OpenAI says AGI is here. Are we ready?
“Welcome to the AGI era.”
Greg Brockman closed OpenAI’s September 3, 2026, Astra briefing with those words. Nvidia Chief Jensen Huang doubled down saying “AGI has arrived.” —>
So has it? I view it less as a milestone marker and more as an invitation. With this incredible new technology, how much more would we attempt?
That is where my mind goes when I hear a declaration like this.
Of course, there are fascinating questions about whether Astra qualifies as artificial general intelligence. What is “AGI”? What does this model understand? How general are its capabilities? How reliably can it operate in unfamiliar situations?
Alongside those questions sits an immediate opportunity: figuring out what people can accomplish with increasingly capable machine workers.
For employers, that opens up questions about capacity, cost, and the design of work. For workers, it raises possibilities for new skills, greater independence, and different careers. For schools and governments, it invites a much larger exercise in imagination.
What could we build if more people had the expertise and capacity to act on their ideas?
And how do we prepare for the changes that come with it?
01\ THE ECONOMIC QUESTION BEHIND AGI
OpenAI’s charter defines AGI as “highly autonomous systems that outperform humans at most economically valuable work.” —>
The emphasis is on autonomy and economic capability.
For knowledge work, I find this a useful starting point:
Can a machine enter an unfamiliar work situation, identify what matters, and produce a useful result without someone scripting every step?
Think about what that would involve.
You give a system an objective. It gathers context, works through the relevant information, uses approved tools, tests its output, and brings back a deliverable. When it encounters a consequential decision or reaches the limits of its authority, it asks for help.
That is what I mean here by a machine worker.
Imagine asking one to investigate why customers abandon a particular service. Within an agreed scope, it could examine support conversations, analyze usage data, identify patterns, and prepare several explanations for your team to test.
Here is OpenAI’s vision for the future —>
The value a system like this generates would depend on the quality of the investigation, the time it required, and what your team could do with the findings.
That gives us an economic question we can actually investigate: How much useful work can we reliably delegate, at what cost?
Every task that clears that threshold creates choices. An organization could reduce a backlog, improve a service, investigate more ideas, or give people time for work that keeps getting postponed.
Enough of those choices, spread across an economy, could change what we expect a person, a team, or a company to accomplish.
02\ WHAT THE EVIDENCE SHOWS
Consider the range of work researchers and companies are reporting.
AI-generated genomes produced bacteria-killing phages. Stanford researchers used Evo 2 to generate viral genomes and tested candidates in the laboratory, identifying 16 effective phages. One combination overcame E. coli resistance to the original phage.
Claude formalized Fermat’s Last Theorem in 11 days. Anthropic reported that Claude worked largely autonomously to produce a complete Lean-checked proof. The achievement involved translating mathematical reasoning into a form a computer could verify, building on existing mathematics.
GPT-5 helped improve a molecular cloning protocol’s efficiency by 79×. In a controlled experiment, model-proposed changes increased the number of sequence-verified clones produced from a fixed DNA input. The gain applied to that specific experimental protocol.
Salesforce projected 500,000 employee hours saved annually through Agentforce in Slack. The company described an internal rollout that developed through months of testing, feedback, and refinement.
JPMorganChase onboarded 200,000 users to LLM Suite in eight months. Its internal platform supported idea generation and content drafting, with further integration into company data and workflows planned.
IBM reported that AskHR handled 94% of common questions within the platform. Its system also supported automated tasks such as employee letters, vacation requests, and payroll access.
Walmart announced AI tools designed for 1.5 million U.S. associates. The rollout included task management, translation, and conversational assistance for frontline work.
Taken together, I see an expanding field of applications across science and everyday operations. What excites me is the possibility of increasing the number of things we can try.
Imagine a laboratory with twenty promising ideas and the capacity to investigate three. What would happen if that team could pursue ten or a hundred? Imagine a small manufacturer able to explore a new product line, or a teacher with the time and tools to develop a richer learning experience.
What is sitting on the “someday” list because the expertise, time, or budget has always been out of reach?
That is where I would start looking for opportunity.
03\ MAKING DELEGATION WORK
Turning that possibility into dependable work is a practical undertaking.
I would begin with a real assignment: a customer escalation, a procurement analysis, a software maintenance task, or a sales proposal. Choose work the organization understands well enough to evaluate.
Then track five things:
Accepted completion rate: How often does the system produce work that meets the required standard?
Review burden: How much qualified human attention does each accepted output require?
Error profile: What goes wrong, how easily can people detect it, and what are the consequences?
Authority compliance: Does the system stay within its permissions and ask for help when required?
Cost per accepted output: What does satisfactory work cost after tools, retries, review, correction, and monitoring?
These measures help answer a straightforward question: can we confidently give this system more work?
For example, consider a procurement team receiving an analysis with the underlying sources attached, assumptions clearly identified, and unresolved questions ready for review. Over repeated assignments, the team could learn where the system performs well and where expert attention adds the most value.
That knowledge would let it expand delegation deliberately.
The unit I would measure is accepted work.
I would also keep a list of newly feasible projects. The economics become particularly interesting when a team can take on useful work that previously had no place in the budget.
There is a learning opportunity here, too.
Picture a junior analyst working through a difficult case with an AI tutor. They try an approach, receive feedback, examine alternatives, and explain their reasoning to a manager. The next case introduces a new complication.
Could we use these systems to give people more varied practice, earlier in their careers?
That possibility deserves attention alongside automation.
In designing such a workflow, I would ask which parts the learner should attempt independently, where AI feedback would help, and how a manager would assess progress. The aim would be to build judgment through increasingly challenging work.
A well-designed deployment could give an organization more productive capacity and a stronger way to develop its people.
That is a future worth getting excited about.
04\ THREE FUTURES AT ONCE
Where could all this lead? I see three possibilities on a spectrum:
A) THE BORG:
The Borg, from Star Trek, represents intelligence organized around a single overriding objective. Everything gets coordinated. Resources are allocated efficiently. The system absorbs more of the activity around it, and its priorities shape everyone’s choices.
B) WALL-E:
WALL-E brings a different possibility to mind: life becomes extraordinarily convenient. Machines anticipate our needs, arrange our surroundings, and handle the effort of everyday life. Comfort becomes the organizing principle.
C) THE FEDERATION:
The Federation, Star Trek’s pluralistic alliance of worlds, is the future that captures my imagination. Technology expands access to knowledge, reduces material constraints, and gives people the means to explore.
We could experience elements of all three futures at once. Different institutions will make different choices about control, convenience, and participation.
That makes the question of objectives especially interesting.
Suppose a future AI could predict the consequences of our choices with extraordinary accuracy. We could ask it to explore ways to improve health, increase prosperity, protect an ecosystem, or give people more leisure.
We would still have a great deal to discuss: whose preferences count, which trade-offs are acceptable, and how people can challenge a decision.
What is worth doing?
That question could become more exciting as our ability to act on the answer grows.
05\ READINESS IS SOMETHING WE BUILD
So, are we ready?
I see readiness as something we develop through use, experimentation, and deliberate choices.
For companies, I would start with a portfolio of real work. Measure results, learn where delegation succeeds, expand carefully, and give teams room to propose new projects. One question I would put to every department: What would you attempt if you had 2X or 10X the capacity?
For schools, I would create opportunities for students to investigate substantial questions, build things, test ideas, and explain their reasoning. AI could become part of a richer learning environment, with teachers designing the right balance of guidance, practice, and independence. What could a student accomplish with knowledgeable support available whenever they got stuck?
For governments, I would explore how broader access to expertise could improve public services and help communities solve problems. I would also put access, competition, worker transitions, and the distribution of benefits into the design of those efforts. What could a small town do with capabilities previously available only to a large city?
These are questions we can begin working on now. Each experiment can teach us something about the technology and about the institutions using it.
The part I find most exciting is how many more people might get to participate.
More people able to investigate an idea. More teams able to build something useful. More communities able to act on what matters to them.
That is the possibility I hear in an invitation to the AGI era.
== What have we always wanted to try and what would it take to begin?
Until next time,
Ram
—
Ram Srinivasan
MIT Alum | Author, The Conscious Machine | Global Future of Work and AI Adoption Leader published in Business Insider, Fortune, Harvard Business Review, MIT Executive Viewpoints and more.
—
A Message From Ram:
My mission is to illuminate the path toward humanity's exponential future. If you're a leader, innovator, or changemaker passionate about leveraging breakthrough technologies to create unprecedented positive impact, you're in the right place. If you know others who share this vision, please share these insights. Together, we can accelerate the trajectory of human progress.
Disclaimer:
Ram Srinivasan currently serves as an Innovation Strategist and Transformation Leader, authoring groundbreaking works including "The Conscious Machine" and the upcoming "The Substrate Shift."
All views expressed on "Substrate" and across all digital channels and social media platforms are strictly personal opinions and do not represent the official positions of any organizations or entities I am affiliated with, past or present. The content shared is for informational and inspirational purposes only. These perspectives are my own and should not be construed as professional, legal, financial, technical, or strategic advice. Any decisions made based on this information are solely the responsibility of the reader.
While I strive to ensure accuracy and timeliness in all communications, the rapid pace of technological change means that some information may become outdated. I encourage readers to conduct their own due diligence and seek appropriate professional advice for their specific circumstances.
