One of the many concerns that business leaders have around automation is how they’re going to adapt their workforce to new technology. The digital age has brought a rapid shift in the skillsets employees need to go about their jobs effectively, leaving little time for business leaders to come up with strategies for how to reskill and retrain their employees, prevent their companies from drawing disdain from the public, and how to compete with their peers.
The era of AI disruption has begun, and with it, many large companies are making press releases on how they’ve launched a new AI initiative—a chatbot or a management tool, for example. For the most part, however, these companies lack long-term AI strategies that consider the talent and skills their workforces will require in the coming decade and beyond.
In this article, we discuss the concepts of reskilling and retraining employees in the enterprise as AI adoption becomes more inevitable. First, however, we’ll dispell the myth that AI adoption is fast-moving in the largest sectors—It’s not.
1 – AI Adoption is Sluggish
AI adoption has been sluggish across industries. Radical shifts are that would lead to massive layoffs are not happening right now. This is for a few reasons, which we discuss below, and it results in one AI trend we’ve noticed in high-risk industries that we call the “Lens of Incentives:”
The Lens of Incentives
The Lens of Incentives is the idea that companies are not going to say things about artificial intelligence that don’t make them look good in the eyes of their customers and in the eyes of their investors.
For example, the top 100 global banks are talking a lot about what they claim to be doing with chatbots, but they don’t talk about how they’re spending much more on fraud detection because doing so would put customers on alert that their money could be at risk if they keep it at one of these banks.
The Lens of Incentives in large part makes AI adoption look as though it’s happening faster in certain industries. Large companies even try to get up on stage at AI events and talk about all the innovation they’re doing in-house. In reality, a lot of this is just talk; these companies just want to appear as though they’re adopting the newest technologies even if they aren’t.
Pilots Projects Are Not Translating to Real AI Adoption
Most enterprises are not adopting much at all: They’re experimenting, and they’re losing a lot of money doing it. The fact of the matter is that pilots are not translating to real traction.
The reason that AI adoption isn’t taking off across industries is that large enterprises lack a culture of enduring R&D risk. AI involves science. If a business leader were to ask a data scientist how long it might take to get an ROI from an AI project, their answer will likely be, “I don’t know.” That’s the only honest answer. There are myriad factors that could increase the time to ROI:
- “I don’t know because I haven’t cleaned the data yet.”
- “I don’t know because I haven’t seen if algorithms will respond to that data. I don’t know because I haven’t tried all the algorithmic approaches.”
- “I don’t know because I haven’t tested all the different individual features of the data.”
There is a lot of uncertainty with regards to when an AI project will eventually pay for itself, and there is a big cultural resistance in the enterprise to making room for large R&D budgets. It’s part of the danger of AI. But large enterprises ruled by their finance departments are uncomfortable with taking on projects that won’t see an immediate ROI, and when it comes to AI, ROI isn’t even a guarantee.
The maintenance requirements of an AI system are also a consideration. Once a company builds an AI product, it has to have dedicated data science and subject-matter expertise staff that is going to constantly upkeep the system and potentially conduct more and more experiments to see if they can improve it. These maintenance costs are often intimidating.
Executive expectations are another barrier to AI adoption. Most executives believe that AI is something they can get a short-term ROI out of. They often don’t expect that their company will need as many cross-functional team members involved in an AI project in order to make it work.
Curiosity and Education
Another reason AI is sluggish in the enterprise is that there aren’t many use-cases becoming the norm. We’re in the Emergence phase of the AI Zeitgeist. In the Emergence phase, companies are just trying to learn what’s possible with AI in their industries.
They’re not that ready to spend a lot of money, but they want to look like they know what they’re doing when it comes to AI. In actuality, companies are spending money on AI events.
They’re not buying from vendor companies that are trying to solve real business problems because these enterprises aren’t ready to buy. If they do buy, they’re buying in a very uneducated way, and, in general, the products they’re buying are not generating an ROI.
Resistance From Middle Management
The fourth reason AI adoption is sluggish is a resistance to integrate AI tools on the part of middle managers. This resistance may in part stem from a fear of job loss.
Some AI applications may indeed lead to employee role restructuring, retooling, re-skilling, if a pilot went really well. Most of the time, it’s going to be hard to get to the point where a company would even have the ability to lay people off because getting an AI application up and running and delivering value is extremely challenging.
However, even with that small chance, middle management is going to take the heat. The person who runs the call center knows that everybody’s going to be asking, “How could you let them use this technology?” They’re the ones that fear the blowback from hinting at the use of AI, the blowback from people feeling like they’re going to be laid off, whether it’s a true risk or not.
A lot of vendors complain that they can get the C-suite to agree, but when it comes time to implement their product, they get a lot of pushback from the managers that are really worried. These managers are going to have to deal with employees who feel their job security is threatened by AI every day.
Related Emerj Interviews
Jan Kautz, Vice President of Learning and Perception Research at NVIDIA, addresses some of the significant hurdles to AI adoption in our interview from mid-2019 – including issues with executive understanding – which we covered in greater depth in our Critical Capabilities article:
Rudina Seseri of Glasswing Ventures spoke with us about AI adoption challenges from a vendor perspective – which enterprise leaders should understand well being booking meetings with vendors:
2 – AI-Related Reskilling Today
Upskilling IT personnel, teaching them to code in a machine learning language, is becoming a new market. There are many training companies that will train people on learning new coding languages, and they’ll take entire technical teams and try to level them up. That said, it’s unclear how successful these endeavors have been.
AT&T is doing this, investing $220 million a year in internal training programs. Similarly, Walmart partnered with Google to invest $5 million in three organizations that claim to be innovating on programs to retrain and reskill workers for the future.
JPMorgan Chase is taking a slightly different approach, investing $350 in its future workforce. The bank worked with MIT to identify the skills it may need its employees to know if it wants to stay competitive in the future. The money is going to community colleges and other educational institutions to cultivate these skills in current and future students.
But these companies aren’t seeing massive layoffs as people transition to using Python. Their investment in upskilling their employees is clearly a future investment. AI vendors are appealing to these efforts in their marketing endeavors.
3 – Layoffs and AI Job Loss
Nobody Wants to be the Bad Guy
Nobody wants to be the bad guy. No bank wants to be the first bank to lay a lot of people off, and so they’re unlikely to do it right now.
Most vendor companies that sell into the enterprise are going to say that they’re not automating anything, that they’re “augmenting people,” helping them do their job better. The truth is, it’s often going to be a little bit of both.
Growth Without Staff Is More Prevalent and Less Detectable
Another factor to bear in mind is that, in myriad different areas of the enterprise, they are growing without increasing staff. It’s much easier to grow a customer service function without hiring new people because of good automation systems. It’s unlikely that a company will come off as the bad guy for this, and so this will likely be how companies try to go about adopting AI and automation.
Well before any large enterprise lays off a tremendous number of people within a given department, it will likely have tested the efficacy of those AI systems really thoroughly by growing the capacity of that function, growing the speed, growing the efficiency, without hiring any new people. This will go relatively undetected before the company starts shutting down offices and laying people off.
Tipping Points for AI Job Loss
A tipping point for AI job loss is when it becomes too difficult for a business to continue keeping people employed in lieu of automation.
Large companies will say that the market is getting too competitive, particularly that startups are eating market share. They’ll say that in order to stay competitive, they need to lay people off. They’ll likely only do this when it truly is impossible to stay competitive without doing so, waiting so that they don’t come off as the bad guy.
At some point, once one bank lays off 1,000 people in New York, other banks will have an excuse to do the same. They will do so with the excuse that everybody else is doing it, and they will be able to blame the market.
Related Emerj Interviews
Dr. Kevin LaGrandeur is a Professor at the New York Institute of Technology and author of Surviving the Machine Age. He shares his perspective on what kinds of job are or are not likely to be automated in the five years ahead:
Marshall Brain is founder of HowThingsWork.com, and a futurist author and speaker. In this episode he explores the changing roles of man and machine in management and workflows – with poignant examples from some of today’s biggest tech companies:
4 – Considerations for Business Leaders
Look at what’s going on with larger competitors. A brick-and-mortar retailer that has 200 stores should look at the retailers that have 2,000 stores and get a sense of what kind of AI they’re adopting and where it’s actually influencing their workflows. Right now, it may not be influencing much, but it’s important to tune into how that might change.
Business leaders should also brainstorm the impact of AI initiatives on the actual workforce. For a given AI application, ask:
- What kind of talent might we need more or less of?
- What kind of workflows will we need more human effort and less human effort?
- What does that mean for our existing workforces?
What new skills can we teach those folks?
- What new places can we take their skills to still get fruitful use out of them? — What are we going to have to do about that?
Follow-up questions may include:
- What kinds of work might we need more or less of?
- What kinds of skill sets might we need more or less of?
- What should we do with our existing workforce as we adopt this technology?
Before adoption even happens, asking these questions will help big businesses prioritize projects for which they can actually manage employee reskilling well and in which they don’t have to be the bad guy.