Reflection on the Future of Work

Discussing Future of Work at the Stanford earlier this year

A few months ago I left my role as VP Product in a Bay Area startup to deeply research topics around the Future of Work. To kick off the new year, I’m summarizing some thoughts and topics I’ll go into deeper this year.

I am greatly indebted to experts who took the time to share their experiences and information. I’ve spoken to leaders from massive global technology companies, start ups who are breaking new ground in this space, think tanks, research institutes, and non-profit organizations. I am also indebted to my family who have supported me in pursuing a rather ambiguous mission.

Overall, there is much to be done in dealing with the impending problems of work. There will be a mismatch. Think of it as a Venn diagram, with two circles: one circle represents the skills that are in demand and the other represents the skills that are on offer. As technology changes the skills demanded, the amount that the circles overlap becomes smaller, and it’s moving at an increasing rate. It’s hard for people to work out what skills are required when requirements are constantly evolving, and to keep their skills up at the required rate. This is perhaps one of the hardest challenges ahead of us.

I spent time looking at both circles of the Venn diagram, but I’d say it hasn’t been 50:50 and I’ve looked more closely at the circle on the right: the ways that labor fundamentally keeps pace — or keeps chasing — the circle on the left, the skills that are in demand.

1/ Job boards are prevalent.

Job boards match people to jobs. As users started to join the internet, job boards started to takeoff. In 1994 Monster, Netstart (which later became Career Builder) and Career Mosaic launched. Today there are 150 job board companies including Indeed, Career Builder, Monster, Angelist Jobs, Glassdoor Jobs and Hired. To get a sense of traction, the biggest of these:

  • Indeed, connects 11,000 employers to resumes daily based on 180 million unique visitors per month
  • LinkedIn has a network of 500M users.
  • Career Builder powers more than 90% of Fortune 500 company job boards and sees 20.5 million unique visitors per month based off 45M resumes.
  • Monster has 35 million unique visitors per month. These provide a valuable service to lower search costs for both sides of the market.

Job boards focus on matching open jobs to candidates. They don’t touch the problem of creating more job-worthy candidates.

2/ Matching Software & Algorithms

Then there’s Applicant Tracking Systems (ATS), automated systems that screen and filter resumes. The ATS market has over 400 companies. These include offerings from established global enterprise software companies to startups. The software has been used by large corporations for their hiring needs since the late 1990s. But today there are solutions available for any business, all the way to down to small operators who may only fill a few positions a year. Numerous ATS vendors are working on providing human users with the ability to search and prioritize resumes. It remains a hard problem and experimenting with new data pipelines, new machine learning algorithms, and designing user interfaces continues at a rapid pace.

Examples of ATS systems (source: Google)

Machine learning is nascent. Google is still working on cracking the matching problem. Amazon has tried, and failed once already, finding bias and poor predictions. Fear of ML is not helping. But even without social protests against ML, ML is hard in terms of designing the right (unbiased) training data at the front end, and explainability of the results at the back end. (This is such a big topic for both employers and candidates, that I will post specifically on this.)

3/ Skills are not yet well-defined and hard to understand between employers.

The digitization of job descriptions and resumes has been helpful.
A powerful 10X multiplier (perhaps, 50X?) would be the development of a skills taxonomy on which other systems could be built. Skills definition is central to helping break down many of the problems in the work space:

  • figuring out who is qualified for a job — do the skills called for in this job match the skills embodied in this candidate?
  • solving the coaching problem, “What skills do I need to get X job?”
  • solving the job search problem, “If I have these skills, where can apply them?”
  • solving the employer problem, “I am struggling to hire for these skills, what non-traditional places can I go to hire?”

Big companies have been working on skills for a while. LinkedIn has 500,000 skills on its platform. The Workday skills cloud has 55,000 skills (based on 31 million workers, derived from 1 million user-entered skills reconciled to 55,000 unique skills). Degreed, which is a central bank of credentials helps it clients measure and certify their people in 1500+ skills. Employers have a tendency to define the skills they want, but not in cooperation with other employers. Bleeding edge AI companies that specialize in finding the right candidates for the enterprise also work on skills, but frequently define jobs criteria for each enterprise client, based on that client’s needs, which reduces the interoperability of skills. This work is exciting and much-needed, but so far it looks like the overall state of skills definition development and adoption is still nascent.

4/ Education is changing rapidly.

There is incredible availability and accessibility of online skill development. MOOCs have made access to incredible, world-class courses available to anyone with computer access. Training units are becoming bite-sized, in weeks not years, and some are incredibly inexpensive. Some, like Coursera, are free to audit and payment is only needed if you’d like a certificate of completion. Expensive in-person boot camps are emerging to give more support to those who find studying alone difficult, as well as the chance to apply skills in group project settings. Better yet, some courses only require payment once you have found a job (Lamda School). While training platforms initially started offering topics in multiple domains, they have converged to the domains that get the highest demand: IT skills.

There is not a tight link between training and jobs for the most part. The investment decision on the consumer side to undertake training continues to be a high risk, “blurry” endeavor.

The bright point is that examples of training linked to jobs are emerging, providing high-ROI, low-risk training paths for individuals:

  • Amazon collaborates with community colleges in LA
  • Google sponsors courses on Coursera that lead to applying for specific jobs

It is worth thinking about what it means to learn. Learning is the acquisition of knowledge that is later applied in the right situations. It’s not about passing a test and forgetting the content afterwards. We know that engagement drives learning — and often that engagement is pushed to higher levels through experiential training or in-person training. Work being done in the virtual reality world may improve engagement and learning outcomes, and the potential to speed up mastery.

The areas that still need work: Completion rates are low for the online courses. Online learning skews to those already with college degrees. 60% of the adult population lacks computer skills, or is fearful of digital-based learning.

5/ Funding of the re-skilling problem is increasing in the US.

Federal and State funding is increasing. There are also numerous non-profits doing on-the-ground programs helping small populations to find new skills and new work. These experts say these efforts are impactful but hard to scale. One of the difficulties is that the human element of advisors/mentors/counselors is a core part of the recipe for success, and does not scale. I’m interested in learning more about how funders are deciding on which programs to fund.

6/ Advice is sorely missing for workers.

Guidance on what path to take, whether it is a short term decision on jobs to apply for, or longer horizon decisions to switch career path, is hard to come by with any level of quality. Personal career coaches exist, but it’s expensive, and it’s not clear to the client how qualified a coach is in getting a particular client from A to B. Combined with ‘Humans are hard’, the role of human advisors won’t go away. But like any professional services field, the gaps are around:

  • matching people to the right coach
  • working out the right problem definition and scope of service
  • lowering the overall cost of service
  • improving access to those who may not be aware of such services

Even inside large companies “Career Pathing” programs aren’t serving users terribly well: it looks like a subway map, rather than more fine-grained personalized paths.

7/ Humans are hard.

The need for individuals to get help exists across the board. Services that are designed to help individuals — figure out a career path, decide what skills to work on, complete those skills, and apply for new jobs — are rather hard to build. Acquiring users, retaining users, and graduating users from any part of that flow is very, very difficult.

  • Signing up to take action on this kind of problem is not easy. When we’re not actually working, we’d much rather be involved with “entertain-tech” or with leisure activities in the real world.
  • Doing new things requires a growth mindset, and many individuals need help on this before they are ready to move to new things.
  • Completion is hard. Looking at the data from online learning, completion rates of training are low, in the low 10% range. And online learning already skews to those already with degrees.

Since it is hard to build a business around serving individuals, most of the investment from startups is on the enterprise side. Bright spots include consumer-focused companies like The Muse — which offers content and services such as paid career coaching — and UK-based Stay Nimble. Both provide content and services in an especially consumer-friendly way.

8/ New models to get the average person involved are emerging from the gaming world.

Built on the idea of showing what you can do, games are being used to assess what a candidate is like. The form factor of games is making is easier and more exciting for users to engage with different kinds of assessment. This development seems very interesting, especially for soft skills. Engagement may be increasing (time spent) but that must be coupled with the effectiveness of training (mastery of needed skills) that ensure job-worthiness.

Talent Marketplaces, The Gig Economy

I’ll be writing about these in a future post.


Lastly, I don’t want this post to be an advertisement or endorsement of specific companies or startups. However I’m mentioning a few companies that I felt are doing interesting work and worth checking out because they help to understand what is possible, and they’re less well known. It’s not they are the best or the only companies doing this work.

  • Aspiring Minds — automated, AI-powered pre-employment assessments and video interview analytics, global
  • Burning Glass Technologies — data about jobs and skills
  • Learning Machine: skills/credentials on blockchain
  • and O-Net: providing career advice
  • — providing career advice
  • Pymetrics — games based traits identification
  • Knack — games based traits identification
  • HireVue — computer vision for body language, tonality and language (key words)
  • — AI for enterprise ATS

Sunita Parbhu is a seasoned startup product executive with expertise in labor markets. Currently on sabbatical undertaking an independent study on the opportunities for improving labor economics in the field of The Future of Work. Sunita is a proponent of thinking from first principles. She has worked closely with tech leaders such as Tom Siebel (Siebel Systems, C3 IoT) and James Currier (NfX Fund) to identify and create new markets from first principles. Sunita is an economist, a Fulbright Scholar, completed her MBA at Harvard Business School, and held leadership roles in multiple technology startups with exits exceeding $6B.

Start ups, emerging technologies, markets, economics, network effects, behavioral insights; application in software products