By Enrique Dans

I still remember exactly where I was when Steve Jobs took the stage at the Moscone Center in January 2007. I was watching the live broadcast, and somewhere during that first demonstration, I had the unmistakable feeling that the ground had shifted. Not because the iPhone had invented anything, because it hadn’t, but because it finally made everything that already existed work together. That was what set the smartphone revolution in motion. The rest is history.

By 2007, we already had email, web browsing, cameras, MP3 players, and what the industry optimistically called “feature phones.” I had used several of those supposedly smart phones myself. Most required more patience than intelligence, and handing one to someone without technical knowledge was an act of outright cruelty. What Apple did was absorb all that complexity and deliver something people could actually use. The multitouch interface replaced styluses and nested menus. Integration replaced fragmentation. The iPhone mattered because it stopped asking users to worry about how everything worked, not because it added one more feature.

The great Edsger W. Dijkstra once said that “complexity sells better.” Every so often, however, simplicity wins by such a wide margin that the entire market has to reorganize around it.

Today’s Enterprise AI Looks Suspiciously Like 2006

I see this pattern over and over in technology. The underlying capability exists, sometimes for years, but the market doesn’t really take off until someone manages to hide the cumbersome or complicated parts. That is exactly where enterprise AI stands today.

Any reasonably serious company can now buy access to excellent models, cloud computing, databases, APIs, and agents. Those are no longer the scarce resources. Turning them into something a real organization, rather than a theoretical one, can fully rely on still requires consultants, data scientists, process redesign, governance frameworks, security layers, and ongoing maintenance. The problem isn’t that the models aren’t good enough; the ecosystem around them is still immature.

Nobody buys an iPhone wondering which task scheduler or memory manager it uses, and thank goodness for that. The platform handles all of it. With enterprise AI, the opposite is true: companies are expected to choose the model, vector database, agent system, memory layer, observability tools, and integrations. A business unit leader shouldn’t need to know any of this. The fact that they still do shows that the technology category hasn’t matured yet. There is something faintly absurd about the current situation: the AI industry talks about intelligence as a utility, like electricity or water, then sends highly paid engineers into customers’ offices to make that utility actually work. Truly mature technology scales by hiding complexity, not by sending specialists to every customer.

The “iPhone Moment” Changes Who Can Use the Technology

Before the iPhone, smartphones were largely the province of tech enthusiasts like me, executives, and IT departments. After the iPhone arrived, adoption went mainstream because users no longer needed technical expertise. I still remember giving my mother her first iPhone. I didn’t explain the gestures, menus, or navigation. I simply handed it to her, and within minutes she was doing almost everything she needed.

The equivalent moment for enterprise AI will come when a business unit can deploy all kinds of highly sophisticated AI capabilities without first commissioning a custom architecture or bringing in a team of consultants across the organization.

The Second Half of the Analogy Is the App Store

The iPhone’s real platform effect arrived in 2008, when Apple launched the App Store. I use this case every year with my MBA students because it is one of the clearest examples of a product becoming a platform. There were more than ten million downloads in the first weekend. Developers could build whatever they wanted without reinventing the operating system, the device, or distribution. The platform handled the complicated parts; they supplied the ideas.

That is also the future of enterprise AI: a platform where business capabilities can be activated without building a custom architecture every time, rather than one enormous AI application. Sales, customer service, procurement, compliance, logistics: none of these functions should require a different technical foundation. Once a platform handles identity, permissions, context, memory, governance, and learning, adding a new capability should feel more like installing an app than commissioning another IT project. McKinsey has found that the organizations getting the most value from AI are precisely those redesigning their workflows around it. The model isn’t the transformation. The transformation is what gets built around it.

Frontier Firms and a Counterintuitive Prediction

Microsoft’s Work Trend Index introduced the concept of “frontier firms”: organizations that combine people and agents and reorganize work around outcomes rather than functions. Once intelligence becomes abundant, obsessing over the model starts to look a little like obsessing over a phone’s processor: necessary, yes, but increasingly less decisive.

Here is the counterintuitive part: enterprise AI’s “iPhone moment” will make AI less spectacular, not more. Today, every deployment seems extraordinary because it involves a complex project. Once the technology matures, a sales team will simply turn on one capability, customer service will turn on another, and procurement will add one more. Intelligence will become part of the company’s everyday routine. When enterprise AI finally works properly, it will probably stop looking like AI.

The Limits of the Analogy

I’m not predicting the emergence of a single “Apple of enterprise AI,” complete with a closed ecosystem and a walled garden. That isn’t the point. What matters is abstraction: an immature category forces users to understand the technology, while a mature one lets them focus entirely on what they want to accomplish.

When enterprise AI reaches its “iPhone moment,” the interesting question won’t be which model your company uses. It will be whether your company is ready to use a platform that makes that question irrelevant. Deploying AI should stop feeling like an IT project and start feeling like installing an app.

No, we aren’t there yet. But I, for one, have seen this movie before.