Generating Value with Generative AI

Generative AI has captured the world's attention, sparking both excitement and uncertainty reminiscent of previous technological upheavals. Like the steam engine or the internet before it, GenAI represents more than just an incremental improvement; it stands as a nascent General Purpose Technology (GPT). Such technologies possess the power to fundamentally reshape industries, economies and societies, unlocking immense value creation. However, history teaches us that the true potential and most impactful applications of GPTs are rarely clear from the outset; realizing their promise requires significant exploration, experimentation, and a willingness to navigate the unpredictable path of innovation.

Just under three years since GenAI captured the world's attention with the release of ChatGPT, companies across sectors are already experiencing productivity gains. Carefully designed experiments–among coders, and in call centres, consulting firms and consumer goods companies–all point to consistent worker productivity gains from access to GenAI. However, scaling these early productivity gains across an enterprise is proving difficult. A recent MIT report quantifies this challenge, noting that only 5% of custom enterprise AI tools successfully reach production. As a consequence of this high rate of pilot failure, a full 95% of organizations report seeing no measurable P&L impact from their broader AI investments.

Yet amid this challenging landscape, some companies are breaking through the pilot-to-production barrier by taking a fundamentally different approach. Rather than treating AI as a standalone technology initiative, they're integrating it deeply into their business operations while investing heavily in their workforce. Ikea exemplifies this approach, transforming its customer support model by using AI to streamline routine inquiries. In 2021, the company introduced "Billie," an AI chatbot that by 2023 was handling 47% of customer queries, freeing up human agents to address more complex issues. Crucially, Ikea paired this automation with a significant investment in its workforce, retraining 8,500 call center workers whose jobs were augmented by Billie. These employees have been reskilled as interior design advisors, offering personalized, fee-based design consultations. This initiative not only improves efficiency but also creates a new revenue stream, demonstrating AI's potential to enhance both customer service and employee roles.

Crucially, Generative AI continues to evolve as model developers achieve algorithmic innovations and scale up model training efforts. The performance trajectory has been steep, as perhaps best described by one Hacker News commenter: “The most straightforward way to measure the pace of AI progress is by attaching a speedometer to the goalposts.” Capabilities are consistently increasing in just about any domain one might think to measure, be it software engineering, conversational abilities, knowledge of obscure hummingbird anatomy, or something in between. New and innovative uses are being developed as well. Efficiency is increasing too: the cost per unit of performance is decreasing, the most notable example of this being Deepseek R1 debuting in January at 20x less expensive than OpenAI’s o1 model. Notably, the revolutionary capabilities that a mere 2.5 years ago required computing infrastructure beyond the reach of any ordinary individual now fits on a cheap cell phone: GPT 3.5 (the original ChatGPT model) has, within a mere 2.5 years, been surpassed by open-weights models two orders of magnitude smaller that can work on your phone, no cloud necessary.

All of this translates into meaningful capacity to augment and build upon human performance. Complicating strategic decisions further is the sheer pace of AI progress: the cost-per-unit-performance that defines a project’s feasibility in May is likely to be surpassed–significantly so–by November, demanding foresight that extends beyond rollout. Research by METR, for instance, shows that the time horizon for an AI agent–which is the time that humans typically take to complete tasks that AI models can complete with a 50% success rate–has been consistently doubling approximately every six months.

An October 2024 survey found that only 11% of employees feel “very prepared” to work with AI and related digital technologies in their role (Gallup 2024). Notably, this is a decrease from 2023, when 17% of respondents identified as very prepared. The emerging picture is one of stratification, where some companies are already “winning”: while only 15% of U.S. employees strongly agree that their organization has communicated a clear AI strategy, of those who do feel that their companies have a clearly communicated strategy, 87% believe that AI will have an “extremely positive impact” on productivity and efficiency. Given the pace of improvement in GenAI models, the possibilities are continuously expanding. Building out your knowledge base and internal AI infrastructure is an investment in organizational dexterity and will ensure that your company is positioned to generate and capture the value made available by GenAI.

Schneider Electric was an early and decisive actor when assessing possibilities for AI, having established a Chief AI Officer position and an AI department (“AI hub”) in 2021, well before the widespread public awareness of tools like ChatGPT. These acts of foresight, which included the creation of a Responsible AI working group tasked with creating a Trust and Ethics charter, positioned it to rapidly integrate generative AI capabilities. This strategic groundwork enabled Schneider to move beyond traditional predictive AI and explore how generative models could create new value while maintaining the integrity of existing operations.

Internally, Schneider has integrated AI into its customer care centers, streamlining the initial triage stages of a customer ticket and providing relevant information to the human agent, boosting efficiency. They also employ internal conversational assistants which help make organizational knowledge accessible and digestible across departments.

Generative AI is also used to enhance existing services. One way Schneider is enhancing existing services is its integration of generative AI into industrial automation software, specifically its Programmable Logic Controllers (PLC) product offerings. This allows engineers to articulate desired outcomes in natural language, upon which the AI can generate suggested code snippets and automate documentation. Schneider’s PLC product offerings provide automation software to customers, and this increased functionality improves usability for Schneider’s customers. The introduction of Generative AI to this product line fundamentally changed how industrial control systems are developed and maintained.

Besides Generative AI, Schneider has also integrated Predictive AI into its work on physical manufacturing facilities, creating digital twins for manufacturing lines, calculating best layouts and workflows, automating maintenance, and substantially boosting efficiency and productivity.

Managing the integration of AI is not only a matter of layering new technologies on top of old ones. Alongside identifying AI-borne opportunities, the role of an AI team is to identify and eliminate the accompanying risks: Rambach cites hallucinations, data loss, and cybersecurity as the three greatest threats. Generative AI tools are dynamic, complicated technologies which, as with new technologies of the past, require troubleshooting, monitoring, and reflection. Schneider Electric shows that AI can be both safely and most effectively integrated into your organization when there are dedicated leadership roles with visibility into different departments, and when there are interconnected mechanisms of oversight and information sharing.

Realizing the potential of Generative AI requires far more than passively appending chatbots to existing workers and workflows; fundamental changes to processes and roles will be necessary. The path forward demands a deliberately ambidextrous strategy—sustaining core operations while simultaneously embracing rapid, systematic experimentation. This exploration must be deeply collaborative, marrying technical prowess with operational insight, and geared towards creative implementations tightly aligned with business strategy, not just superficial fixes. Such a proactive pursuit of innovation must advance hand-in-hand with responsibility: proactively mitigating risks, anticipating consequences, and thoughtfully managing the human impact of automation to ensure progress yields truly beneficial and sustainable value.