The AI rollout is here - and it's messy | FT Working It

Key Concepts

AI Investment Wave: Unprecedented levels of financial investment in Artificial Intelligence.
AI Adoption Gap: The discrepancy between the availability of AI technology and its full integration into various economic sectors.
Generative AI (GenAI): A type of AI capable of creating new content, such as text, images, or code.
AI Strategy: A comprehensive plan for how a business will leverage AI.
Productivity Gains: Improvements in efficiency and output resulting from the use of AI.
Upskilling/Reskilling: Training employees to acquire new skills or adapt existing ones for the AI era.
AI Agents: AI systems that can perform tasks autonomously, often viewed as digital co-workers.
Shadow Use Cases: Unofficial or personal use of AI tools by employees outside of formal corporate initiatives.
AI Literacy: The ability of individuals to understand, use, and critically evaluate AI technologies.
Prompt Engineering: The skill of crafting effective instructions (prompts) for AI models to achieve desired outputs.
AI Bubble: A period of rapid growth and investment in AI, potentially followed by a market correction or "burst."

State of AI Rollout in Industry

The current investment wave in Artificial Intelligence is described as unprecedented, with hundreds of billions of dollars being spent on workplace automation. Despite this significant investment, AI adoption is not yet widespread across all sectors of the economy. A key statistic highlighted is that only 1% of CEOs have a fully formed AI strategy, indicating a significant gap between investment and strategic implementation. The video explores whether businesses will see a return on this substantial investment.

Early Stages and Adoption Challenges

The discussion emphasizes that AI is still in its very early days, with new models emerging every six months, causing constant shifts in the marketplace. While the technology promises significant productivity gains, only about 10% of companies are actively integrating AI into their core processes. This integration is expected to take years.

Investment vs. Realized Gains

A staggering amount of investment has been made in AI, accounting for 40% of US GDP growth this year. Globally, over 75% of businesses are using generative AI in at least one function. However, a study by MIT Media Lab found that 95% of GenAI pilots in the workplace failed, pointing to a significant disconnect between investment and successful implementation.

Two Speeds of Adoption

Companies are adopting AI at two distinct speeds:

Tech Companies: These are far ahead, viewing AI agents as co-workers.
Other Companies: These are still grappling with the concept of AI adoption, often struggling to get employees to use basic tools like ChatGPT or Claude, and are not yet seeing productivity gains.

Upskilling the Workforce for AI

The concept of upskilling the workforce for AI is a major topic, but its practical implementation is unclear. There is uncertainty about what the ideal worker's skills will be in three to five years as AI becomes more pervasive. Discussions revolve around whether the ideal worker will be a deep specialist or a generalist who can effectively communicate and validate AI outputs. This uncertainty necessitates more experimentation, which can be uncomfortable for leaders accustomed to avoiding failure.

Analysis of Corporate AI Usage

An analysis of S&P 500 companies revealed a discrepancy between public statements and actual filings regarding AI usage. CEOs often spoke of AI's transformative potential and productivity gains in earnings reports, using terms like "Cambrian explosion of innovation." However, regulatory filings, which are more risk-averse, offered little concrete evidence of AI implementation. In these filings, the risks associated with AI often outweighed the benefits.

Abstract AI Use Cases

While the S&P 500 index has seen growth, much of it is attributed to seven major tech companies. Other companies have not necessarily grown significantly despite claiming AI use. AI usage is often described abstractly, with mentions of productivity without specific examples. Coca-Cola, for instance, lauded its use of generative AI in earnings reports but could only cite its use in creating a Christmas ad in its filings.

The Role of AI Upskilling Platforms

The growth of AI has fueled a boom for consultancies and learning platforms focused on AI upskilling. Euan Blair, CEO of Multiverse, an AI upskilling platform, highlights that companies are not hesitant but rather are facing the challenge of turning potential AI gains into actual realized gains. This is where the training gap becomes critical. Many companies are using AI tools at a basic level, akin to using a smartphone only for calls and texts, missing out on their full capabilities.

Tangible AI Gains in Specific Teams

While macro-level productivity gains are not yet widely apparent, specific teams are seeing tangible benefits:

Accounts Teams: Processing invoices 50% faster with half the errors.
Software Engineering Teams: Increasing code shipping speed by 75% in some cases.

The lack of widespread macro-level gains is attributed to the training and capability gap. Unlike previous software, AI's inherent capabilities require fundamental changes in how people work, necessitating extensive training. The stakes are higher due to the significant investment, leading to a realization that success will depend on an AI-enabled workforce rather than just AI investment.

The Desire to Do More Faster

There's a pervasive feeling of being behind the curve with AI, driving a desire to implement it more rapidly. While money is flowing into tech companies, consultants, and AI firms, adopting companies are not always seeing the promised financial gains. However, it's crucial to remember that these technologies are still in their early deployment stages.

Challenges in AI Adoption and Usage

The adoption of AI tools is astronomical, with ChatGPT's growth outpacing the internet's initial rise. However, a gulf exists between work-related and personal usage, leading to "shadow use cases" where employees use AI tools they prefer outside of official corporate initiatives. This often stems from a lack of communication between leadership and staff regarding needs and desired tools.

Workplace Specifics and Risks

Workplaces have unique considerations, including sensitive information and the critical need for accuracy. AI models can make factual mistakes, which can be embarrassing or catastrophic for organizations. Therefore, every organization must consider how these tools apply to them and what employees need to know about their usage.

Readiness for Digital Transformation

A significant challenge for businesses is their lack of readiness for digital transformation. Effective AI use requires good structured data, robust cyber defenses, and, most importantly, AI-literate staff.

Google's Approach to AI Training

Amanda Brophy, Director of Grow with Google, emphasizes the importance of making AI work for individuals within their specific roles. Translating AI capabilities into immediate, personal applications is key to overcoming skepticism and realizing benefits. Examples include marketers using AI for social media captions and customer service agents using it for polite responses to upset customers.

The "And" of Technology and Training

Brophy stresses that both technology and training are essential for AI rollout; it's an "and," not an "or." Simply deploying technology is insufficient. Google's AI Essentials course teaches people how to use AI, prompt effectively, and ensure reliable usage, fostering daily practice and upskilling.

Daily Practice and Intrinsic Interest

AI requires daily practice to become a habit, unlike simply learning about it. An intrinsic interest in AI is needed to see its value in professional and personal life. Employers must provide accessible training for employees to consume this information.

Effective Prompting as a Key Skill

Effective prompting is critical for obtaining desired AI outputs. This involves clearly defining the audience, goals, context, and providing reference materials. Journalists are highlighted as excellent prompters due to their ability to ask the right questions and understand audience needs.

Cisco's Internal AI Usage

Sarah Walker, Cisco's UK and Ireland CEO, describes a broad spectrum of AI usage within the tech company. AI agents are integrated into product development, such as the Webex platform, enhancing efficiency. Employees also use various AI platforms, with adoption levels varying.

Leading by Example and Adoption

Walker emphasizes that leaders must lead by example to drive AI adoption. She advocates for a "pro workforce and pro AI" stance, clarifying that AI is about enhancing efficiency through automation, not replacing roles. She notes that human nature naturally seeks more efficient ways of doing things, and AI is simply the latest manifestation of this.

Common Mistakes and Experimentation

A common mistake is assuming adoption will follow the availability of AI applications. Encouraging employees to experiment with AI tools is the quickest way to learn their potential benefits and limitations. AI should be used for specific, beneficial use cases, not as a universal problem-solver.

The Future of AI and Potential Bubble Burst

The current optimism about AI's future is based on the assumption that it will lead to amazing things. However, if this proves to be a bubble that bursts, only AI use cases that demonstrably benefit employees and bring tangible advantages will survive.

Conclusion: A Path Forward for AI in Workplaces

There is no one-size-fits-all solution for AI rollout challenges. Businesses need input from staff, and staff require support and training from leaders to realize AI's promised financial and productivity gains. The current stage of GenAI adoption is compared to the mid-1990s internet rollout, suggesting a period of potential boom and bust, disruption, and hopefully, excitement in the workplace.