Tech • AI • Robotics • Game

ENFR

AGI is HERE: ChatGPT 6.2 Astra can automate your work!

7/10
AIParlons IANovember 18, 2026 at 07:00 AM17:29
Audio player

TL;DR

New releases branded ChatGPT 6.2 Sol and ChatGPT 6.2 Astra are being presented as capable of handling real office workflows, from guided visual exploration to partially automated administrative tasks.

KEY POINTS

Shift from benchmarks to work automation

The main claim surrounding the new models is not higher test scores but their ability to complete practical business tasks. The technology is framed as useful only if it can generate measurable output, reduce labor time and improve margins for companies rather than merely produce impressive demos.

Interactive Paris tour showed multimodal control

A live demonstration used ChatGPT 6.2 Astra to conduct a guided tour of Paris, moving from Trocadéro to the Eiffel Tower, Invalides, the Louvre and Notre-Dame. The system described landmarks, changed viewpoints, displayed images, shifted to an aerial view and even altered the scene lighting to 10 PM, illustrating coordinated control of navigation, vision and narration.

Louvre sequence highlighted scene awareness

Inside a modeled version of the Louvre, the system identified major works including the Mona Lisa, The Wedding Feast at Cana, Liberty Leading the People, The Lacemaker, The Coronation of Napoleon and The Raft of the Medusa. It also responded to directional prompts, performed a panoramic sweep of the room and explained which wall the viewer was facing, suggesting a stronger grasp of spatial context than a standard text chatbot.

Model build time remained short

The interior model of the Louvre used in the demonstration was said to have taken 1 hour and 15 minutes to build. That detail is important because it suggests the workflow is being pitched not as a one-off spectacle but as a repeatable production method for tailored environments and interfaces.

Administrative workflow targeted secretarial tasks

The more consequential demonstration involved assigning the model the role of a secretary in a law firm, consulting office, accounting practice or notary environment. The system opened La Poste online services, selected a registered mail workflow, filled recipient details, attached a PDF, added a return receipt and advanced the process to the stage just before payment.

Registered letter example showed supervised execution

During the postal workflow, the model narrated each click and status change: opening the registered mail page, selecting R1 service, handling address entry manually, importing a file and confirming a total of €10.91 with return receipt included. A safeguard prevented autonomous payment, leaving the final validation to a human user while the system prepared a completion report.

Cost-savings pitch focused on back-office functions

The technology was presented as a way for firms to save more than €7,000 per year while improving productivity. The underlying argument is that routine administrative tasks such as mailing, form handling, document submission and interface navigation can be delegated to AI agents under human oversight.

Accessibility benefits were emphasized

Beyond cost reduction, the system was described as potentially useful for people with physical disabilities, difficulty typing or degenerative illnesses. Voice-driven navigation across websites and work interfaces could widen access to office work for users who struggle with conventional input methods.

Human role recast as AI orchestration

Rather than removing people entirely from workflows, the proposed model places humans in a supervisory role. The AI handles the repetitive operational sequence, while a person remains responsible for approval, payment and legal accountability, a setup described as the work of an AI orchestrator.

Job market implications are central to the pitch

The broader message is that companies will increasingly value workers who can build and supervise AI agents that automate real tasks. In that view, the most marketable skill is no longer prompt-writing alone but the design of agent architectures that connect tools, interfaces and approval checkpoints into production workflows.

CONCLUSION

The release of ChatGPT 6.2 Sol and ChatGPT 6.2 Astra is being cast less as a chatbot upgrade and more as a move toward AI systems that can navigate software, complete office procedures and hand final decisions back to humans. If those capabilities prove reliable at scale, the immediate impact is likely to be strongest in administrative work, accessibility tools and agent-based business automation.

Ask a question

More from AI