Apple may be the next firm to incorporate a ChatGPT-like chatbot into its Support app, likely as a method to automatically manage cases outlined in its documentation and reserve only the more complex ones for human agents. Overall, this appears to be a positive direction.
## A ‘Support Assistant’ possibly powered by an external model
The Support app already features a live chat option that connects users to real people, but response times can become somewhat unpredictable based on the time of day or following busy periods like Christmas and product launches.
As stated by MacRumors, Apple is working on an AI-driven “Support Assistant” feature for its Support app. According to strings discovered within the Support app, the assistant “utilizes generative models,” which is broad enough to suggest that it likely won’t be employing Apple’s own models, at least not at this time.
## Introducing RAG
If Apple opts not to use its own models, it could take advantage of the same B2B infrastructure already provided by organizations like IBM, NVIDIA, and Salesforce, all of which have developed robust enterprise tools utilizing a method known as Retrieval-Augmented Generation (RAG).
The fundamental concept of RAG is straightforward: instead of depending only on the tokens for which a language model was trained, you also supply it with precise, real-time information, such as the full scope of Apple’s support documentation in all languages, enabling it to generate an answer to a query.
Even more advantageous is that RAG is designed so the system smartly identifies the relevant documents, or even specific sections of documents, needed to address each unique prompt, constructing its response based on that select information rather than needing to sift through everything with every inquiry.
Remarkably, the report also notes that users may have the ability to upload images or documents. This could enable the AI to examine a picture of a damaged screen or interpret an error message from a screenshot, features that would align naturally with a RAG-enabled backend.
This type of task could also seamlessly operate on Apple’s own infrastructure, even if it utilizes an external model provider, ensuring that privacy remains intact.
## 9to5Mac’s perspective
Despite the expected “ugh, AI” immediate reactions, this is a scenario where generative AI is entirely logical.
Assuming Apple genuinely pursues the RAG approach, which it should, RAG continues to be one of the most effective and field-tested methods for creating support chatbots. In this way, they could adeptly manage straightforward cases such as warranty details and battery policies, allowing real human agents to more promptly address the more intricate issues.
Naturally, as with most of Apple’s AI-centered initiatives, there’s no timeline for when (or if) this will be implemented. However, given how commonplace such features have become over the past few years, and the plethora of high-profile corporations offering this precise solution as a plug-and-play option, it might not be long before this one is launched.
