🔮 GenAI at work: The year of scale?

An update on the opportunities and challenges to AI adoption

The world’s first AI software engineer, data leaks and GPT-powered Pulitzer nominees. GenAI has announced itself to the world of work, dazzling and disorienting.

Last year, theory and experimentation dominated the discourse on AI’s economic impacts. 2024 is the year we start to discover if this discourse translates into tangible value.

We previously covered genAI’s impact on the workforce in May 2023, and again in October 2023. These pieces are still relevant, and I invite you to revisit them.

But a lot has happened since October. We’ve seen…

• 50% of large companies launch a genAI pilot. Impressively, 10% already say that they have scaled.

• 55% of workers have used genAI for a workplace task.

• Amidst workplace bans, 55% of workplace genAI users resort to unapproved tools.

In the rest of today’s edition, we’ll share what you need to know about the latest developments, including…

• GenAI’s growing pains — are they real? And how are firms tackling them?

• The guerrilla adoption — Are productivity gains forcing employees to break their organisation’s genAI ban?

• The freelance market early impact — What do shifts in the gig economy tell us about the effects on various professional sectors?

  

  Exponential View via Dall-e 3

---

The year of scale

In 2023, an astonishing 50% of large companies launched a genAI pilot. This is a remarkable feat for a technology that effectively became usable a mere year ago with ChatGPT’s debut. 

Some firms have already seen gains. 

Klarna, a Swedish fintech leader, handles two-thirds of its customer support chats using AI, contributing $40 million in profit from adoption. Martin Elwin, their Director of Engineering, said three-quarters of Klarna’s workforce now embrace genAI in their daily routines.

Software development has also benefited from genAI, with companies using it to automate aspects of unit testing, allowing them to radically scale up the code test coverage. This should result in improved code quality, better maintainability and faster iteration.

IBM has implemented a system that allows 94% of the interactions between its ~275,000 employees and HR to take place without human intervention, helping streamline backroom operations.

These examples underscore the low-hanging fruit of enterprise AI adoption. As time progresses, there will likely emerge radically innovative applications that we have not yet anticipated.

So far, 10% of firms have successfully scaled applications within their operations, suggesting that harnessing AI’s potential at an organisational level is possible, but not without a challenge. While LLMs can automate straightforward decision-making tasks, complex decisions with safety or legal ramifications will remain under human purview. LLMs’ stochastic nature leads to responses which vary widely even for conceptually similar queries. The hallucinations can result in unpredictable risks: for example, the referencing of non-existent software packages in the source code of several big businesses like Alibaba. These risks have prompted 98% of Fortune 1000 CEOs to pause genAI initiatives, according to one survey. The rationale behind this is to allow organisational guidelines and policies to catch up, rather than an explicit lack of faith in the technology.

Movement in the shadows

These and other hurdles have not stifled genAI’s proliferation among employees — 55% of the workforce use genAI at least once a week.