetitive tasks.
Although machines are fast and scalable, they must not degrade the quality of our work. Otherwise, it may not be worthwhile to trade efficiency for lower quality work. Today, certain AI can outperform humans in specific tasks, but machines cannot replicate human quality in most of our work just yet. However, if sufficient data is available to train AI well enough to mimic part of our work, then we move into the next phase.
Related Article: 3 Strategies to Maximize the Benefits of Intelligent Applications
Phase 2: Teaching the AI to Refine Its Model of How We Work
Once we have AI that is good enough to help us, we move to collaboration. We let AI attempt to replicate our work, and we judge to see if it’s able to match our quality. Who better to do this than us humans?
This will essentially divide up the work so that AI and humans can each do what they do best, respectively. In cases where AI can replicate near-human quality work, AI can be leveraged to automate those and offer efficiency. Where AI fails to meet the acceptable quality, humans can take over those cases. Meanwhile, because all of our work is digitized, those challenging cases become new data from which the AI learns and refines its model of how we work.
In the above scenario, AI is like an apprentice learning from its human masters through the digitized records of our work.
Phase 3: Full AI Automation and Job Shift
As we continue to collect more data on the challenging case where AI fails, AI may eventually learn to automate those cases as well as we can. Now, our apprentice is ready to graduate. This AI can now automate our work with minimal human supervision.
So what do we do then? We do what we’ve always done at work: We move on and do something else. For example, we become managers. This shifts the nature of our work significantly. We assume more responsibility, and to some extent, we do less hands-on work.
Likewise, the complete automation of our work by AI means that our jobs will shift. This will shift the nature of our work to something more important and more strategic. It will also shift our work towards tasks that require more human empathy (e.g. management and leadership), which is hard to digitize, and therefore not easily automated by AI. Frankly, this work is likely more interesting and more fulfilling.
A Concrete Example: Automating Tasks for Sales Reps
Although enterprise sales reps have to do many tasks, we’ll focus on a short stereotypic sequence of tasks in the work:
- Task 1: Analyze the transaction and CRM data to determine which prospect to pursue.
- Task 2: Email the prospect to engage them with marketing collaterals to create interest.
- Task 3: When it’s time, set up a meeting to meet the prospect in person.
- Task 4: Go meet the prospect to sell and negotiate the deal.
We already have AI tools today that take transaction/CRM data, automate their analysis (Task 1), and recommend opportunities for sales reps to pursue. They cannot replicate human-level decisions yet, so sales reps still need to either accept or reject the recommendations. As they do so, the AI analyst will refine its model so it can recommend more relevant opportunities in the future. Eventually, the recommendations will be so good that the sales rep will accept most of them. At this time, the sales rep can choose to fully automate the decision of which opportunity to pursue and accept the recommendations from the AI analyst and go after them.
The full automation of Task 1 will shift the sales rep’s job downstream. Now, the sales rep just needs to email the prospects and engage them. Soon, however, we may be able to take all the prospecting emails from a sales rep and use them to train an AI to generate emails. This AI writer will take a list of recommended opportunities from the AI analyst and automate the creation of emails (Task 2). The sales rep will need to decide if the generated email is good enough by either sending it as is (i.e. good enough) or tweaking it before send (i.e. not good enough). Meanwhile, the AI writer will learn this sales rep’s writing style, how he engages different customers, etc. from his modification of the generated emails. Eventually, this AI writer will learn and be able to generate emails that are so good that the sales rep will just send out most as is. Then he can choose to fully automate this process to save time and let the AI writer engage the prospect until the prospect is ready for a meeting.
The full automation of Task 2 will again shift the sales rep’s job downstream. Now the sales rep just needs to set up meetings for prospects who are ready to meet. But soon, we may be able to take the sales rep’s calendar data and use it to train an AI planner that will automate the planning of his itineraries (Task 3). Again, the sales rep can either accept the itinerary, or reject and change it to his liking. As he does this, the AI planner will learn this sales rep’s travel habits and preferences. Ultimately, this AI planner will be able to create an itinerary that is so good that the sales rep will accept most of it. Then he can choose to automate this again to save time and let the AI planner plan his travels.
The full automation of Task 3 will shift the sales rep’s job further downstream. Can you see the pattern? Now the sales rep only has to follow the itineraries created for him and go meet the prospects. If a sales rep is free from doing Tasks 1, 2 and 3, and now only required to do Task 4, imagine how many more prospects he can meet. This could significantly increase the productivity and efficiency of the entire sales operation. Moreover, this job is now more important and more interesting for salespeople.
The Last Mile of Creating a Resilient AI-Augmented Economy
With the help of remote conferencing technologies, I can hold meetings, and even give a keynote across the world. But what’s preventing us from digitizing all of our work?
At the moment, we can’t, because our economy is so service-dominant. Although this should, in theory, make digitization easier, many service industries have a well-known “last mile” delivery problem. Consumers can explore, research, configure, and purchase products or services online, but not all of them are downloadable with a click of a mouse. Many products and services still need to be delivered or performed in-person by humans.
Currently, the primary use case of autonomous AI is in self-driving cars. But this will shift to other autonomous systems, such as robotics, autonomous drones, IoT, and other smart devices and connected machines.
If you live in the San Francisco Bay Area, you can already see autonomous AI helping ecommerce automate this “last mile” delivery of physical goods. Food and package delivery robots (e.g. Starship, Kiwibot, etc.) are appearing on select campuses and cities. Robotic wait staffs are already taking orders and fetching food in restaurants while robotic baristas, chef assistants and bartenders are automating the roles of their human counterparts. Although these robots may not match their human counterparts in everything they do, they are often able to execute most of the essential tasks equally as well, if not better.
To create a workforce that is resilient to a global pandemic like coronavirus, companies have two choices: they can shift more of their physical work to digital, which can eventually be automated by AI, or they can directly automate their physical work using advanced robotics powered by autonomous AI.
By allowing AI to augment the roles of humans, we can eliminate the need for human interaction without halting a significant part of our economy. Because these robots can function autonomously, receive instructions remotely, or be controlled remotely, crucial functions of our society can still be executed when human interaction is not advisable.
Related Article: Before You Hand Human Resources Over to AI ...
What Would a Future AI-Augmented Society Look Like?
What would such an AI-augmented future look like? It would seem that such a resilient AI-augmented workforce could also render a large part of the human population redundant. What happens after the pandemics then? Wouldn’t this create another crisis where people’s jobs are displaced?
Some jobs will inevitably be replaced by AI in the immediate future. But this will happen, and it has happened with every technological innovation in human history. The effect is more dramatic with AI because it is a general technology that is very versatile and can be applied in many different ways across many business domains.
However, as I look further into the future, I am constantly reminded that we should not use today’s standard to make a judgment about what may be a wildly different future.
In a more distant future, jobs may no longer be required for the survival and well-being of individuals. This may sound farfetched, but it’s not implausible. Numerous countries have already been experimenting with Universal Basic Income. Moreover, humans have dramatically reduced the amount of time they spend working ever since the Industrial Revolution due to the efficiency we gained through the employment of mechanical machines. It’s conceivable this trend could progress further with the help of intelligent machines.
In this distant future, AI and robots could automate many of the crucial operations that are necessary for a society to function without too much human supervision.
AI can not only help us build a more resilient economy able to withstand a pandemic like coronavirus, it may even eliminate our dependency on an economy altogether. We are a long way away from this utopian future, but I believe we can get there with the help of AI.
In the meantime, stay healthy and live to witness the AI-augmented future.