- I attended attended the World Economic Forum in Davos, Switzerland.
- I had more than 20 conversations about AI over the span of four days.
- Here are my takeaways on task automation, workforce reskilling, human augmentation, and more.
The annual World Economic Forum just wrapped up in Davos, Switzerland, and talk of artificial intelligence was just about everywhere.
The annual event brings together world leaders, billionaire founders, CEOs, academics, and nonprofit types to discuss the biggest topics of the day.
The official WEF theme this year was rebuilding trust. And while there was plenty of discussion around changes to the global order, increasing geopolitical risk, and it being a huge election year, this was unquestionably the first Davos of a new AI era.
Tech companies dominated the Davos Promenade, with AWS, C3.ai, Infosys, IBM, Builder.ai, Cisco, Qualcomm, Salesforce, and more paying for prominent spaces on the main street. OpenAI CEO Sam Altman spoke at the event.
BI
I had more than 20 conversations about AI over the span of four days. Here are my takeaways:
Copilots everywhere
BlackRock COO Rob Goldstein said the investing giant is rolling out AI copilots to staff, anticipating that AI will be able to produce first drafts of content such as meeting docs or primers. Jason Girzadas, the CEO of Deloitte US, said he’d just gotten trained on Microsoft 365 Copilot. And the consulting firm Oliver Wyman Group created a generative-AI tool called LenAI that some users said saved them eight hours a week.
Tasks versus jobs
Amid lots of concern about how AI will impact the workforce (scroll down for more on that), several of those I spoke to drew a distinction between tasks, which have a narrowly defined process and outcome, and jobs, which are a collection of tasks.
AI won’t generally replace jobs, these people said, but it will absolutely automate tasks.
Take someone who works in recruitment. Maybe their job involves paperwork, identifying and meeting with candidates, and providing strategic counsel to employers and prospective employees. The paperwork can most likely be heavily automated, potentially freeing up that recruiter for more human-to-human interactions.
Big savings
Deb Cupp, the president of Microsoft Americas, said in a panel discussion that the software giant saved $100 million in its customer-service operation by implementing AI. Similarly, Mihir Shukla, the CEO of Automation Anywhere, said the company cut customer-service costs by 40% while seeing improved performance since introducing AI.
Workforce concerns
Of course, whenever there’s talk of cost savings and increased productivity, there’s understandable concern that this is code for job losses.
“There is an equal measure of excitement and wanting to engage with it and nervousness about what it’s going to do for them and their jobs,” Tanuj Kapilashrami, the chief human resources officer at Standard Chartered, said of employees and AI on one panel.
A recent survey found that 51% of global executives, 50% of global CEOs, and 52% of US CEOs said job replacement could be one of the effects of adopting AI in their businesses.
Ana Kreacic, the chief knowledge officer at Oliver Wyman, said that in some cases, productivity was actually declining because workers were paralyzed by their fear of the technology.
Reskilling
That means it will be crucial to find ways to retrain workers for the human tasks that are still necessary in the AI era.
At Automation Anywhere, which, as its name suggests, works on automating work, the customer-service team shrank when AI was implemented. But some of those customer-service agents found new roles at the company, either in customer success or in technical roles, based on their transferable skills.
Shukla said the business world had defined roles increasingly narrowly as companies looked to industrialize their operations and scale. In the age of AI, the workforce will be more fluid, he said.
Becky Frankiewicz, the chief commercial officer at ManpowerGroup, echoed this, saying the age of the static job is over, and employees will have to continually upskill and reskill in the future.
This means employers should look for potential hires who are adaptable and have the ability to learn quickly rather than focus on specific skills that could soon be redundant.
From small to scale
If 2023 was the year of the pilot study, then 2024 will be the year of operationalizing AI, Lareina Yee, a senior partner at McKinsey, told me. Others I spoke to agreed: Many businesses have small-scale AI experiments running, often with promising results so far. The challenge is in taking that pilot and specific-use case and scaling it.
Different models
While “LLM” (large language model) might have been up there with WEF as the acronym of Davos 2024, many of those I spoke to stressed the difficulties with multipurpose models trained on giant datasets like OpenAI’s GPT-4.
In many cases, the most effective AI tools, particularly for corporate use, are those that have narrowly defined purposes and have been trained or fine-tuned on exclusive, bespoke datasets.
For example, McKinsey built an AI tool that’s trained on the consulting firm’s rich archive of PowerPoint presentations, white papers, and Excel spreadsheets and helps consultants get up to speed on specific topics more quickly.
“There will be different classes of models for different contexts,” said Satish H.C., an EVP and cohead of delivery at Infosys.
Cost
Then there’s the cost involved. These AI projects are expensive, and not every company has the financial or human capital to pursue an ambitious AI strategy. Plus, not every AI project will generate a positive return on investment, especially when factoring in not just the cost of the tech itself but the costs of changing workflows and retraining staff.
Augmentation … for now
There are always a few buzzwords that emerge at every Davos, and augmentation was one of those this time around.
The idea is simple: Rather than replacing humans, AI will superpower them, taking on some of the most boring and repetitive tasks and freeing them up for higher-level work.
“As machines get better at being machines, humans can get better at being humans,” said Kapilashram from Standard Chartered.
That certainly seems to be supported by the research so far. But some questioned how long that could last.
Mustafa Suleyman, the cofounder of the AI pioneer DeepMind who now runs the startup Inflection AI, predicted that AI will be so business savvy by 2030 that it may be capable of acting like an entrepreneur, mini project manager, or inventor by manufacturing, marketing, and selling products for profit.
“We are going to have not just those capabilities, but those capabilities widely available for very cheap,” he said. “I think that completely changes the economy.”
At a panel where Kapilashram was speaking, Azeem Azhar, who writes the Exponential View newsletter, noted that when AI-powered game engines emerged, the hybrid of human players plus AI outperformed for about eight years. Now, though, the AI game player wins out.
Dan Vahdat, the founder and CEO of Huma Therapeutics, said that while we might think of empathy as a distinctly human characteristic, research published last year showed that patients found AI chatbots to be more empathetic in their communication than medical professionals.
A few final words
The computer scientist Roy Amara is often credited with first saying that we overestimate the impact of a technology in the short term and underestimate its impact in the long term. AI will likely be no different.
