Agentic AI’s intersection with devices like phones, speakers and personal assistants hadn’t been visited much in the Prompt Economy. At least that was the case, up until last week when we heard from an upgraded Alexa+.
Amazon’s newly rebuilt Alexa+ represents a significant shift toward agentic AI, according to Wired. The article reports that Amazon redesigned Alexa around generative AI to move beyond scripted commands and toward autonomous task execution. This rearchitecture required solving complex integration challenges across tens of thousands of devices and services, as well as combining multiple large language models into a coordinated system capable of sustained, natural conversation. Wired notes that Amazon’s internal collaboration across AWS, Devices & Services, and its AGI teams gave Alexa+ the technical foundation to operate as a true digital assistant rather than a command-and-response interface.
Wired also highlights Amazon’s development of Nova Act, a visual-reasoning agent that enables Alexa+ to complete tasks on websites when APIs do not exist. This capability allows the assistant to navigate pages, fill out forms, and execute actions much like a human would, extending Alexa’s reach across the broader internet. Combined with a new multi-model orchestration system built on Amazon Bedrock, Alexa+ can evaluate multiple providers in real time, minimize latency, and carry a request through from intent to fulfillment. The article concludes by describing Amazon’s broader ambition for “ambient assistance,” in which Alexa+ becomes a background system that coordinates services seamlessly across a user’s digital life.
And speaking of speaking into devices, last week was also notable for an Indian Express report that Chinese telecom company ZTE, in partnership with ByteDance, has introduced what may be the world’s first fully agentic AI smartphone, a prototype called the Nubia M153. Unlike current AI assistants that operate within limited app boundaries, this device integrates ByteDance’s Doubao model directly into the operating system, allowing the AI to see the phone’s screen, navigate apps, tap and type, make payments, and execute multi-step tasks autonomously. Demonstrations show the agent selecting the correct app for a job, completing forms, and finalizing bookings without user guidance. The article notes that this level of “full-stack control” over the smartphone represents a major leap beyond Samsung, Apple, and other global competitors, whose AI actions remain permission-gated and restricted to a narrow set of partner integrations.
According to the article, the phone’s capabilities rely on a hybrid architecture: cloud-based semantic reasoning performed by Doubao and on-device execution handled by Nebula-GUI, a 7-billion-parameter model trained by ZTE. This approach enables the agent to complete complex tasks such as booking hotels from a single photo, coordinating robotaxi trips across Baidu Apollo, or negotiating with other automated systems, such as Meituan’s delivery bots. The prototype also demonstrates deep app-level knowledge, OS-level visibility, and multimodal understanding, including identifying objects in the physical world and generating modified images.
The Indian Express concludes that while it is unclear whether the device will become commercial, it signals that the first true agentic smartphones may emerge not from Silicon Valley but from China’s integrated AI and mobile ecosystem.
The Prompt Economy is also an active ecosystem for medical devices. HealthTech reports that agentic AI is emerging as a major new category of technology in healthcare. Unlike traditional generative AI, which relies on user prompts, agentic AI can synthesize information across systems, revisit prior steps, and proactively initiate follow-up actions. Experts cited in the article, including NVIDIA’s Amanda Saunders and UiPath’s Jason Warrelmann, note that while these agents remain forms of narrow AI rather than true general intelligence, they represent a significant shift in how healthcare organizations may manage clinical, administrative, and operational workflows. The article highlights rapid projected growth: Gartner expects one-third of enterprise applications to include agentic AI by 2028, and Market.us estimates the global market will near $200 billion by 2034.
The article outlines several high-value use cases across drug discovery, clinical trials, insurance operations, and patient care. Agentic AI can screen billions of compounds, identify eligible trial participants, generate and submit insurance appeals, recommend treatments, and serve as virtual health assistants that deliver instructions, reminders, and symptom-based escalations. Because these agents operate autonomously, HealthTech emphasizes the critical importance of data governance and oversight, including strict control over which systems agents may access and how protected health information is segmented. The article notes that many major technology vendors, including NVIDIA, Microsoft, IBM, Google, and UiPath, are already embedding agentic capabilities into their platforms, enabling health systems to adopt the technology through existing tools.
Ultimately, the publication frames agentic AI as a transformative but carefully managed step toward more automated, efficient healthcare operations.
Source: https://www.pymnts.com/