Employees are Hiding AI at Work and Why Airlock AI Was Started

Axios says that 75% of knowledge workers (who isn’t a knowledge worker, exactly?) are using generative AI at work, but most worry about their bosses knowing about it. So says a 2024 Work Trend Index Annual Report from LinkedIn and Microsoft.

According to Axios, there are several remarkable findings in the report. Among them:

• 73% of boomers and 85% of Gen Z use GenAI at work. This is unlike, say, social media platforms like TikTok or Snapchat, which skew, generationally, in one direction while Facebook skews in the other. Even smartphone adoption took several years before the more senior work generation used it productively like their younger counterparts. There is no dramatic divide between “digital generations.” In other words, everybody at work is plugging into AI.
• 66% of leaders said they wouldn’t hire somebody with AI skills – but employees are afraid to tell those leaders that they’re using AI. And since they’re not getting training at work, they’re trying to figure it out on their own. This is a huge disconnect and every business executive should put addressing this on their radar.
• “Power users” have started to “fundamentally reorganize their workdays around AI” because these users save 30 or more minutes a day using AI. 30 minutes a day is 2.5 hours a week. At $48/hour – call it $50 – that’s $250. Scale that to a team of, say, 20 people on a team, and we’re up to $5k a week in costs saved or redirected. That’s like hiring 2 new people.

The AI moment is here, and it’s arriving from the bottom up in workplaces. – LinkedIn CEO Ryan Roslansky

LinkedIn CEO Ryan Roslansky is quoted from Axios, and one of my favorite pieces of advice he offers is to stop equating roles with job titles. Instead, figure out what needs to be done, then figure out what can be done easier, faster, or better with AI and automation. Then, you’ll realize what new skills you require (maybe from those 2 extra hires) to stay competitive.

We’ve Seen This Before And Many Weren’t Ready

Quick story. In a previous career iteration, I designed and led a team that developed digital learning. At that time, in the mid-2000s through early 2010s, a browser plugin called Flash was the status quo for just about anything digital learning. Flash was everywhere – it was originally how YouTube rendered its videos. It was ubiquitous. Part of the “real” internet. It was also slow, a resource hog, and full of stability and security concerns. It was ideal for the days of beige boxes bolted to big furniture: desktop computers. The first real shot across the bow of Flash’s imminent demise was in a famous letter Steve Jobs wrote in Thoughts On Flash on why Flash would not, on purpose, work on iPhone.

In the early 2010s, we had an opportunity to pitch to an iconic brand, and in the requirements for the pitch, which were sensible to the era, they wanted to see us come up with ideas using Flash. So we did. But in my unique position as part supernerd, part tech enthusiast and part strategist, I knew where we were headed. People would soon prefer faster, leaner experiences on smaller mobile devices. So, parallel to building a working prototype for this pitch using Flash, we also built a mobile-first non-Flash version. I had practiced a mighty flourish to perfection when, at the end of the Flash pitch, I thanked them for the opportunity to show them what we could do with Flash and then said, “But that’s not what you really want.” If I remember right, we handed devices to a few in the room who still didn’t have smartphones or tablets to showcase how a mobile-first web-app version would look as I explained with charts and graphs how Flash was on the decline and the future was the native web.

I love telling this story because it makes me seem smart (I was one of many people involved, they were all awesome) and because, to my knowledge, it was the start of the first mobile learning platform, at least in consumer electronics. (We won the business. It was a lot of fun.)

Fast forward a few years to 2016 when I started Not Really Rocket Science. Of the potential clients I’d speak to, maybe 60% or more still didn’t have responsive websites – websites that automatically adjust their experiences from wider, clicking around on a desktop or laptop to narrow, scrolling down on a smartphone. These people usually had outdated websites that they’d paid a lot of money for just five years or so previous that now were a nightmare on mobile, or in a few cases they’d actually built quick, smaller versions of their website – an entirely alternate website to host and maintain – just for mobile. Many decision-makers didn’t understand why what they had wasn’t good enough, while salespeople and marketers were pleading with them to realize that everybody using a smartphone was just up and leaving for a competitor who seemed to know which way the wind was blowing.

Change happens fast. Faster than we’re ready for sometimes. AI is happening so quickly that any reticence will be costly. Complacency might be unsurvivable.

Why I Started Airlock AI

Remembering my job title as supernerd/tech enthusiast/strategist, when ChatGPT came into common consciousness seemingly overnight (it didn’t, we just all jumped onto it together at around the same time) in November 2022, I was fascinated. Probably like you, I could instantly see the possibilities and potential. It also tapped into a deep, energetic curiosity that drives much of my creative work. I wanted to understand how it all works. I applied to a program at MIT to, in essence, understand the source code. How did Large Language Models work, exactly? What’s the difference between generative and diffusion AI? What’s a parlor trick and what’s deeply impactful and important to know?

Our projects involved understanding and reverse engineering the AI used in, for instance, recommendation engines like Netflix or sentiment analysis from Amazon. It was hard and interesting. All along, I was thinking—what can my clients gain from this? What can I take back to them?

The answer was simplicity. Most small businesses don’t need Netflix-scale AI. They don’t necessarily know they need AI at all—they’re frustrated with inefficiency, redundancy, or lack of productivity. The solution isn’t even clear—they’re just familiar with the frustration.

I started Airlock AI to relieve that frustration for businesses and business leaders who don’t have time or interest in figuring it all out but want to be ahead of change, or at least not stuck behind it. So they won’t build their businesses on Flash when everybody’s about to be using smartphones.

Stop Hiding AI At Work

Back to the LinkedIn report and Axios article, here’s my guidance on how to, as Roslansky puts it, “Build for agility instead of stability and invest in skill-building internally.”

• Develop a common vocabulary for AI. The report suggests that some employees spend a lot of time with GenAI like ChatGPT. Some are seeing huge productivity gains. Start by getting everybody on the same page. We do easy, 101-level intro-to-ChatGPT training so the whole team can get aboard through more advanced, customGPT, automation training. But start by getting everybody familiar with the tools available and how to use them. That includes knowing how not to use them. For instance, if you want employees never to use AI in a customer email, make those expectations known. (By the way, I’m talking about services from Airlock here but the point is to find some resource that works for you, some service provider that can help you move from where you are to where you could be.)
• Start small. You don’t need half your business running on AI next week. In fact, I usually suggest starting by fixing stuff. After some education, determine where things fall through the cracks—those redundancies or inefficiencies. From the study, if “power users” are saving 30 or more minutes in productivity a day, what’s creating unproductivity in the first place? We have a service where we “scan” your business for these opportunities.
• Then, innovate. In the process of fixing stuff, you’ll almost always find opportunities. If you just automated this with that, and AI could do this over there, you could have this whole new way of approaching that scenario.

In all this, some kind of goals should exist. Here’s how we’ll know we succeeded. This addresses how many leaders aren’t sure how to quantify the effectiveness of AI, or where to invest. In some ways these questions are stuck in paradigms we’re realizing don’t apply, or at least won’t apply in the same ways. What should that employee saving 2.5 hours each week do instead? If they’re among the early high performers who figured out how to save 2.5 hours each week using AI, should they be doing something new or different? Wait, what characteristics does that person have that might change my ideas on who I should be hiring in the first place? We don’t know all the new questions yet. So we have to ask the questions we do know, and measure what we can against those.

The trick, as ever, is determining what those questions are.