I still remember where I was when Steve Jobs walked onto the stage at the Moscone Center in January 2007. I was watching the livestream, and somewhere around the first demo I had the distinct feeling that the ground had shifted. Not because the iPhone invented anything—it didn’t—but because it made everything that already existed finally work together. That’s how he managed to spark the smartphone revolution.
In 2007, we already had email, web browsing, cameras, MP3 players, and what the industry optimistically called “feature phones.” I’d used several of those so-called smartphones before. Most of them required more patience than intelligence, and handing one to a nontechnical person was practically an act of cruelty.
What Apple did was absorb all that complexity and hand you something you could actually use. The multi-touch interface replaced styluses and nested menus. The integration replaced the fragmentation. The iPhone wasn’t important because it added one more function. It was important because it stopped asking the user to care about how any of it worked.
Edsger W. Dijkstra once said that “complexity sells better.” But every so often, simplicity wins by such a margin that the whole market has to reorganize around it.
Corporate AI today looks a lot like 2006
I see this pattern again and again in technology. The underlying capability exists, sometimes for years, but the market doesn’t really take off until somebody manages to hide the ugly parts. That is precisely where corporate AI is right now.
Today, any reasonably serious company can buy access to excellent models, cloud computing, databases, APIs, and agents. That is no longer the scarce resource. But turning all of that into something a real organization—not a theoretical one—can actually trust still requires consultants, data scientists, process redesigns, governance frameworks, security layers, and continuous maintenance.
The problem isn’t that the models aren’t good enough. The problem is that everything surrounding the models is still immature.
No one buys an iPhone asking which scheduler or memory manager it uses (and thank God for that). The platform absorbs all of that. In corporate AI, it’s still the opposite: Companies are expected to choose which model, which vector database, which agent system, which memory layer, which observability tools, which integrations. A line-of-business manager shouldn’t need to know any of that. The fact that they still have to means the category hasn’t grown up yet.
There’s a moment I find quietly absurd about the current state of things: The AI industry talks about intelligence as a utility, and then sends extremely expensive engineers into clients’ offices to make the utility actually work. A truly mature technology scales by hiding complexity. It doesn’t scale by dispatching specialists to each customer.
The “iPhone moment” changes who can use the technology
Before the iPhone, smartphones were fundamentally in the hands of technophiles like myself, executives, and IT departments. After the iPhone, the product became massively adopted because it stopped requiring technical knowledge from the user. I still remember handing my mother her first iPhone. I did not explain gestures, menus, or navigation. I just gave it to her, and within minutes she was doing almost everything she needed.
The equivalent moment in corporate AI will come when a business unit can deploy all types of highly sophisticated AI capabilities, without first having to commission some sort of bespoke architecture and deploy a team of consultants across the organization.
But here comes the second part of the analogy . . . the App Store. The real platform effect of the iPhone arrived in 2008, when Apple launched the App Store. I use this case every year with my MBA students because it’s one of the cleanest examples of how a product becomes a platform. On the first weekend, more than 10 million downloads. Developers could build whatever they wanted without reinventing the operating system, the device, or the distribution. The platform absorbed the hard parts; they supplied the ideas.
That’s the future of corporate AI, too—not a single enormous AI application, but a platform on which business capabilities can be activated without building a bespoke architecture every time. Sales, customer service, procurement, compliance, logistics: None of these should require different technical foundations.
Once a platform handles identity, permissions, context, memory, governance, and learning, adding a new capability should feel more like adding a new app than commissioning a new IT project. McKinsey has found that the organizations extracting most value from AI are precisely those that redesign workflows around it. The model isn’t the transformation. The transformation is what you build around the model.
Frontier firms and a counterintuitive prediction
Microsoft’s Work Trend Index introduced the concept of “frontier firms”—organizations that combine humans and agents and reorganize work around outcomes rather than functions. Once intelligence becomes abundant, obsessing over the model starts to look a bit like obsessing over the processor inside the phone. Necessary, yes. Increasingly decisive, no.
Here’s the counterintuitive part: The iPhone moment in corporate AI will make AI less spectacular, not more. Each deployment today looks remarkable because it is a complex project. In a mature phase, a sales unit will simply activate a capability. Customer service will activate a different one. Procurement another. Intelligence will be absorbed into the normal routine of the company.
When corporate AI finally works properly, it will probably stop looking like AI altogether.
The limits of the analogy
I’m not predicting a single “Apple of corporate AI” with a closed ecosystem and a walled garden. That’s not what matters here. What matters is the concept of abstraction: An immature category forces users to understand the technology; a mature category lets them focus entirely on what they want to achieve.
When corporate AI reaches its iPhone moment, the interesting question won’t be which model your company runs. It will be whether your company is ready to use a platform that makes all of that irrelevant. Deploying AI should stop feeling like an IT project and start feeling like installing an app.
We are not there yet. But I’ve seen this movie before.