From AI Assistant to AI Colleague
ChatGPT Work agent is an autonomous AI system inside ChatGPT that can handle multi-step office tasks such as research, document creation, spreadsheet management, and web app development for hours with minimal human oversight, effectively behaving less like a chatbot and more like a junior digital employee.
That shift is the headline: ChatGPT now works like an employee. OpenAI’s new ChatGPT Work agent can research, build documents, manage spreadsheets, and create web applications for hours at a time without constant human supervision. It runs on the new GPT-5.6 family, designed for complex, chained tasks rather than one-off prompts. In other words, this is not “type a question, get an answer” AI; it is “assign a project, review a deliverable” AI. The result is a direct encroachment on the day-to-day work junior professionals used to own—especially the repetitive research, formatting, and basic coding that form the backbone of many first jobs.
Agentic AI Automation and the New Desktop Workbench
ChatGPT Work is designed for multi-step tasks that would normally require a human assistant: monitoring connected apps, running scheduled jobs, and producing finished deliverables. It can tap into local files, a built-in browser, and third-party plugins across web, mobile, and desktop, then keep going without someone constantly steering it. A new ultra acceleration mode even coordinates multiple agents across parallel workstreams to finish complex tasks faster. This is agentic AI automation in practice: many small decisions, taken continuously, without a human in the loop every minute.
OpenAI is also merging its Codex developer tool directly into the ChatGPT desktop app for macOS and Windows, and the new unified app combines Chat, Work, and Codex in one interface. Existing Codex users can update while keeping their projects, settings, and workflows. Add the new Sites beta, which turns work into interactive websites or lightweight web apps from inside ChatGPT, and you get a single pane of glass for knowledge work. This isn’t a cute productivity add-on; it is an AI control room where a single agent can research, code, write, and ship artifacts end to end.
Signals of AI Job Displacement in Finance and Beyond
The technology story sits on top of a blunt labor signal. Investment bankers might be next in line to be rendered obsolete by artificial intelligence if OpenAI's latest push into the financial space is any indication. The company has opened a new position for an investment banking expert whose responsibilities include making ChatGPT and related models better at handling complex transactions such as mergers, acquisitions, fundraising, and other major deals. That expert is expected to define “the quality bar for AI-assisted investment banking” across research, analysis, valuation, modeling, diligence, transaction execution, and client materials.
Here is the uncomfortable part: in exchange for teaching an AI to do their own job, OpenAI’s future investment banking expert will be offered as much as USD 205K (approx. RM943,000) a year, plus equity. The job is explicitly about translating banking workflows into representative evaluation tasks so AI can improve its performance on financial work. Put plainly, the firm is paying a senior banker to industrialize the replacement of junior analysts. You cannot read that as neutral upskilling. It is AI job displacement, targeted squarely at the lowest rung of a very lucrative ladder.
Access, Incentives, and the Risky Middle Ground
Access to this new AI autonomy is already stratified. ChatGPT Work is available today for Pro, Enterprise, and Edu users, with Plus and Business subscribers getting access in the coming days, while Free and Go users are excluded from the agent. Free and Go users instead default to GPT-5.6 Terra, while Plus and above can choose among Sol, Terra, and Luna. Sol, the flagship model, costs USD 5 (approx. RM23) per million input tokens and USD 30 (approx. RM138) per million output tokens, Terra runs USD 2.50 (approx. RM12) input and USD 15 (approx. RM69) output, and Luna costs USD 1 (approx. RM5) input and USD 6 (approx. RM28) output. OpenAI claims Sol delivers state-of-the-art results across coding, cybersecurity, and science while using fewer tokens than competing frontier models.
At the same time, OpenAI has been adding connectors for personal financial accounts so ChatGPT can access bank records and other financial data, a feature rolled out to Plus and Pro users at the end of June. All this rolls out while the company is in the process of going public and has slipped from being the darling of the AI world to playing second fiddle to a rival that filed its IPO documents first. Some in the financial industry are already warning that the AI bubble OpenAI helped inflate could pop, potentially taking the global economy with it. That is the risky middle ground we are entering: more AI autonomy, more financial integration, but still models that have been known to get things wrong and even make things up rather than admit uncertainty.
The Future of Junior Work When AI Does the Busywork
The launch of ChatGPT Work on GPT-5.6 and the parallel hiring of an investment banking subject matter expert point in one direction: the entry-level spreadsheet-and-PowerPoint phase of professional life is no longer safe. When an AI agent can watch your connected apps, run scheduled jobs, and output finished deliverables, the economic temptation is to treat it as a drop-in substitute for junior staff, not a mere assistant.
Over the next 24 hours, GPT-5.6 models will roll out, GPT-5.4 will be retired on July 23, and GPT-5.5 models will remain available. That timeline underlines how fast this shift is happening. The question is not whether agentic AI automation will change knowledge work; it is whether organizations will use the savings from automating AI autonomous tasks to build better human careers or to hollow out the bottom of the ladder. If junior roles collapse into oversight of AI workflows, then “experience” becomes something you supervise rather than something you gain. In that world, the real scarcity is not intelligence—it is opportunity.
