Apple’s Siri Struggle: New Report Explains What Went Wrong

A lot has been said about Siri in the past year after Apple failed to deliver on its promise of a supercharged voice assistant. Multiple delays and lawsuits later, we now have details on what went wrong at Cupertino. 

According to Bloomberg’s Mark Gurman, Apple rushed the development of its AI features after OpenAI’s ChatGPT launched in 2022. Things began falling apart when the team tried merging Siri’s old code with the new one.

Sources tell Gurman that Apple had no plans to launch Apple Intelligence until ChatGPT arrived. Apple’s software chief, Craig Federighi, realized the chatbot’s potential a month after its launch when he used it to write code for a personal project.

Soon, Federighi, along with Apple’s then AI head, John Giannandrea, and a few other executives, started meeting with OpenAI, Anthropic, and other AI companies to learn more about these AI models. Federighi also told his team that iOS 18 should infuse as many AI-powered features into an iPhone as possible.

Giannandrea then assembled a team and started working on building large language models (LLMs) for AI features. However, as Apple inched closer to Apple Intelligence’s public reveal in June 2024, internal tests showed that the chatbot lagged significantly behind ChatGPT, with OpenAI’s product delivering 25% better accuracy.

The desperate need for AI forced Apple to look for partners. For stronger user privacy, Giannandrea suggested Google, but Apple announced OpenAI as its first AI partner at WWDC 24. Many AI-powered features were announced at the June event, including the ability to summon ChatGPT for requests Siri can’t fulfill. A delay pushed that feature to December 2024, but many others that were announced are still MIA, resulting in lawsuits and a new chapter in Apple’s ever-growing Siri struggles. 

Some Apple Intelligence features were too buggy when Federighi started testing them on his personal phone before iOS 18.4. Worse, the communication between Apple’s product development and marketing teams was poor. Apple had to pull down TV ads for iPhone 16 features that were nowhere close to being ready.

Internally, Giannandrea has absorbed much of the blame for Siri’s failure. Several employees claim that Giannandrea lacked urgency for generative AI; another Apple executive says Giannandrea believed consumers don’t trust it enough and that AI agents are far away from replacing humans, Bloomberg reports. The exec has blamed overhyped marketing. CEO Tim Cook eventually lost faith in Giannandrea and replaced him with Vision Pro’s head, Mike Rockwell, in March. 

Aside from these issues, Siri’s code was seen to have a major technical flaw. To add AI features, Apple engineers split Siri’s infrastructure in half. When they tried to merge Siri’s new AI features with legacy features, such as setting the alarm, things started falling apart. 

A team in Zurich is now creating a new software architecture built entirely on an LLM-based engine, called LLM Siri. Their aim is to make Siri more conversational and better at processing information. In the EU, Apple will also let users replace Siri with third-party assistants, according to a source.

According to an earlier New York Times report, some of Siri’s announced but unreleased features are expected this fall. Meanwhile, Apple is prepping for a huge iOS revamp scheduled for WWDC 25 next month.