Credit: Sébastien Thibault
Ideas Made to Matter
What happens to industry and employment after COVID-19?
Numbers tell the story about the COVID-19 pandemic’s immediate impact on the U.S. economy. A 4.8% decline in gross domestic product in the first three months of 2020. More than 40 million unemployment claims in 10 weeks. An unemployment rate close to 15%.
These numbers are stunning, even compared to the Great Recession of 2007-2009. But the comparison helps explain what the economy may look like once the immediate impact of the pandemic has subsided.
“Firms used the Great Recession to change the structure of their workforce to focus on more technology-centered production,” said Sarah Bana, a postdoctoral associate at the MIT Initiative on the Digital Economy. “This will be at the forefront of the response to COVID-19.”
The presence of technology to support remote workers has taken center stage in the last several weeks. Video conferencing, collaboration software, and customer service chatbots have allowed knowledge workers to remain productive while working at home. At the same time, the absence of technology such as machine learning has left the manufacturing industry at a loss to manage rapidly shifting supply chains and the health care industry struggling to compile and analyze data on the effectiveness of COVID-19 treatments.
Ideas Made to Matter spoke with Bana and four other experts at the MIT Initiative on the Digital Economy to get their perspective on how technology will impact the future of work, for companies as well as employees, in the months and years ahead. Interviews have been edited and condensed for clarity and conciseness.
‘There’s going to be a big division between nimble and slow companies’
Erik Brynjolfsson: director, MIT Initiative on the Digital Economy, professor, Information Technology
This is a huge shock to the economy, but it’s also a big restructuring of the economy. There’s going to be a fundamental change in the way companies work. We’re compressing 10 years of structural change into 10 weeks.
A big part of that is a shift to a more digital economy. That has been going on for some time and will continue for some time, but the current situation is compressing the efforts of a lot of businesses, managers, and workers to become more digital. Even when you have technologies in place like Zoom or Slack, people don’t always make full use of them; they stick to the old ways of doing things, and for most companies, if it ain’t broke, they don’t fix it. Most of us have been forced to think about whether we can work remotely. It hasn’t worked for everyone, but overall digital tech has been quite impressive, so that’s a win.
According to the productivity J curve, companies don’t adopt powerful technology right away — whether it’s artificial intelligence and digitization, or electricity and the steam engine. When they do adopt the technology, productivity goes down, and it can be one or two decades before they realize the benefits.
Right now, companies have no choice but to restructure. Because it’s dangerous for people to be next to each other, companies are trying to think about how to produce value and input without people getting together. There are two options: Have employees work remotely, or have robots and machines do some of the work. Companies are scrambling to see if they can do one or both of those things.
A paper I recently co-authored shows a huge increase in remote work. One-sixth of Americans were already working at home, and another one-third switched from working in the office to working remotely. That means half the U.S. workforce is working at home. In particular, information workers like managers and professionals are much more likely to be working from home — and once you have people work remotely, they don’t have to be living in the same neighborhood. Countries might become more protectionist, with trade restrictions for moving around atoms, but there will be more globalization for moving around bytes.
At the Initiative on the Digital Economy, we run a weekly seminar, and we had to switch that to Zoom. To my surprise, it’s working better. More people are participating, the discussion is more egalitarian, and we’re seeing people from around the world. Moving forward, we’re going to be doing the seminars partly digital. For us, and for many companies, the adoption of remote tools isn’t going to switch back. Economists call this hysteresis — the effect persists after the cause has been removed.
As this continues to happen, there’s going to be a big division between nimble and slow companies. There’s also been a very disparate economic impact between knowledge workers (who can work remotely) and blue-collar workers (who cannot). There will be some big winners and losers. The smart companies aren’t just looking at how to survive, but how to come out on the other side and be structured for a more digital economy. They are thinking two steps ahead.
The rise of machine learning is also going to play a role here. We’re in the early stages of a tsunami of machine learning. COVID-19 is forcing managers and workers to figure out if machine learning can keep operations going. We developed the Suitability for Machine Learning rubric explicitly to let people know where the economy was going so they could position themselves for that.
‘We call this the reorganization of work’
Sarah Bana: postdoctoral associate, MIT Initiative on the Digital Economy
The Suitability for Machine Learning rubric takes in the detailed work activities involved in an occupation, applies 21 different questions, and sees how suitable they are for machine learning. Our goal is to examine where the labor force is most and least exposed to machine learning technologies.
Based on our work so far, there are three examples where machine learning adoption, coupled with the impact of COVID-19, is going to affect the work that people do. These are not exhaustive but demonstrate possibilities we might see coming out of COVID-19.
The first is chatbots for customer service. When you go to a bank or a store, you talk to someone in person. But in the new normal, more interactions happen through email, phone, or chat. These conversations are often recorded and saved. Advances in natural language processing can better parse conversations and identify appropriate responses. Our adaptation to the new normal, combined with the new data for these interactions that are now being recorded, will speed up the deployment of chatbots to provide consistent service at all hours of the day, with customer service professionals handling the most challenging problems.
The second is dynamic pricing for accommodation. Consumers are used to online travel sites that offer dynamic pricing, but some large hotels still have sales agents who book weddings, conferences, and other events at set rates. Post-COVID-19, an agent or manager's intuition might not be as good as it once was in determining whether they should book a space today and what price they should book it for or wait for another offer. Accurate predictions of occupancy are an important use of machine learning. Because the industry is being hit so hard, it will be even more important to bring in revenue from supplemental spaces.
The third is visual pattern recognition for inspection. Most inspections are in person, and traveling has virtually stopped — but with the right tools, a lot of inspections can be done from home. One example is the growth of vegetation near power lines. When you interlace machine learning and the Internet of Things, you can take pictures, calculate the rate, distance, and direction that tree limbs are growing, and determine whether they need to be cut down.
In a lot of ways, machine learning will take this from being a reactive process to a predictive one. Plus, a utility company will be able to analyze many more locations. Humans are limited in how far they can drive in a day, or how many times they can safely climb a ladder. That will impact prices, too — if the power line is less likely to fail, then the cost of electricity goes down.
We call this the reorganization of work. We know that virtually no one’s job is 100% suitable for machine learning. If that’s the case, what elements of the job get emphasized and de-emphasized? Jobs like climbing a ladder to inspect a power line are dangerous; it’s something people generally don’t want to do. Now we can move away from the routine components, from climbing the ladder to analyzing vegetation growth patterns.
‘Some areas are better suited to recovering from the pandemic than others’
Andrew McAfee: co-director, MIT Initiative on the Digital Economy; author, “More from Less: The Surprising Story of How We Learned to Prosper Using Fewer Resources – and What Happens Next”Compared to the last recession, the digital toolkit today has gotten even more powerful. The cloud, artificial intelligence, and machine learning are all available now. In “The Second Machine Age,” we showed how companies now have a digital toolkit to trim costs and resources. They no longer have to buy DVDs or print out atlases or invest in servers.
All those decentralized, profit-driven efforts to save money on resources add up. That uncovered a surprising phenomenon, which I discussed in “More From Less:” We learned to decouple GDP and economic growth from taking from the Earth. If you were you weigh the U.S. economy year after year, it used to weigh more, because you needed more resources to build the economy. Now, it weighs less.Technology is helping companies cut material costs. That’s important, because economic contraction focuses companies on their cost structure with laser-like precision. There’s an unfortunate consequence, though, because technology also helps companies reduce labor costs. And as we have seen in past recessions, companies don’t immediately rehire everyone, just like they don’t spend money on steel or other raw materials.
Some areas are better suited to recovering from the pandemic than others. We’ll need just as many kindergarten teachers — we’ll want to educate our children even after the pandemic ends. We’ll need as many home health aides, nurses, and doctors. Where people are doing in-person services that robots can’t do and that we can’t virtualize, those professions will be fine.
The category of hospitality workers is an interesting one. Eventually — and I don’t know how quickly — we will get out of this. We’ll get a vaccine, the lockdowns will end. We’ve already seen some economies end the lockdown. Does that mean that travel and leisure come back to 100%? Or is it 90%, because 10% of us are scared? Or is it worse?
We will get a bit of clarity about which in-person services are valuable to us. One example is business travel. We’re already seeing how much work can be done via Zoom and Slack. Not all business travel needs to happen to the extent that it did. Those industries and companies that center on hospitality for business are going to go down and stay down.
We’re also going to automate a great deal more knowledge work. The role of the customer service representative was already under pressure, and it’s going to be under even more pressure. There’s also a lot of white-collar clerks left in the economy, with jobs that double-check or oversee important transactions because someone’s health is on the line or there are a lot of dollars at stake. There’s a set of transactions we haven’t automated yet because we need a way to make it right.
Cloud technology providers were doing well before the pandemic, and right now they are providing the infrastructure to let things get better. Alphabet [Google’s parent company] probably doesn’t want to get into farming, but it is getting into cloud and machine learning infrastructure for agriculture. There’s a lot of business to be made for helping other businesses run more efficiently, through data collection or task automation.
Finally, technology has provided a silver lining in the pandemic. The extraordinary speed at which we’re learning about the virus, the intense collaboration, the discovery happening around the world — I don’t think it would have gone as quickly 10 or 11 years ago.
‘You expect slight shifts but not gargantuan shifts’
Geoffrey Parker: research fellow and visiting scholar, MIT Initiative on the Digital Economy
Within the World Economic Forum’s Advanced Manufacturing group, we’re working on a white paper discussing the need for a new operating model for manufacturing firms. A lot of these companies have been heavily disrupted. Their supply chains have been long-distance, and they were optimizing cost but not flexibility. During the pandemic, that has been crushed.
That’s why we found ourselves in the situation where we couldn’t redeploy assets such as personal protective equipment and respirators to critical areas of need. Flour was another example. The supply chain was optimized to supply restaurants in bulk, and when the demand for restaurants fell, the supply chain wasn’t easily reconfigured to respond to the exploding demand from households. It wasn’t as flexible as it could have been. You expect slight shifts but not gargantuan shifts, and you’ve got to be able to meet the demand in the form and location where it exists.
This is a broader signal that manufacturers have failed to absorb the digitization and flexibility that was taking place at the factory. There’s a digital thread going through the organization, but many firms didn’t link this end-to-end to their supply chain. The firms that had redesigned and redeployed their manufacturing capacity before the crisis, the firms that weren’t just working with a predefined set of supplies, were much more responsive than the firms that were highly optimized for cost and stable demand. Other firms need to learn from that and embrace it.
Health care is another area we’re interested in. There’s been a call for a long time for electronic health records to be interoperable, but EHR vendors have been pushing back. It’s too bad, because easier sharing of health records has both long-term and short-term benefits. Hopefully this crisis will spur action to resolve it.
In the short term, health systems have been quick to redeploy doctors and nurses — to make them available through telehealth or in pop-up triage centers. However, those services may or may not be connected to other electronic record systems. You have this awkward situation where you don’t have access to the full medical history at the point of care. That’s just ridiculous in 2020.
In the long term, you have a whole system that could operate with higher capabilities if it was open for external parties to build on it. Instead of keeping a system closed and locked down, if you could let third-party developers in, they could do a lot of innovation.
Data aggregation is an important example. Right now, we don’t really have large medical history or demographic datasets to apply artificial intelligence and machine learning to try to understand which COVID-19 treatments are going to be effective. A lot of small-scale trials and experiments are running when we could run a much larger set of observations, but we need a much more flexible data system to be able to respond to that. Right now, the best sources of data are the Johns Hopkins University site or the New York Times site. Those are hand-curated.
There is a danger that firms will snap back and go to what the old normal was, but I don’t think that’s going to happen. It’s implausible that we are going to move fully back to business as usual in manufacturing, health care, or other industries. In the future, every business is going to be asking three key questions: “What is the new normal? What part of the new normal is going to be standard operating procedure? What can we learn from firms that adapted more quickly?”
Social media will be ‘the central nervous system of humanity’
Sinan Aral: incoming director, MIT Initiative on the Digital Economy; professor, Information Technology and Marketing; author, “The Hype Machine: How Social Media Disrupts Our Elections, Our Economy, and Our Health – And How We Must Adapt”
Social media has the potential for tremendous promise but also tremendous peril. Four months before the pandemic hit, [Facebook CEO] Mark Zuckerberg was testifying before Congress. There was the 2016 election scandal, the fake news scandal, the livestreaming of a mass murder in Christchurch, the #DeleteFacebook movement, and so on.
Overnight, the pandemic sent people scurrying off the streets, into their homes, and onto social media. When we were denied this physical contact, social media became an indispensable source of human connection, social support, fundraising, medical information, collaborative art projects, live group chats, and free concerts.
The question is, what is social media’s impact on society — and how do we regulate it, use it, design it, and harness it to create positive outcomes rather than peril? COVID-19 put this question front and center because everyone was forced into remote socialization.
I think we’ll see Facebook, Twitter, WhatsApp, Instagram, and even Slack essentially becoming the central nervous system of humanity. Even the digital Luddites were forced to use social media to connect with people. WhatsApp's automated information service, COVID Connect, is being used by the World Health Organization, and it's the official information service in more than 20 countries. The Facebook and Carnegie Mellon COVID-19 symptom survey is collecting 1 million responses a week, and the behavioral survey we're developing with Facebook and WHO will collect 1 million responses as well.
Social media was able to do a lot of positive things in response to COVID-19. It helped debunk misinformation and the spread of fake news. Usage data allowed us to analyze the effectiveness of government policies on social distancing. User surveys enabled us to collect information from people around the world, analyze it, and make it available to policymakers and other stakeholders. The impact of all this work will steer social away from the peril and toward the promise.