Successes and failures in companies – the debate about AI productivity flares up and CISOs learn hard lessons from CrowdStrike

Cover story – Welcome to the debate on AI productivity

The true impact of artificial intelligence on labor productivity is a hot topic. Some claim that artificial intelligence could bring us productivity gains that we have not seen in decades.

I dispute this assumption and have therefore read with interest Cath’s “A new corporate gap to bridge – employers are buying new generation AI to boost productivity, their employees are not so sure”:

In fact, according to a study by freelance platform Upwork, just over three-quarters (77%) believe that the technology simply increases their workload. This is partly because they have to spend time learning how to use the new systems (23%). However, two in five also complain that they have to spend more time reviewing or moderating AI content.

Right – a big reason why I argue that generational AI will not be a collective productivity boost. In most cases, it will require adult supervision of production. That impacts the distribution of labor. Cath continues:

One in five even say they demand more of themselves as a direct result of new technology. And to make matters worse, 47% of employees have no idea how to achieve the productivity gains they want. Two in five believe their employers demand too much, which is why one in three are planning to quit in the next six months.

Second problem: When we succeed in doing more, we are usually asked to do more. In theory, there might be modest productivity gains by taking on new tasks or doing more work faster. But morale doesn’t grow by settling for less. So what should we do differently? One of the key lessons from Cath’s post: put people first, not technology.

Deepika Adusumilli is Chief Data & AI Officer at telecommunications company BT Group. She believes that there are three key phases to successfully implement AI initiatives. The first phase is to identify suitable use cases and develop relevant key performance indicators and metrics to measure success. The second phase involves training and educating employees. Possible approaches to this include gamification and group learning. Equally important is the introduction of change management practices to win people’s hearts and minds and ensure their adoption.

These kinds of policies make generational AI more tied to project metrics. Productivity gains, however, are another matter—and indeed a rarity in recent decades. Aside from a brief spike in total factor productivity in the late ’90s and early 2000s in the U.S. (thanks to the personal computer and, to some extent, the Internet), we have not seen major productivity gains from any technology since then—not even cell phones (phones are an instructive example of how constant connectivity does not necessarily lead to productivity gains in the workplace. It may lead to another, less corporate benefit known as consumer surplus, but that is another issue for economists to argue about).

No modern technology has increased productivity as much as the inventions of electricity and the internal combustion engine in the early 20th century. That’s not a knock against generative AI, which despite its performance shortcomings is a significant engineering achievement that humbles me every time I delve into the math behind it. (I’ll share more research in future articles, for now, read: This expert thinks robots won’t eliminate many jobs. And that’s a problem.)

I wouldn’t be surprised if we ultimately see tangible productivity gains through generative AI. but should that be the point we obsess over? Are there not better metrics than pure task performance to judge the impact of AI? The productivity imperative limits our most interesting AI experiments while circumventing the ethical Moor pit problems it often raises. Granted, squeezing every last ounce of labor out of employees is a bit better than using generative AI to justify ill-conceived staff cuts, but are we really getting anywhere with this kind of framework? I would rather choose ROI for the desired outcomes – and start with Cath’s human-centered premise, before new systems are implemented:

The same rules apply as with traditional technology implementations, including ensuring effective change management. Employers who neglect this do so at their own peril.

Meanwhile, smarter thinking about the impact of AI on humans is extending to regulation. Read Derek’s article “US, EU and UK sign world’s first international treaty on AI to protect human rights, democracy and the rule of law”.

Diginomica selection – my top stories on Diginomica this week

Vendor analysis in Diginomica style. Earnings coverage brought news from some vendors currently on the rise:

A few more recommendations from the providers:

Jon’s lucky bag – We don’t often write about crypto on diginomica, but George set that straight this week with a relevant book review: How Crypto Confidential explains the psychology and technology behind the hype. In quantum technology, the winner takes all, claims EY. But what does that mean? Chris revisits the advances in quantum technology, which I believe are due to security first and foremost: “Isn’t there a more immediate challenge: ensuring quantum-safe systems and planning for the post-quantum world where classical systems may no longer be secure?“

The best of the enterprise web

My Top 7

• The Enterprise Security Customers’ Dilemma: Can You Have It Both Ways? – The hard lessons of enterprise security are back at the top of the list, starting with Constellation’s Larry Dignan. For enterprise customers who want to be less dependent on the CrowdStrikes of this world, this is a tricky maneuver indeed: “On Zscaler’s fourth-quarter conference call, Choudhury said, “Since the CrowdStrike outage, the importance of mission criticality has increased significantly. Our customers want resiliency, but they also want consolidation, but not consolidation that makes them dependent on a single vendor.” In other words, organizations are trying to walk a fine line between consolidating security vendors and building resiliency.“
• How Meta’s CyberSecEval 3 can help combat weaponized LLMs – Louis Columbus dives deeper into the LLM threat problem. See also: Louis’ How Badge’s device-agnostic MFA is revolutionizing identity security.
• Cruise’s robot taxis coming to the Uber app in 2025 – that’s interesting – and not the only item (or car provider) I could have picked. While I believe there are still major safety issues to be resolved (such as dealing with adverse weather conditions), there’s no doubt that self-driving cars are seeing some momentum after several years of worrying incidents that garnered a lot of press attention. I’m not the first to chime in, but the point is that in more urban areas, you may soon be able to make that decision yourself.
• Auto giants are reversing their electric vehicle plans – see Ford, Hyundai, Volvo and others change course – Electric vehicles are suffering setbacks in the US, but as Techradar reports, automakers aren’t really backing down, they’re embracing hybrid.
• Authors sue Claude AI chatbot creator Anthropic for copyright infringement – ​​Just a grumpy reminder that even supposedly “security-conscious”/ethical LLMs still find themselves embroiled in high-profile lawsuits that impact their users’ intellectual property…
• Has AI changed the meaning of transformation – in ERP and beyond? – For this video I said: “AI is said to have a huge impact on application areas like ERP – but is that true in practice? And what about the big picture view of digital transformation? Bring your best and/or most incisive questions for Eric Kimberling of Third Stage Consulting.” The audience obliged and provided smart/incisive comments.“Here is the YouTube replay.

touch

I walk on asphalt fumes here in Corporate Event Paradise (lost) Las Vegas, so I have just a hint for you for next time, and maybe it’s not really a hint, but rather extravagant fan service:

This earned me a first-class retort:

I’ll make up for the smell next time… If you have a #ensw piece that qualifies for hits and misses – for better or for worse – let me know in the comments, Clive (almost always. Most Enterprise Hits and Misses articles are from my curated @jonerpnewsfeed.