Main content

Has AI really caused a massive retraining need? Then we need a retraining roadmap

Jon Reed Profile picture for user jreed September 20, 2019
It's Friday afternoon, and caffeine is running low. It's time for a fresh critique of the AI skills gap. My foil? A fresh report from IBM, predicting a retraining need for 120 million workers. Grab a cup and join me.


I've been writing about thriving amidst robots since diginomica was in the early days of our enterprise media experiment subversion beat.

The strident proclamation of IT skills shortages is primarily a circular conversation. Attention is gained, and nothing happens.

Companies continue to under-invest in training (a mistake) and insist there is a limited applicant pool of qualified candidates (one part BS, two parts failure of imagination).

The potential impact of AI and automation on jobs remains a fierce debate. However, I have seen a shift the last five years from the alarmist tone of huge swaths of jobs will be gone soon (typical in 2014) to a less alarmist, but still sensational More Robots Mean 120 Million Workers Need to be Retrained. This lightweight Bloomberg headline is based on a meatier (free) report from IBM, The enterprise guide to closing the skills gap, which I recommend.

I don't mind the urgent tone. Even if some of the claims are impossible to verify (120 million), the changes are profound enough that enterprises - and job seekers - should heed a skills wake-up call.

Unfortunately, the "AI and jobs" debate is still stuck, and it's stuck in a most unfortunate place: what to do next. It's a lot easier to sound a shrill alarm than to provide a practical skills roadmap for individuals (and companies) to follow. IBM attempts to do this in their report, but in a very generalized way.

Robots and job impact - what we know so far

Here's what all these robots-and-jobs studies have correctly figured out:

  • Repetitive work across industries is the most threatened. Any jobs where humans are going through the motions and acting like robots, from fast food to bank tellers to grocery baggers to blackjack dealers, are threatened by machines.
  • The AI-and-jobs optimists believe that the amount of new jobs, including "higher value" customer-facing roles, will blunt the impact of jobs lost. Whether or not that's true doesn't matter for our purposes. For now, there is middle ground. If chatbots and self-service can automate the bulk of a call center, for example, then the remaining call center jobs are sure to go to those individuals with the social sophistication to nurture high value customers, or to the subject matter experts who can resolve escalated problems.

We also know this:

Most of the skills in demand will be the so-called "soft skills" - all the skills machines will struggle to duplicate. Often misunderstood or trivialized, the "soft skills" needed go far beyond the ability to communicate effectively at the water cooler. It's about being a process expert, and more:

  • creative marketing ideas and exceptional content
  • making sense of data and choosing the proper action (analytics)
  • the savvy to solve complex problems and get teams to gell (emotional intelligence)
  • ethics and policy-making skills, including expertise is assessing human/machine bias

Of course, those skills still need the "teeth" of industry expertise to generate job offers, but you get the idea. There is also a debate on the technical skills needed to thrive amidst the machines, and what technical tasks can be automated. I'd argue for a few:

  • creative software design/UX (including coding expertise)
  • advanced mathematics and statistics (data science and algorithmic know-how)

The robots and jobs debate leaves companies hanging

Here's where we are now falling short:

  1. We haven't figure out how to combine these skills in one education. Even the skills on the same bullet points above are not easy to combine in one person (e.g. coding and design). Our entire education system is misaligned with what is coming. Liberal arts grads might have advanced creative/critical skills, but have they been exposed at all to the ethics of artificial intelligence?
  2. We have defined the "soft skills" needed, but we haven't articulated how human talent will work amongst, with, and for machines. Yes, we've heard assurances that "digital natives" glued to their smart phones will thrive in an automated workplace. But I don't believe smart phone savvy equates to the machine savvy humans will need within specialized settings like shop floors, and IoT-powered retail. The fusion of advanced tech and soft skills has not been well explained.
  3. We issue excited calls for "retraining," but we have not defined how the retraining needed for emerging roles will happen.

Alas, defining the needed skills is the easier part. Getting those skills to stick is the tough row. New skills can be acquired in a few ways:

  1. On the job, or in continuous training, including online classes -  this is the easiest type of training to foster in today's workplace, though it requires follow-through.
  2. Short-term immersion - taking a week or two off to learn a new skill. This could work well for a new programming language, design thinking, or perhaps some diversity/inclusion training.
  3. Long-term immersion - some skills are flat-out difficult to learn in the first two formats. Example: developing your personal and professional ethics. I'm highly skeptical anyone can develop an ethical framework without a deeper inquiry of some kind. A one week course in ethics seems close to a joke. Interestingly, the same long-term immersion can be important to advanced math and stats, topics which can be difficult to tackle in an online format. For more on that, see Overcoming the AI, ML and data science skills gap - hashing it out with Vijay Vijayasankar of IBM.

We need a skills acquisition plan - enter the IBM report

So, for each necessary skill, we need a clear idea of how it can be fostered, how long it takes to develop, and the environment in which it should be cultivated. I turned to the IBM skills report for answers. First, we rehash the problem: most companies aren't ready. IBM's Institute for Business Value team, which conducted the survey, notes that less than half of the CEOs surveyed said they had the resources needed to close the skills gap brought on by these new technologies.

I'm a bit skeptical on that one. If you asked those same CEOs, "If you had to choose between your business going bankrupt and reskilling your employees to lead your industry, could you find the money?" I'll bet 99 percent would say they could find the money. So it's really back to priorities and a failure of imagination on how a skills investment avoids worst case scenarios - and angles for the best ones.

But it's not just about the resources. It's the lack of a coherent skills plan. IBM's press release for the report adds:

Organizations are facing mounting concerns over the widening skills gap and tightened labor markets with the potential to impact their futures as well as worldwide economies. Yet while executives recognize severity of the problem, half of those surveyed admit that they do not have any skills development strategies in place to address their largest gaps.

Cnet's piece highlights the mounting training problem:

IBM says companies should be able to close the skills gap needed for the "era of AI," but that this won't necessarily be easy. The company said global research shows the time it takes to close a skills gap through employee training has grown by more than 10 times in the last four years.

IBM makes useful training distinctions, similar to my "immersion" views - for both soft skills and advanced tech skills. As Bloomberg wrote:

Some skills take longer to develop because they are either more behavioral in nature like teamwork and communication or highly technical, such as data science capabilities.

"Reskilling for technical skills is typically driven by structured education with a defined objective with a clear start and end," Amy Wright, IBM managing director for talent, wrote in an email. "Building behavioral skills takes more time and is more complex."

Depending on the skill, just-in-time training approach can fall short:

Behavioral skills, such as the ability to work well on a team, communication, creativity, and empathy are best developed through experience rather than structured learning programs like a webinar.

Bloomberg also noted the rise in behavioral training priorities versus technical:

When employers say they're facing a skills shortage, the first thing that comes to mind is coding experience or another advanced technical skill set. Yet, today, employers are calling for more emphasis on soft skills like communication skills, ethics and creativity rather than technical, a switch over the last few years, the survey notes. Behavioral skills are now seen as critical from digital and technical capabilities.

I think that's a false distinction, as it's the fusing of the two that will prove most beneficial.

Retraining is personal

No one called out IBM on the need for more examples on how this "retraining" is done - especially with more complex behavioral skills. Consider my hypothetical example of three call center employees:

  • One call center employee is very polite and helpful, but has no subject matter expertise. They can walk through a service call script, but could easily be replaced by a bot as long as the documentation powering the bot is decent. Possible solution: this call center rep could be retrained into advanced customer success metrics and focused on top shelf accounts, or immersed in a deeper technical training to provide level two or three human support. However,  there isn't an immediate job for them to step into just because they are pleasant, and have what I call "light soft skills."
  • Another call center employee is very sophisticated and knowledgeable, skilled at handling difficult issues, and enjoys resolving complicated problems for customers. They get bogged down in the routine calls and would happily give up the routine calls and administrivia and join a customer success team. Possible solution: this person likely needs very little retraining, except more analytics tools as they go forward. As soon as satisfactory automation is in place to clear their desk of routine FAQ calls, they are good to go.
  • A third call center employee enjoys enforcing policy and inflexibly explaining the reasons why late payment fees are never refunded. However, when confronted with possible job loss, they express interest in retraining. However, they have no technical or soft skills foundation.

Each of these profiles has a different skills roadmap. Some retraining is more feasible than others. Example: training the first profile into a more technical resolution role seems doable, if they are motivated. But retraining the third employee is going to be a project. If they are an outlier, perhaps they are simply relieved of their duties. Not very nice, but this is the wake-up-call world we are living in. However: if there are plenty of folks like this third profile inside the company, now you're talking about a deep culture change that goes beyond a training program.

The wrap - IBM cites retraining use cases, but we need more

By now, you'd think we'd have examples of corporate exceptionalism around this. I went looking in the IBM report. IBM cited AT&T. Not a company high on my personal list of innovative companies, but here's what they shared:

In 2013, AT&T initiated a massive retraining effort after discovering that nearly half of its 250,000 employees lacked the skills needed to keep the company competitive. Core to their strategy? Transparency. AT&T started a dialogue with the workforce about the importance of skills and skills relevancy and now provides a robust portfolio of programs and tools for employees to continually gain new skills.

What I could not figure out from this report is: how is AT&T approaching "soft skills development", and addressing know-how such as the ethics of AI and data privacy, a big deal in their industry? I would be interested to learn more on how AT&T is tackling the entire skills problem, and how they are addressing the amount of immersion needed for each type of skill. Not surprisingly, IBM thinks AI can help solve the problem AI is creating. As per Cnet:

One way companies can keep up, according to IBM: Use AI to discover what skills are already available throughout their organization and share that info with employees to drive a culture of "continuous learning."

That seems useful, though not magical pixie dust. I did like this:

AT&T employees have access to a career intelligence tool that helps them make self-directed, informed career decisions by analyzing hiring trends within the company. For example, employees interested in a U.S.-based network services job could determine that in 2015, AT&T offered nearly twice as many of these positions as it had in 2012. They could also see that information technology roles trended down by more than 200 jobs during the same period. This tool also provides links to options for developing their skills developed through a partnership between AT&T, Udacity, and Georgia Tech.

Other companies cited by IBM using a similar "career intelligence" approach included the eclectic mix of Ernst & Young, Banfield Pet Hospital, and IBM itself.

Surprisingly, I liked IBM's emphasis on personalization. Ordinarily I treat "personalization" with the same toxic wariness I do towards a wasp or a mosquito. But the need for personalized skills roadmaps is clear - even if the problem of doing that at scale is still a work in progress.

I was also encouraged to learn that the executives surveyed are thinking about talent in a less insular way. IBM says the tactics executives are using include: 

  • Acquire talent from outside the organization
  • Move talent across business units and divisions
  • Reskill employees based on business priorities
  • Leverage visa programs to source international talent
  • Leverage apprenticeship/internship programs to train talent
  • Leverage new and emerging educational programs/platforms to enhance employee skills
  • Apply analytics to analyze and predict skill supply and demand
  • Implement skill recognition initiatives to recognize and track skills progression
  • Leverage talent through ecosystem partners

Lots of "leveraging" going on - a word that needs to be purged from modern business, aside from its finance origins. But still - a list that shows companies are thinking beyond an outdated approach to hiring.

Overall, this report is a solid attempt at advancing the skills conversation.Though IBM fell short of the detailed use cases needed, the emphasis on a transparent, company-wide approach is welcome.

I won't call for an end to sensational "skills shortage" headlines - that's a useless howl in the wind. Here's where I think we need to take this conversation next:

  1. Continue a dialogue on remaking a modern education that combines disparate skill sets, from "soft skills" to advanced math to creative design (see my comment below).
  2. Define and detail examples of companies that are taking a company-wide approach to modern skills sourcing and training - with an eye towards lessons learned. If you know of a good one, contact me on LinkedIn.

Updated, 10am UK time Saturday, with a number of additional resource links and section headings.

This article is part of my diginomica series on robots, the future of work, and differentiating skills.

A grey colored placeholder image