Gustavo’s Corner: AI News for CFOs - #22
Refined AI News for CFOs
AI is moving quickly from answering questions to performing real work. This edition highlights a clear shift toward more autonomous agents, local AI processing, automated research, public AI infrastructure, and even robotics inside data centers. From ChatGPT Work signing into websites and completing tasks to Anthropic automating parts of AI research, the common theme is that AI is becoming more capable of acting independently across both digital and physical workflows.
ChatGPT Work is moving closer to becoming a true digital operator. Plus and Pro users can now let it log into websites and complete account-based tasks using credentials pulled from a connected third-party password manager. OpenAI says usernames and passwords are never seen or stored, although users may still need to approve security checks such as two-factor authentication and can clear saved browser sessions later. OpenAI is also expanding scheduling: free users can now run up to three active tasks, while paid users can trigger automations when something changes in Gmail, GitHub, or Slack.
Perplexity and Nvidia are bringing more AI work back to the desktop with Portable Computer, a version of Perplexity’s Computer agent that runs entirely on a user’s own machine. Files stay local, and work completed on the device does not consume credits. The system can asks permission before sending any step to more than 15 cloud models. The message is clear: powerful AI agents no longer need to live entirely in the cloud.
Apple is pushing deeper into local AI with new Mac Mini and Mac Studio models built for increasingly demanding workloads. The Mac Mini now offers the M6, Apple’s first 2-nanometer chip manufactured by TSMC, alongside an M5 Pro option. Apple says the M5 Pro Mac Mini can process LLM prompts 8.5 times faster than older Pro models, while linked M5 Ultra Mac Studios can pool memory to run trillion-parameter frontier models together. The catch is price: the Mac Mini now starts at $899, up from $599 before this summer’s memory-driven increase.
ChatGPT Work is also expanding what “automation” means by allowing the system to sign into frequently used websites through a password-autofill setup designed to keep credentials private. OpenAI says ChatGPT Work can book appointments, complete applications, and handle essentially any browser-based process that requires a login.
Anthropic is showing how AI could begin improving AI with far less human involvement. In new research led by Chen Yueh-Han, automated AI systems improved model performance across all 10 alignment benchmarks tested without reducing overall performance. The system works much like a human researcher: it searches the literature, proposes an approach, trains the model for 30 minutes, keeps what works, discards what does not, and repeats the process. Anthropic estimates the automated researcher costs about $4 per hour versus roughly $150 for a human researcher.
South Korea’s science ministry is turning national AI policy into public infrastructure. Three consortia led by SK Telecom, KT, and Kakao have been selected to build chatbots and public-service agents that every citizen will be able to use free of charge and without usage limits. The government is providing a combined 512 Nvidia B200 GPUs this year and plans to cover part of the operating costs from 2027. At least half of each service’s model usage must run on certified Korean models.
Meta is now testing automation inside the data centers that power its own AI ambitions. Robots from Watney Robotics, Kinova, and ABB are being evaluated for jobs such as swapping networking cables, rebooting servers, moving equipment racks, and inspecting hardware. Human supervision is still required because tangled cables, battery limitations, physical obstacles, and visual inspections remain difficult for today’s machines. One worker estimated that a reliable cable-swapping robot could eventually perform up to 80% of certain jobs, although this was not an official Meta forecast.
Why this news is important to CFOs and their teams:
These developments point to a new phase of AI adoption focused less on experimentation and more on execution. Agents that can complete workflows, local models that improve data control, and automation that reduces the cost of research and operational work could materially change productivity and operating models. At the same time, greater autonomy brings new questions around security, governance, access controls, investment priorities, and human oversight. Finance leaders will increasingly need to evaluate not only where AI can save time and money, but also where organizations are ready to let AI act on their behalf.
—
/Gustavo
Already a Member? Login HERE.
Identify your path to CFO success by taking our CFO Readiness Assessmentᵀᴹ.
Become a Member today and get 30% off on-demand courses and tools!
For the most up to date and relevant accounting, finance, treasury and leadership headlines all in one place subscribe to The Balanced Digest.
Follow us on Linkedin!
