Apple is working hard to rebuild its digital assistant, Siri, after a rough start with its artificial intelligence (AI) features. The company’s recent attempts to upgrade Siri using AI haven’t gone smoothly, and now Apple is changing direction by rebuilding Siri from scratch using a new approach called ‘LLM Siri’.
Apple’s AI struggles started with delays and doubts
The issues began when Apple rolled out its AI platform, Apple Intelligence, last year. According to a new report by Bloomberg’s Mark Gurman, the rollout did not meet expectations. Apple didn’t jump into the AI race quickly enough. Unlike rivals such as OpenAI and Google, Apple held back from making major investments in AI technology.
Craig Federighi, Apple’s software chief, was hesitant to pour money into AI without knowing the results. AI development often requires powerful and expensive graphics processors (GPUs), which Apple didn’t buy in large enough numbers early on. As a result, they struggled to keep up with other companies in the AI space.
Even Apple’s AI leader, John Giannandrea, didn’t believe people wanted AI chatbots like ChatGPT. He told his team that customers often prefer to turn off such tools. This slow and cautious start meant Apple was already behind when ChatGPT launched in late 2022. At that time, Apple Intelligence wasn’t even being planned.
Siri’s update efforts failed behind the scenes
Apple’s first idea was to improve the old version of Siri by adding generative AI features. But this patchwork solution wasn’t working. One Apple employee told Gurman that fixing the old Siri was like “playing whack-a-mole”—one problem would be solved, only for more to appear.
There were also internal challenges. Giannandrea, who joined Apple in 2018, didn’t fit in with the company’s leadership style. Gurman says he wasn’t aggressive enough in asking for the funding needed to compete in AI. Some workers also said he didn’t push his team hard enough and didn’t view companies like OpenAI or Google as serious threats.
Despite marketing promises of a smarter Siri and Apple Intelligence features that could understand context across apps, Apple had to delay many of these functions. They were announced too soon and weren’t ready in time, worsening the situation.
A new Siri and a new strategy
Now, Apple is starting over. The company focuses on building a new version of Siri using a large language model (LLM), the same kind of AI that powers tools like ChatGPT. Apple’s AI team is developing this new version in Zurich and will build it entirely on this new AI foundation. The goal is to make Siri better understand you, have more natural conversations, and pull together information from different sources.
One of Apple’s ideas is to improve Siri’s knowledge by comparing fake training data with actual language used on your iPhone, such as in your emails. This comparison will happen on your device to keep your data private. Only the synthetic data will be sent back to Apple to train its AI models. Apple also wants to make Siri smarter by letting it gather information from the web, similar to what tools like Perplexity AI do.
There are reports that Apple has approached outside companies for help with AI-powered web search, including partnerships for Safari’s search tool.
However, Giannandrea is no longer leading this effort. As of this spring, he has been removed from Siri and robotics projects. Some executives even discussed preparing him for retirement, but they worry that his team might leave if he goes. Gurman says Giannandrea plans to stay at Apple, though he is no longer central to its AI future.
As Apple moves forward, all eyes are on how well the new Siri will perform—and whether it can finally catch up to the competition.
