Enterprise AI is still waiting for its 'iPhone moment': hiding complexity, not better models

Writing in Fast Company Middle East, Enrique Dans offers an analogy he already uses regularly in his MBA classes: corporate artificial intelligence is today at a stage equivalent to that of mobile phones before 2007. He recalls Steve Jobs launching the first iPhone at the Moscone Center in January of that year and points out that the device did not invent any new function — email, web browsing, the camera and the MP3 player already existed on so-called 'feature phones'. What Apple did was absorb all that technical complexity and hand the user an interface that finally worked coherently, replacing styluses and nested menus with the multitouch screen. Dans also quotes a line from Edsger W. Dijkstra, 'complexity sells better', to underline that, every so often, it is simplicity that ends up reordering an entire market.

According to the author, that same pattern is now repeating with AI in the corporate environment. Any reasonably serious company can already buy access to good models, cloud computing, databases, APIs and agents: that has stopped being the scarce resource. The problem, he argues, is not the quality of the models but the immaturity of everything surrounding them. Turning that technical capability into something a real organization can use with confidence still requires consultants, data scientists, process redesign, governance frameworks, security layers and ongoing maintenance. Nobody buys an iPhone asking which task scheduler or memory manager it uses; in corporate AI, by contrast, companies are still required to choose a model, a vector database, an agent system, a memory layer and observability tools — something Dans sees as proof that the category has not yet matured. He finds it almost absurd, he says, that an industry that talks about intelligence as if it were a basic utility needs to send hugely expensive engineers into every client's offices to make that utility work; a truly mature technology scales by hiding its complexity, not by deploying specialists case by case.

The article illustrates the change of scale that an 'iPhone moment' represents with a personal anecdote: before the iPhone, smartphones were in the hands of tech enthusiasts, executives and IT departments; afterwards, the product went mainstream because it stopped demanding technical knowledge from the user. Dans recalls handing his mother her first iPhone without explaining gestures or menus, and that within minutes she was already using it fluently. The equivalent in corporate AI, he argues, will arrive when a business unit can deploy sophisticated AI capabilities without first commissioning a bespoke architecture or mobilizing a team of consultants across the organization.

The second part of the analogy is the App Store, launched by Apple in 2008, which recorded more than 10 million downloads in its first weekend. For Dans, that is one of the clearest examples of how a product becomes a platform: developers were able to build applications without having to reinvent the operating system, the device or distribution, because the platform absorbed the hard parts and they only contributed the ideas. That, in his view, is the future of corporate AI: not a single great AI application, but a platform on which areas such as sales, customer service, procurement, regulatory compliance or logistics can switch on business capabilities without building a different technical architecture each time. When a platform handles identity, permissions, context, memory, governance and learning, adding a new capability should look more like installing an app than launching an IT project.

Dans backs this idea with two outside references: a McKinsey finding that the organizations extracting the most value from AI are precisely those that redesign their workflows around it, and the concept of 'frontier firms' from Microsoft's Work Trend Index, which describes organizations that combine people and agents and reorganize around outcomes rather than functions. From this he draws a conclusion he considers counterintuitive: when corporate AI reaches its 'iPhone moment', it will become less spectacular, not more. Today every deployment draws attention because it is a complex project; in a mature phase, a sales unit will simply switch on a capability, customer service will switch on another, procurement yet another, and intelligence will be absorbed into the company's normal routine. When corporate AI really works, he concludes, it will probably stop looking like AI.

The author closes by qualifying the limits of his own analogy: he is not predicting the emergence of 'an Apple of corporate AI' with a closed ecosystem and a walled garden, but rather that 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, he suggests, the interesting question will no longer be which model each company uses, but whether that company is ready to use a platform that makes the choice irrelevant. We have not reached that point yet, Dans concludes, though he says he has seen this movie before.