Generative AI at the Crossroads: Light Bulb, Dynamo, or Microscope?

More than 80% of companies using generative AI still observe no tangible impact on their financial results, according to McKinsey. Yet field studies document spectacular gains: 14% improvement in customer service productivity, 56% acceleration for developers, and a 50% explosion in AI-related patents since 2018. This apparent contradiction reveals a crucial question for leaders: does generative AI represent a simple temporary improvement or a lasting technological revolution?

A new Brookings Institution study provides a rigorous analytical framework to answer this question. The authors examine whether generative AI possesses the characteristics of two types of revolutionary innovations: general-purpose technologies (GPTs) like electricity, which transform the economy for decades, and inventions of methods of invention (IMIs) like the microscope, which sustainably accelerate the research and development process.

Three criteria to identify a lasting revolution

Diffusion: beyond the hype

Real adoption of generative AI reveals a contrasted landscape. In large companies, 72% declare using it in at least one business function, compared to only 9% for all American companies. This disparity is explained by a crucial nuance: large organizations often employ it for support functions (marketing, HR) while SMEs remain focused on their core production processes.

Even more revealing, 40% of American workers already use generative AI individually, often without their employer's knowledge. In the tech sector, 84% of programmers use it. This "underground" adoption suggests authentic demand that goes beyond the fashion effect.

Complementary innovation: the ecosystem is structuring itself

Unlike an isolated technology, generative AI catalyzes a wave of complementary innovations. User interfaces are rapidly evolving from simple chatbots to autonomous agents capable of complex actions. Robotic integration progresses thanks to multimodal models, while "copilots" integrate directly into existing workflows.

More strategically, generative AI forces organizations to rethink their processes. Leading companies don't just "plug" the technology into their current systems. They restructure their teams, centralize their data governance, and optimize their supply chains around new capabilities.

Continuous improvement: the race isn't over

The third criterion concerns the core technology's capacity for continuous improvement. Here, the signals are exceptionally positive. Since the introduction of the Transformer in 2017, models progress at a sustained pace thanks to three levers: algorithmic optimization (the training cost of a given model halves every eight months), hardware improvement (24% annual decline in computing cost), and dataset enrichment via synthetic and sensory data.

A double revolution: technology AND research method

The analysis reveals that generative AI also possesses the characteristics of an invention of a method of invention. It simultaneously improves four dimensions of R&D:

Observation: reconstruction of imperfect images, completion of incomplete datasets with superior accuracy to traditional methods.

Analysis: ability to identify patterns in complex social and scientific data, opening new research fields.

Communication: automation of time-consuming writing tasks (literature reviews, funding requests, presentations).

Organization: emergence of "digital twins" for R&D, reducing dependence on expensive physical infrastructures.

Your roadmap this week

This dual GPT-IMI nature of generative AI suggests a profound and lasting impact on productivity. But unlike apocalyptic or utopian predictions, the transformation will be gradual and will require massive complementary investments.

Three immediate actions for your organization:

Assess your technological maturity: Map your processes according to the four IMI dimensions. Where could generative AI improve your observation, analysis, communication, or organization of R&D and innovation activities?

Identify your critical complementary innovations: Beyond purchasing AI licenses, what investments in interfaces, training, and organizational restructuring will be necessary to capture value?

Build your transformation narrative: The Brookings study shows that performing companies treat generative AI as an integrated cognitive partner, not as an isolated tool. How will you evolve your corporate culture to embrace this human-machine collaboration?

Generative AI is neither the electric light bulb (temporary improvement) nor the magic wand so hoped for. It resembles more the electric dynamo and microscope combined: a technology that will transform the economy for decades while accelerating the pace of innovation itself. The question is no longer whether it will revolutionize your sector, but how quickly you will adapt to this double revolution.

Going Further

The Projected Impact of Generative AI on Future Productivity Growth

The Brookings study shows that generative AI possesses the characteristics of a lasting revolution, but on what economic scale? A new Penn Wharton Budget Model analysis provides valuable quantitative projections to calibrate your strategic investments. This research evaluates the potential impact on US GDP and identifies the sectors most exposed to this transformation.

The figures reveal a substantial but gradual impact: +1.5% GDP by 2035, +3% by 2055, with a peak contribution to productivity growth around 2032 (0.2 percentage points per year). About 40% of current GDP is potentially exposed to AI automation, but real gains will depend on effective adoption and realized cost savings. Administrative, financial, and IT occupations are most affected, while manual jobs are less so.

The study highlights a crucial limitation of current projections: they don't account for changes in product quality, the emergence of new tasks, or effects on innovation. This methodological caution reminds us that the true benefits of generative AI could far exceed these conservative estimates, reinforcing the argument for proactive adoption.

Innovation Report 2025 – How Top Companies Scale Bold Ideas

If generative AI transforms research methods as Brookings indicates, how do leading companies concretely integrate this technology into their innovation strategies? The new Bain & Company report offers valuable insights based on Fast Company's 50 Most Innovative Companies. This analysis reveals the distinctive practices that separate true innovators from followers in the AI era.

The data contradicts conventional wisdom: performing companies don't necessarily spend more on R&D, but allocate their budgets differently. 94% of studied innovators plan to enter new sectors, and 88% place innovation among their three strategic priorities. AI plays a central role in accelerating processes (20% reduction in time-to-market for 31% of leaders), but doesn't eliminate human creativity.

The key to success lies in a dual approach: one system to optimize the core business (rigid processes, financial KPIs), and another to explore disruptions (agility, rapid learning, long-term potential KPIs). 79% of innovators use separate operational models for disruptive projects, reflecting strategic centralization of innovation resources.

Behind the Curtain: Slow, Hard AI

The Brookings analysis and Wharton projections paint a promising future, but what about ground reality today? An exclusive Axios article based on interviews with Julie Sweet (Accenture CEO) reveals why AI adoption remains slow and complex in large companies, despite media enthusiasm and massive investments.

Accenture's figures are revealing: 85% of leaders plan to increase their AI spending in 2025, but most struggle to reinvent their processes and train their teams. Julie Sweet emphasizes that success doesn't depend on the technology itself, but on the ability to rethink work and empower managers. A historical phenomenon, the "J curve," explains this delay: disruptive technologies first reduce productivity before boosting it.

Major obstacles include resistance to change, the need to simplify obsolete processes, and the risk of job losses for workers unable to adapt. This ground reality confirms Brookings' analysis: generative AI is indeed a general-purpose technology, but its deployment requires massive complementary investments and profound organizational transformation.

This convergence of analyses - theoretical (Brookings), quantitative (Wharton), practical (Bain), and ground-level (Axios) - paints a coherent picture: generative AI will transform the economy, but the road will be long, demanding, and reserved for organizations capable of combining strategic vision with operational excellence.

Thank you for following this exploration of ongoing transformations. To not miss our weekly analyses and receive exclusive insights, subscribe and share this newsletter with your network.

Full transparency: this newsletter was designed by a human (me, Marc!) with help from Claude by Anthropic and Le Chat from Mistral for design and inspiration. The core ideas, composition, and narrative are the product of my three decades of leadership experience. I believe in practicing what I preach: using AI as a collaborator, not as a substitute for human creativity and insight.