Enterprise AI is waiting for its iPhone moment

Enterprise AI has not yet reached its "iPhone moment" because the technology surrounding AI models remains immature, according to a Fast Company analysis. The piece argues that while companies can already buy access to excellent models, cloud computing, databases, APIs, and agents, turning that into something a real organization can trust still requires consultants, data scientists, process redesigns, governance frameworks, security layers, and continuous maintenance. The author compares the current state to January 2007, when Apple's multi-touch iPhone absorbed existing complexity and replaced fragmentation, and says line-of-business managers should not have to choose which model, vector database, agent system, memory layer, observability tools, or integrations to use.

I still remember where I was when Steve Jobs walked onto the stage at the Moscone Center in January 2007 https://urldefense.proofpoint.com/v2/url?u=https-3A youtu.be VQKMoT-2D6XSg&d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=X5IrC8Ph7Xcr67JA8Xc2wAcSekdqQz9WWCIWouibSRw&e= . 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 https://urldefense.proofpoint.com/v2/url?u=https-3A www.apple.com newsroom 2007 01 09Apple-2DReinvents-2Dthe-2DPhone-2Dwith-2DiPhone &d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=R14UhYsDW0gal0cBrashfte5HO6C7feg SCku4BjFLg&e= 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 https://urldefense.proofpoint.com/v2/url?u=https-3A en.wikiquote.org wiki Edsger-5FW.-5FDijkstra&d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=WpwtCq9G-ae3dTwOGKBWXW9QSpHncq-eXDir6SPvLCQ&e= .” But every so often, simplicity wins by such a margin that the whole market has to reorganize around it. 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 https://www.fastcompany.com/section/artificial-intelligence 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 https://www.fastcompany.com/91528182/ai-enterprise-failing-llms . 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 https://urldefense.proofpoint.com/v2/url?u=https-3A openai.com index openai-2Dlaunches-2Dthe-2Ddeployment-2Dcompany &d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=rroddkXz-PfmaKEibDZAOkSeUNUVOG33SmI4np4yexo&e= . A truly mature technology scales by hiding complexity. It doesn’t scale by dispatching specialists to each customer https://www.fastcompany.com/91544792/why-big-ai-companies-embedding-engineers-customers-what-does-that-mean . 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 https://urldefense.proofpoint.com/v2/url?u=https-3A www.apple.com uk newsroom 2008 07 10iPhone-2D3G-2Don-2DSale-2DTomorrow &d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=bbv8AD1EziPZnYRRNfIC3WzZMR8K7w-sAn3QQz7AMkg&e= . 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 https://urldefense.proofpoint.com/v2/url?u=https-3A www.apple.com newsroom 2008 07 14iPhone-2DApp-2DStore-2DDownloads-2DTop-2D10-2DMillion-2Din-2DFirst-2DWeekend &d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=PDKTlmlHBG-ieCOsRTeCdFaZWdm8ZFyqUCr 24p54Xc&e= . 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 https://urldefense.proofpoint.com/v2/url?u=https-3A www.mckinsey.com capabilities quantumblack our-2Dinsights the-2Dstate-2Dof-2Dai-2Dhow-2Dorganizations-2Dare-2Drewiring-2Dto-2Dcapture-2Dvalue&d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=J6lbyeVXCYbBXLuh0W3gavgq5NbWW26ivzywww7eNlI&e= . The model isn’t the transformation. The transformation is what you build around the model. Microsoft’s Work Trend Index https://urldefense.proofpoint.com/v2/url?u=https-3A www.microsoft.com en-2Dus worklab work-2Dtrend-2Dindex 2025-2Dthe-2Dyear-2Dthe-2Dfrontier-2Dfirm-2Dis-2Dborn&d=DwMFaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A CdpgnVfiiMM&r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE&m=twNS1Jn06V DZsWHO2g3g7e0Ev6mcHmFKVRgW0yqAqXwD7oDQLZbw1vsDp861pnI&s=REUjkd2bGfpFrptAOp0jBetNo4kgrXf4s7uGGTVQGYc&e= 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 https://www.fastcompany.com/91536400/when-enterprise-ai-finally-works-it-wont-look-like-ai . 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.