Let’s hit some model news.
Apple has launched a new artificial intelligence chatbot named Asa, intended exclusively for its retail employees as an automated digital sales assistant. This rollout comes just weeks before the anticipated launch of the iPhone 17, with the goal of enhancing support for the retail team during the busy sales period. Asa is integrated into Apple’s internal sales training app, SEED, and is currently in beta testing. Reports indicate that employees may use this tool to better understand how iPhones can serve various industries. Notably, while other tech companies focus on customer-facing AI products, Apple’s strategy appears to prioritize internal applications to bolster employee performance. This approach aligns with findings from a recent study by the Massachusetts Institute of Technology, which noted that successful business AI applications often stem from back-office, employee-centric innovations rather than customer-facing solutions.
Google has announced a significant upgrade to its Gemini chatbot, introducing a new AI image model called Gemini 2.5 Flash Image, which allows users to make more precise edits to images through natural language requests. This update aims to enhance user experience and compete with OpenAI’s popular image generation tools, which have seen rapid adoption, with OpenAI reporting over 700 million weekly users. The Gemini image model focuses on maintaining the consistency of details in edited images, such as faces and backgrounds, which has been a challenge for competing tools. Nicole Brichtova, a product lead at Google DeepMind, emphasized that this update improves the quality of edits and enables users to visualize projects more effectively. Google is also implementing safeguards to prevent misuse, such as generating non-consensual imagery, and plans to watermark AI-generated images to help users identify their authenticity.
Microsoft has officially ended its collaboration with OpenAI’s voice models, unveiling its own speech models capable of generating a minute of audio in under a second. The newly developed MAI-Voice-1 by Microsoft utilizes a mixture-of-experts architecture and is trained on 15,000 NVIDIA H100 GPUs, showcasing the company’s ambition to establish independence in artificial intelligence’s critical user interface. With the global voice artificial intelligence market projected to reach $7.63 billion this year and an expected $139 billion by 2033, voice AI funding surged eightfold in 2024 to $2.1 billion. OpenAI’s latest offering processes audio through a single neural network, eliminating traditional handoffs that can lead to loss of tone and context. This technological rivalry is set to define how billions of users will interact with machines in the coming decade.
Why do we care?
Apple rolls out Asa, but not for you—it’s only for their store employees, helping them sell iPhones better. Internal AI can be a safer proving ground—helping employees first reduces risk before exposing customers to half-baked automation.
Google tweaks Gemini with a new image model, chasing OpenAI while promising better edits and watermarks so it’s harder to fake images. Better AI image tools don’t just mean new tech—they change how you produce marketing, proposals, and training content
And Microsoft? They cut loose from OpenAI on voice and drop their own model that spits out a minute of speech in under a second. As voice becomes a standard interface, expect clients to ask whether their helpdesk or customer apps should sound more like people.
What does this mean for us? Each vendor’s sending a message: Apple’s saying “AI for employees first,” Google’s saying “trust us with your content,” and Microsoft’s betting on it’s own models.
For your customers, that means choices—do they need productivity, content, or voice? Your job isn’t to pick the winner; it’s to help clients figure out which of these flavors actually drives outcomes, because AI hype only matters when it solves real problems. Remember AND, not OR. You can combine too, which is your real superpower.