.jpeg)
Every transformative technology seems to follow a familiar pattern: a slow start, a sharp takeoff, and eventually a plateau. In business, we often visualize this as an S-curve of adoption.
Early on, only a few pioneers use the new tech; progress is gradual as kinks are ironed out. Then comes the steep climb—suddenly the technology goes mainstream, and adoption skyrockets. Finally, growth tapers off as the market saturates or the innovation reaches its limits. If you chart it over time, you literally get an S-shaped curve rising from near-zero to widespread use. This pattern has repeated from the telephone to the smartphone, and it’s how most game-changing innovations spread through society.
For example, think about radio or television—in the early years, only a few had them; then there was a boom as they became affordable and essential, and finally virtually everyone had one. The Internet followed a similar S-curve: it was a researchers’ toy in the ’80s, a niche in the early ’90s, but by the 2000s it became ubiquitous and eventually hit a saturation point where being online is a given. The S-curve captures a fundamental truth: adoption of new technology isn’t linear. It starts out painstakingly slow as people experiment and figure out what to do with it, then hits a tipping point where it seems to be everywhere all at once, and finally slows again when most of those who will adopt it have done so.
Understanding S-curves matters for business leaders because it helps set realistic expectations. If you’re introducing a disruptive tool in your company, the initial phase might be underwhelming. That’s normal. It doesn’t mean the tech won’t ultimately deliver—it means you’re likely still in the lower, flat part of the S. Patience and persistence (and some smart change management) are needed to reach the boom phase. As a CEO, I remind my team that early adoption can be deceptively sluggish. We have to endure that early crawl to get to the sprint. And as I’ll argue in this post, Generative AI (GenAI) is following this script—albeit with its own twists.
Some technologies are more than just tools—they’re platforms for countless innovations. Economists call these General Purpose Technologies (GPTs), meaning they are broadly applicable and radically transform economies and life. Classic examples include the steam engine, electricity, the automobile, computers, and the internet. These weren’t one-trick ponies; they sparked waves of complementary inventions and changes across almost every industry. GenAI—AI that can generate content and perform a wide array of tasks—is shaping up to be a new GPT, potentially as impactful as those earlier ones.
Here’s the key insight from history: even once a breakthrough technology exists, its benefits aren’t felt everywhere overnight. In fact, there’s often a lengthy lag between the capability and the impact. I like to say capability is not diffusion—just because we have the technology doesn’t mean it’s instantly deployed at scale or delivering productivity gains across the board. We saw this with electricity: inventors like Edison and Tesla gave us electric light and power in the late 19th century, but factories didn’t see major efficiency gains until the 1920s, after managers reengineered factory layouts and processes to leverage electric motors instead of steam power (it took decades to redesign workflows and build new infrastructure) . Early electricity adoption was slow and clunky; only after a period of adaptation did productivity soar. The pattern repeated with computers: the semiconductor and PC were invented by mid-century, but by the 1980s economists puzzled over the “productivity paradox”—you could “see computers everywhere except in the productivity statistics” (to quote Nobel laureate Robert Solow). It wasn’t until the 1990s, after companies restructured and learned how to harness IT, that output and efficiency really jumped. These examples underscore a crucial point: revolutionary tech often requires complementary changes—new skills, processes, business models—before its potential is fully realized. The innovation itself might arrive in a flash, but its economy-wide impact unfolds over years in an S-curve fashion.
GenAI in 2025 sits in a similar spot. The capabilities have leapt ahead—today’s AI models can converse, write code, generate imagery, and more—but most organizations are still in the early stages of figuring out how to use them effectively. As a CEO deeply involved in AI, I’ve observed tremendous hype around tools like ChatGPT, but also the inevitable early missteps and learning curve. And that’s okay. History tells me that we should be excited about what GenAI will do, but also realistic about the timeline. Just as steam power, electricity, and the internet didn’t transform the world in their first year or two, AI’s most profound impacts will take time to diffuse. (In a previous post I even argued that if true AGI—AI with human-level intelligence—were achieved tomorrow, it still wouldn’t instantly upend the economy, because adoption lags invention.) GenAI is a transformative new capability; now we face the work of spreading it and adapting to it—the hard journey from capability to broad diffusion.
When I look at GenAI’s trajectory, I no longer treat OpenAI’s five levels and the cascading S-curves of adoption as two different stories. OpenAI’s five-level ladder to AGI captures the supply side—what the tech can do. The familiar S-curve explains the demand side—how fast enterprises absorb it. Put the two together and you get a single compass for the next decade of transformation.
They’re simply two lenses on the same roadmap: each new capability level ignites its own S-curve of enterprise uptake—rapid rise, plateau, baton-pass to the next wave:
The first wave is now background noise. Consumer adoption is basically done, and inside most companies chat-first tools—ChatGPT, Copilot, Claude—sit quietly in customer-support queues, marketing copy workflows, and dev environments. They turn blank pages into first drafts, translate jargon, and knock 5-15 percent off the busywork every quarter.
We’re still ironing out prompt hygiene and data-security guardrails, but the productivity lift is real and compounding.
2025 is the pilot year for true reasoning engines. Contract-review bots flag clause risks that junior associates miss; AI pair-programmers debug multi-service stacks in minutes; medical summarizers pull differential diagnoses straight from unstructured notes.
Accuracy isn’t perfect yet, but it’s crossing the 90-percent trust barrier that moves a tool from “interesting demo” to “default teammate.” Most firms you and I talk to are either spinning up their first sandbox or expanding a successful proof-of-concept.
This one is no longer theoretical—it’s operational. A growing list of mid-market firms now run goal-driven agents that renew software subscriptions, triage support tickets overnight, and compile the ops dashboard before the human team signs in. They loop for hours or days without a prompt; they also force new conversations about audit logs, kill switches, and who’s legally on the hook when an AI makes a bad call.
Governance has moved from the innovation lab to the board agenda.
Next on deck are models that don’t just solve problems— they originate ideas.
Think drug-discovery agents spitting out novel molecules, supply-chain optimizers sketching entirely new network topologies, or strategy copilots surfacing go-to-market angles no human proposed.
Early pilots are brewing inside pharma and materials-science labs, but broad deployment waits on models that can generalize creativity across domains and on workflows that can vet machine-born ideas at speed. Expect first commercial wins late this decade.
The endgame is an enterprise where most decision-making, operations, and even strategy execution are AI-run. It will start as skunk-works subsidiaries or green-field startups that live on cloud credit and continuous fine-tuning. Once a few of those out-perform their human-heavy peers, the rest of the market will feel like dial-up competing with broadband.
That horizon is late-2020s at the earliest, but everything we do in the first four waves sets the runway: data quality, oversight discipline, and a culture willing to let algorithms take the wheel when they prove they can drive.
History keeps reminding me that capability is not the same as impact. From electricity to cloud computing, the pattern is constant: a breakthrough appears, hype skyrockets, and only years later do the productivity curves bend upward. Gen AI is no exception. Wherever you sit on the five-level ladder—whether you’re perfecting chatbots or piloting autonomous agents—the real lift only arrives after you re-wire data pipelines, redesign processes, and shepherd people through the change curve. Ignore that groundwork and the tech will stall, no matter how dazzling the demo. citeturn0file0
Complicating matters, these waves don’t queue up politely. Conversational tools are now background noise, reasoning copilots are spreading fast, and agents are already grinding through overnight workloads—all at once. So budget cycles, hiring plans, and risk models have to assume overlapping roll-outs: at any given moment you’re stabilizing one wave, scaling the next, and prototyping the one after that. Treat it like a relay race and you’ll be forever dropping the baton. citeturn0file0
That’s why I describe the road to AGI not as a single tsunami but as five overlapping swells. In 2025 we’ve largely tamed the first swell, we’re riding the second, and the third is foaming around our ankles. The winners will be the companies that learn to layer those swells—stacking capability, governance, and fresh value propositions at every crest—while everyone else is still scanning the horizon for the next big wave.
As a CEO who lives and breathes this stuff, I speak with a mix of excitement and urgency. AGI will change everything, but not all at once—it will come in waves. That means right now is the time to position ourselves to ride those waves rather than be drowned by them. Companies that treat AI as a passing fad or defer action until it’s “proven” will wake up to find their competitors several curves ahead on the adoption journey. Don’t let the current early-stage hiccups fool you into thinking AI isn’t ready for prime time. Instead, recognize we are in the early innings of a massive long-term change. The winners of the AGI era will be those who learn faster and adapt through each phase.
In practical terms, every organization should be asking: What’s our GenAI gameplan? How are we using today’s AI (Wave 1) to deliver value and build competency? What experiments are we running to be ready for more autonomous AI (wave 2 and 3)? Are we tracking advancements in the field so we’re ready to pounce when, say, reliable AI reasoning becomes available or when a trusted AI agent platform emerges? This isn’t hype; it’s strategic foresight. Just as savvy leaders in the 1990s positioned their companies for the internet age (even while others said “maybe we’ll wait and see if the internet matters”), today’s leaders must position for the coming AGI age.
I’ll be candid: planning for an AGI world is challenging when the technology is still evolving. But planning doesn’t mean predicting exact timelines; it means building flexibility and capability. It means investing in talent who understand AI, upgrading your data infrastructure, creating an internal culture that welcomes automation and change, and perhaps most importantly, reimagining your value proposition in a world where AI is abundant.
Ask yourself, “If AI can do X and Y (where X and Y are core tasks in your business), what will we uniquely contribute? How will we leverage AI to offer something new or better?” Companies that have good answers to those questions will thrive. Those that don’t…well, history hasn’t been kind to businesses that ignored electricity, or computers, or the internet. AGI will be no different—except maybe even bigger in its effects.
The bottom line is to start riding the wave now, and keep paddling.
We have a once-in-a-generation (perhaps once-in-a-century) opportunity to reinvent how work gets done. It’s not happening in one fell swoop, but it is happening faster than previous revolutions. As the waves of GenAI adoption cascade through the economy, there will be shakeups and surprises. Some jobs will change, new ones will emerge, entire markets will evolve.
Our task as leaders is to guide our organizations through it—staying agile, seizing the advantages of each new capability, and mitigating the risks. If we do this thoughtfully, we can surf these S-curves and come out on top of each successive wave. I believe the next 5–10 years will separate the companies that merely react from those that proactively shape their destiny in the AI era. Let’s make sure we’re in the latter camp.
The AGI world is coming, wave by wave, and now is the time to prepare to ride those waves to success.
