Tackling Job Displacement at Scale: My Ideal Solution

Job displacement is not new, but the scale and speed will increase in the coming years – a 2013 study by Frey and Osborne suggested that 47% of workers in America held jobs with a high risk of potential automation.

A few months ago, I wrote “Tackling AI-Driven Job Displacement: A Primer”, which argued that current avenues for people to retrain and find new jobs appeal mainly to those who are already motivated. When changing jobs is a choice, we don’t have a problem. However, when changing jobs is a necessity – which it will be as the speed of job displacement increases – we need the whole population to be motivated and proactive about re-skilling and finding new jobs.

How can we achieve this? For countries, this will be one of the defining challenges of our generation. If we cannot get a large proportion of our population to continuously re-skill to fit emerging jobs, we will face economic slowdown, increasing unemployment and most probably political upheaval.

In this article, I give a big picture of my ideal future, some of the outstanding efforts by tech companies and Governments, and the two big challenges that require collaboration.

My Ideal Future: The Big Picture

The job market today is wrought with inefficiencies which create barriers for job-seekers and employers.

  • For individuals, the job search is intimidating because most of us do not have a good understanding of our skills, and how these map onto other jobs. There is also little incentive to obtain new skills if the link between these skills and future employment seems tenuous. It is a simple cost-benefit analysis.
  • Employers are similarly in the dark. They can pay exorbitantly for skills that seem to be in limited supply when they unaware of people with adjacent skill-sets who could easily fill those roles with some investment in training. According to the Manpower Group, 46% of US employers reported that they had difficulty filling jobs.
  • Education providers can profit greatly from the lack of transparency, when people are willing to pay for credentials that have no bearing on employment outcomes. On the other hand, they may struggle to reach a wider audience because only motivated people are seeking out their services. Adult education providers suffer here.

Imagine a different future with me for a few seconds.

If you are a regular person who wants to stay employed

Imagine a day when making plans for your next job is as much a part of your life as making plans for your next purchase: the barriers to your job search are so low.

  • Just as you have a bank account, you have a skills account which elucidates all the skills you have accumulated through your education and previous jobs. These include self-reported skills, as well as skills that are verified by an authority that has studied the relationship between job descriptions and skills across sectors.
  • Just as Facebook and Google push you advertisements based on your click history, you are consistently notified about emerging jobs with a close skills match to your account.Good morning, we noticed that these jobs, which match your skillset, are trending. Salaries in these jobs are 5% higher than your current salary”.
  • Suppose these jobs are sorted by degree of match to your existing skillset – 90%, 80%, 70%. You click on one and get an assessment of the best next steps – “Hi Karen, looks like you’re an 80% match to this job, and people with your profile have successfully entered this job by [taking this course] [doing a side project in this area] [taking a 1-month internship with X company]. Click to proceed”.
  • You are also notified if your job is at risk, and nudged to take action. “Based on trends, your ‘risk modelling’ skill is increasingly automated by companies. Job listings for this are are decreasing at a rate of 5% a year and this is expected to speed up. Here are some adjacent skills you should pick up ASAP.”

When it’s so easy, won’t more people start to inculcate this as part of their lifestyle?

If you are an employer

Now imagine you are an employer. You need the right talent to serve emerging needs and want to access this talent, fast. Imagine if you not only had access to candidates who applied for their jobs, but candidates who had a >80% match to needed skill-sets, and could be a perfect fit if they underwent prescribed training or certification. You could reach out to these individuals and nudge them towards this, or sponsor training where it makes economic sense.

Rather than firing staff when functions are automated, you may prefer to move them into emerging jobs within the company, as long as the cost is not prohibitive. What if you could assess the degree of match between a potential lay-off with all the emerging jobs in your organization? This could help you make a choice on whether to invest to train the potential lay-off for these roles, or to favour of a new hire.

The information and incentives are better aligned to help you invest in training, rather than to go through rapid cycles of hiring and firing, which could undermine the fabric of your company.

If you are an education or training provider

Now imagine you are an education or training provider. When your courses are not standalone units but part of a pathway that quantifiably increases someone’s chances at securing an emerging job, your user base will likely increase. With more employers investing in education and skills training, your revenue also increases.

If you are a country Government

Now imagine you are a country government who has realized that the traditional model of centralized labour force planning is far too slow for the rate of job displacement. You can breathe easy because there is no longer a need for this outdated method. Individuals are motivated to reskill by good information and clear outcomes. Employers are willing to invest in training that will bring them the skills they need. Training providers fill real gaps and less people are overspending on education that yields little outcomes. The market is finally working – not just for the motivated few, but for your whole population.

Existing Efforts by Tech Companies are Laudable, but Insufficient

What I have laid out is pretty ambitious. It requires:

  • A closed loop between individuals, training providers and employers (graph below). Having a common language to describe skills and jobs lies at the core – if we continue to describe the same skills and jobs in different words, employers do not know the full extent of their candidate pool, job-seekers do not know where they fit, and educational institutions do not know how their programs really help. It’s the wild, wild west.
  • User-centric design that caters for the full range of our population: from the highly-motivated time-rich youth, to time-strapped parents and people who have worked in the same industry for decades. Some may not even use the internet. Employers similarly have to come on board – from the large multinationals to our mom and pop shops.


Several companies have already been working on parts of this puzzle. For example:

Google for Jobs, announced in June this year, leverages its powerful search engine to pull jobs which are most relevant to job searchers. To achieve this, Google’s powerful machine-learning model normalizes jobs and skills which are described very differently by employers and users. For example, if you search for “collections call centre job”, google will be able to flag out all the jobs that you will likely be interested in, even those that include none of your search terms, e.g. “Revenue Management and Care Representative.”

In my opinion, no one is better placed than Google to do this because of their experience with relational models, which group relevant search terms and results. Relational models form the fundamental basis for how Google yields relevant results even when you are not sure how to search for it. I wrote about how Google uses relational models in health search here, which has strong parallels to the job search.

Linkedin is well-positioned to create a closed loop between job-seekers, training providers and employers because all these players are in their platform. Linkedin helps individuals develop their skills portfolios through self-reporting and endorsements. It is also starting to map individual skills and jobs. If you have a premium subscription, you can see how your skills compare to other candidates and the percentile of applicants you fall into based on your Linkedin profile (example below). With the acquisition of Lynda.com, Linkedin is increasingly able to recommend courses to help people up-skill towards their desired job outcomes. To provide greater transparency to job seekers and employers, Linkedin’s Economic Graph team is also mapping the demand and supply of various skills across the world. However, Linkedin’s current business model is expensive for users, which could exclude a large part of the population and limit the data it collects.

Linkedin: initial efforts at mapping skills to jobs

Start-ups are also playing important roles in this ecosystem. Just two examples:

  • Interviewed, an SF-based start-up, helps to close the loop between skills training and jobs. It started out by successfully developing a range of assessments to help employers assess the skills of candidates. It has since developed platforms such as Rightskill, where employers commit to hiring or at least interviewing people who obtain a certain score on hosted course. MooCs like Udacity have also made strides to closely link skills training to job placements – an essential move if they want to motivate a wider population.
  • JobKred, a Singapore-based start-up, uses predictive analytics and data science to connect job seekers to their best opportunities, recommend personalised learning plans, and help employers zero in on their top prospects. They provide a customised user experience in the job search, making the job search less frightening.

Two Big Challenges Ahead

The creativity and experience of these companies will go a long way to achieving my ideal future. However, there are two big challenges ahead that I hope companies, Governments and non-profits will tackle together.

First, aligning the way that everyone describes skills.

Google has made leaps in working out the objective relationship between skills and jobs on the back-end. Their relational model is based on proprietary occupational and skills ontologies that they built in-house.

  • Google’s occupational ontology gives a top-down understanding of 1,100 job families, a task typically undertaken by Governments.
  • Google’s skills ontology maps out 50,000 hard and soft skills and the relationships between these skills.
  • The skills and jobs ontologies are then mapped onto each other to paint an objective picture of the hard and soft skills required in each job, regardless of how they are described.

This is an amazing first step – let’s just take a moment to appreciate it. The fact that they make their relational models available through the Google Cloud Jobs API is even more amazing.

However, I believe these ontologies should not be kept on the back-end, used only when someone conducts an internet search for jobs. Many aspects of the job and education journey take place offline. Constraining the benefits to people who make proactive internet searches may miss out on a huge swathe of potential job-seekers.

Public and non-profit organizations should help with nudging individuals, employers and training institutions towards describing skills and jobs in the same language – perhaps Google can consider working with Linkedin, to combine Google’s job and skills ontologies with Linkedin’s data on self-reported skills, and make these a shared resource. However, the business model must make sense for the companies. Instead of users, Governments, charities or philanthropists should pay as this is a public good.

To reach the widest population, we also need to proactively help individuals populate and manage their skills accounts, just as they manage their bank accounts. UX designers need to work on interfaces that give people a bias to action. I covered some ideas in my “ideal future” section.

Second, bringing users on at scale to enable powerful Artificial Intelligence.

AI will play a huge role in enabling my ideal future: it underpins recommender systems and relational models, just to name a few areas. In turn, these AI-powered systems are enabled by data. Here we face a chicken-and-egg issue: unless we bring on the wider population to use the system, the software will not be able to serve the wider population. To collect data, we need to bring employers, individuals and educational institutions onto this system at scale.

This is an area where technology companies will benefit from the help of public and non-profit institutions. Singapore’s decision to give every citizen $500 for skills training and to create a unified skills and jobs database for individuals is a first step towards bringing a larger segment of the population on board. For segments of the population who are not digitally savvy, we need humans to come alongside and help them get onto the system.

Final thoughts

My ideal future is one where everyone – not just the highly motivated, time-rich or digital savvy – is empowered to continually update their skills and move to emerging jobs, rather than get left behind in the wave of job displacement. Finding your next job must be made as easy as finding the next place you want to eat at.

In all likelihood, this vision will not come through 100%. New safety nets must be put in place for people who simply cannot find new jobs. However, I believe that keeping people in jobs must be our utmost priority: not only are jobs the best safety net, there is also something dignifying about work that I do not believe can be replaced by a handout.

The goal of motivating a population at scale and overcoming inefficiencies in the job market is an area ripe for solutions by technology companies. I am encouraged that many – from conglomerates to start-ups – are working hard to achieve this vision. At the same time, I want see more collaborations that will enable large segments of our population to benefit. I would love to hear your thoughts – what would you like to see in this space?




3 Ways Technology Can Help Keep People in Work (Genevieve Ding)

We’ve heard it before: robots and AI are eating our jobs. New jobs will emerge, but we may not be equipped to do them. I’d written an overview of the issue here.

This week, Genevieve Ding gives an overview of how technology can help KEEP people in work, by mitigating job displacement. Gen speaks from a position of experience: over the past five years, she headed economic strategy in the Ministry of Finance, where her team introduced the SkillsFuture initiative in 2015. She currently works at the National Trade Union Congress (NTUC), where she is interacts with employers and workers to understand the employment landscape across the spectrum: old and young workers, blue collar and white collar jobs. She was previously in the Foreign Service for 4 years, and was posted to Beijing, reporting on all aspects of the Chinese economy.



Photo from ntuc.org.sg 

Technology and Job Displacement: Not a Foregone Conclusion

Auntie Sally is a union member at an electronics factory in Singapore producing wire bonds, the tiny wires in your mobile phone, tablet or step tracker that connect the semiconductor chip to its housing. A toothy lady with boundless energy and a knack for making you feel like her daughter, she has been in this job for 15 years. With no more than a primary school education, she began by operating the machines which make aluminium wire bonds, and has risen through the ranks to become a supervisor in her department. Recently, she was transferred to a new department, where she has to use software she does not understand.  Frustrated and under immense pressure, she feels lost and fears that she may soon be let go. Once fiercely proud, her sense of self tied up so closely to the value she used to provide at work is now deflating. Worse still, her company’s prospects are bleak because the process of wire bonding risks being entirely replaced by advances in materials engineering.

There have been many involved government policy discussions on how rapid technological disruptions and the progress of artificial intelligence will inevitably displace workers like Auntie Sally along the entire value chain, hollowing out the lower and middle classes.  The pace of disruptive technologies makes it ever more difficult to train workers fast enough to transition to new jobs and sectors.

Underlying this is a deep, visceral and very real fear that the diligent, dutiful employee has of losing his job. The US elections and Brexit last year and the ongoing protests against Uber in part reflect this widespread, deep-rooted fear. The sheer potential and associated impact of disruptive technology could trigger the same fears that offshoring has, driving an even deeper divisive cleave between proponents who are able to extract benefit from it, and those who fear they will lose out.

There is plenty of literature on how we can partner technology to improve productivity that make jobs easier, but risk displacing workers, whether by hovercraft technology to reduce the need for painful manual work; collaborative robots to increase productivity, data mining programmes such as Verifi and Margin Matrix that perform time-consuming routine research and drafting so lawyers can focus on more complex and meaningful work; or smart manufacturing systems that use predictive data analytics to increase yield rates and optimise operations in manufacturing.

In a technologically-enabled nation, can technology be used to tackle some of the very problems arising from its development? So that workers are not mere spectators or recipients of help, but active players with a call to action, and who can see the reality of a better, more hopeful life?  Workers will be more invested in the adoption of emerging technology if they can see themselves jointly sharing in progress.

How can technology can play a role in mitigating job displacements and making workers more valuable? I suggest three ways.

1. Helping companies identify their skills needs

As technology evolves, the skills required in the workplace to enable companies to keep ahead are evolving more quickly than ever. The biggest challenge for the labour movement or employment agencies is how keep up with the rapid change in the skills profile of their workforce.

What was somewhat surprising to me when speaking with companies, is that they too are often feeling around in the dark for skills they will need in the future, before suddenly realising that they’re behind — leaving too short a runway to train workers. Data science for instance was a skill which was suddenly hot, prompting companies to scramble to develop expertise. But for some time many didn’t really understand how data science would add real value and what specific skills were required, so job descriptions for data scientists had grandiose expectations looking for “unicorns” that were almost impossible to fill.

Technology has a role to play in helping both government and companies, especially smaller ones, better anticipate the changing skillsets needed to remain competitive.  Data analysis has the potential to trawl through thousands of online job descriptions to fish out skills that are trending in particular sectors, or even identify emerging skills across sectors which may give rise to new niche areas of growth. Once identified, companies are better able to start training workers as soon as possible before their old skills become outmoded.

2. Matching skills, eliminating biases

Just as Tinder and Coffee Meets Bagel help people find their other halves based on users’ preferences and profiles for a better fit, finding good matches between skills and skills in demand is essential to helping workers.

Algorithms, not unlike those that help you find a compatible date, have the potential to match job seekers to jobs based on skills, interests, aspirations, and cultural fit. At the same time, algorithms can help workers identify skills gaps in their resumes based on the skills most in demand or trending in job descriptions, helping them to identify training opportunities.

Digital labour platforms like LinkedIn or CareerBuilder also create more transparent job markets and disrupt previously closed labour markets by increasing workers’ access to a wider variety of jobs and employers’ access to a wider pool of job seekers, reducing the advantage of “old boys clubs”, often driven by wealth and connections.

The playing field is levelled even further by technology platforms that attenuate hiring biases such as paper qualifications and gender, by enabling testing for the specific aptitudes required on the job. Platforms like Codility and GitHub help employers seek out and test for quality of coding and development skills, not certifications.  Catalyst DevWorks’ Catalyst Talent Platform uses machine learning on thousands of variables from hundreds of thousands of individual engineer and developer candidates to identify innate capabilities and predict whether someone will be exceptional talent in the job, whether or not they have a degree or a good resume.

3. Real-time, real-world training

Lifelong learning is much easier said than done. Massive open online courses (MOOCs) have already democratised learning, providing easy access to countless new courses and possibilities. However, much of this learning remains theoretical and does not train or test  “on-the-job”, so it is less useful for industries such as manufacturing.

Even more depressingly, upskilling won’t get you a job. I’ve had the unenviable position of speaking to a electronics engineer Mr. Lee who was retrenched. In tears, he related how he tried to take professional courses in the biomedical sector, with hopes of entering what was then one of Singapore’s growth sectors. Despite his burnished qualifications, all the companies he approached felt that he didn’t have the job experience commensurate with someone else his age in the industry.

Virtual and augmented reality (VR and AR) open up new possibilities of providing “on-site” ,  “hands on”  training for workers and might provide a solution to learning that accelerates job transition and meaningful skills acquisition throughout one’s life.

In manufacturing for instance, AR smart glasses that overlay computer-generated graphics and real-time instructions can improve productivity without prior training. This will shorten the time required for onboarding new workers and help close skills gaps.

Significantly, these upskilling technologies can also help companies “test” out potential employees during the hiring process in a simulated environment, assuring them that the job seeker – even if, like Mr Lee, did not have prior work experience – can perform to standard. Real-time, real-world training with AR will also workers help existing workers learn continuously and at an accelerated speed, increasing organisational learning agility.

Why is this so difficult? The challenge of scale

Using technology to help mitigate the impact of job displacements can only be really effective however if we can adopt them at scale. This can be challenging.  For instance, identifying in-demand skills across sectors or on a national level, or skills matching through data analytics will be most robust if there is open access to large volumes of  job offerings on the demand side. Markets are more transparent the larger the source data.

However, much of this information is fragmented across various platforms and job portals–with a significant proportion of hiring done through personal referrals or headhunters. National job portals where all employers are required to list job openings with job descriptions and skills needed–such as Singapore’s national online Jobs Bank–would go some way to address this. Google for Jobs, which was recently launched, will also contribute to this. 

Technology will also affect various constituents to differentiated degrees.  Eliminating biases through Codility or GitHub for example is limited to skills that are more quantifiable and thus demonstrable on a platform. Less quantifiable skills such as learning agility or strategic thinking may not be as easily evaluated through mediated platforms.

Last, technologies such as VR and AR for training are most impactful if they can both be customised and scaled up. Cost constraints and access to these technologies in the near-term will limit their scalability. Addressing these challenges  in-depth is certainly worth a separate discussion.

Conclusion: Technology as a force for social resilience and collective progress

Challenges notwithstanding, by deliberately harnessing technology in these ways, we are negotiating a new narrative: one that empowers workers and shows them that they too have a stake in our collective progress. Technology no longer divides, but instead buttresses society’s resilience. It provides Auntie Sally a vision of progress that she can once again take pride in contributing to.


Tackling AI-driven job displacement: A Primer

I’m starting a series of posts on solutions to AI-driven job displacement.

I want to answer one big question: in light of AI-driven job displacement, what technology and policy innovations do we need to give people a great sense of hope for what the future holds?

I use the word “hope” because it is the prospect of a better future within reach which will inspire people to continually re-invent themselves when they face setbacks like job loss, rather than to give up. It’s not enough to give them concessions.

Lots of great articles have diagnosed the problem of technology-driven job displacement. A quick summary with links to articles I really enjoyed:

  • In the long run, there will be a net positive in jobs, but in the short run there will be pain: especially among the middle class. The White House published a useful overview on the impact of AI on society last December and Gary Bolles has great resources to understand the hollowing out of the middle class. One very concrete example of middle-class job loss: as shared mobility becomes an increasingly attractive option for city-dwellers, less people will feel the need to own a car. The GMs, Fords, Toyotas and Tier-1 suppliers of the world will scale down their manufacturing business and move into new lines such as mobility services. New jobs will most definitely be created – for example, operators who manage fleets of autonomous vehicles and software engineers whose algorithms will continually improve the cost and energy efficiency of fleets. It’s likely these jobs will hire new people, not those who were displaced.
  • Progress is at stake if we can’t give people the confidence that the new system will work in their favor. We see this in widespread support for anti-trade platforms, even though no economist doubts the overall benefits of trade. America has taken a few steps back, and while some of this may be attributed to the ignorance of one man, it is also the collective anxiety of people speaking through their votes – and this anxiety is not unfounded. This is not a particularly American problem. Brexit is the other most popularly cited example, but you see it everywhere, both in the developed and developing world: Egypt, the Philippines, Singapore: a push back among those who feel that the system no longer works in their favor.

So when it comes to job displacement, it’s not a matter of “us” and “them”. It affects all of us, albeit differently.

Here’s the challenge: how can we make the future full of hope for everyone – so that broad-based support for technological progress will push us all forward?

A Framework to Understand our Target Group

To start off, it’s useful to have a framework to understand the universe of people we need to reach. I offer two dimensions that differentiate people in our target group.


*these are just illustrative examples! Obviously, the reality is a lot more nuanced.

Proactive vs Reactive

On one end, we have people who are “proactive” about keeping themselves relevant to the job market. Geeks like my husband, who completed a dozen Coursera/Udacity courses, taught himself new coding languages, mastered Tableau and ran a successful math blog, all while managing his day job.

On the other end we have people who are “reactive” – they are hardly thinking about the possibility of losing their job; even if they are, they aren’t taking action. This could be the 45-year old who worked a 9-5 job for the past 20 years or a new parent who has no bandwidth to think about anything but the present.

Inherent personality traits are one determinant of where a person falls on this spectrum. Life stage is another – it affects how much time we have to think about alternative plans. The prevailing culture of family and community also shapes what we value such adventure or stability, ambition or work-life balance. Much of this is determined by geography.

Where someone falls on this spectrum is never static. People can shift left and right, and motivation is key.

 New Entrant vs Old Timer

The “new entrant” vs “old timer” distinction is useful in framing solutions. To illustrate this, let’s imagine Uber can fully automate all its fleets by 2025, and will no longer need human drivers.

“Old timers” are existing Uber drivers. The challenge is to give them sufficient fore-warning of impending job loss, and opportunities to pivot into new industries. The difficulty is in encouraging them to take action before it’s too late, and to ensure there are good alternatives.

“New entrants” are those who will be entering the job market in 2025. The challenge is to equip them with skills that will be relevant to the job market in 2025, including skills required to operate Uber’s fleet of autonomous vehicles in localities throughout the world. The difficulty is that no one knows exactly what skills will be needed in 2025, and how many jobs will require these skills.

Here are some topics I’d like to address based on this framework.

First, most solutions today address those who are “proactive” – it’s the easiest group to address because it’s sufficient to make learning resources available; they’ll find them and take them up. How can we motivate or guide those who fall into the “reactive” half of the framework? What can tech companies do? What can Governments and employers do? How do these efforts tie together?

Second, how can tech work with higher education institutions to prepare new entrants (our children today) for the future job market? About five years ago, some predicted the death of higher education institutions as Massive Open Online Courses took off. Now, most of us acknowledge that this is not going to happen at a large scale any time soon. Most people aren’t as confident enough to ditch traditional paths, even if these don’t make sense anymore. How can tech companies help to reform higher education and make it far more responsive to the job market than today? I will draw on my experience working in higher education to address this question.

Third, what social security policies do we need in an era of accelerating job displacement? In the Silicon Valley, the idea of a Universal Basic Income – where everyone receives a basic monthly stipend from the state – has gained much traction as a policy solution to job-displacement. How will a Universal Basic Income impact the different groups in this framework? What are the alternatives or complements to a Universal Basic Income? I will draw on my experience working in and studying social policy to address this.

A summary here: Stay tuned!