
When Will Superintelligence Happen? Why Companies Should Not Set a Date
No one can reliably say when superintelligence will arrive. Learn why your business plan should not depend on a date and which signals to track instead.
Want to apply this in your company?
Book a call with Juan →No one has a reliable date for superintelligence, and experts do not agree on when, or whether, it will arrive. For a company, the practical problem is not guessing a year. It is deciding what to do with the AI systems available today and the ones likely to appear in the next purchasing and operating cycles. The recommendation is to remove dates from the plan and track observable signals that change concrete decisions.
Key takeaways
- Date forecasts for superintelligence are hypotheses, not facts that can support a budget, a contract or a hiring decision.
- A plan built around a date tends to fail in two ways: buying too early, or postponing decisions that should already be made.
- Useful signals are the ones your company can measure: performance in its own workflows, cost, applicable rules and incidents.
- Review the plan at fixed intervals, with criteria written down before each decision.
- The map, limit, pilot and transfer method works with or without any forecast.
Why nobody can back up a date
The term superintelligence is used with different meanings. For some people, it means a system that outperforms humans at nearly every economically relevant cognitive task. For others, it is simply a way of talking about models that keep getting more capable. A date depends on a stack of assumptions: that technical progress continues at its current pace, that available computing power keeps growing, that today's architectures lead to the goal, and that regulation and the economy keep up. Each assumption is uncertain, and the combination is even more uncertain.
Forecasts from well-informed people also change often, and the confidence with which they are presented is not the same as evidence. To separate the concepts, it helps to read the superintelligence glossary for executives and the comparison between AGI, ASI and AI agents. Both help distinguish what exists in products sold today from what is still a research and debate topic.
What a deadline does to planning
When leadership picks a date, it turns a hypothesis into a rule. This usually creates three problems. The first is buying capacity for a scenario that may never materialize, with long contracts for tools that change quickly. The second is delay: the team waits for the next leap, and processes that already work stay manual. The third is pressure to move without criteria, when a competitor announces something and the company reacts without understanding whether it affects its own processes.
There is also a legal consequence. Obligations such as Brazil's LGPD apply now, to the data the company handles now. The Brazilian National Data Protection Authority (ANPD) publishes guidance and is the official source for following the topic in Brazil. No forecast about superintelligence changes what is already required today.
Which signals to track instead
Instead of a date, track signals you can verify with data from your own operation and from official sources. They should be few, clear and tied to a decision.
- Performance in your workflows: the share of responses that needed correction, cases handed to humans, and tasks the system completed without error.
- Cost and availability: price changes, usage limits, and discontinuation of models or versions you already rely on.
- Real new capabilities: features that reached the product you pay for and were tested on your own cases, not just announced.
- Rules and obligations: developments in Brazil's AI legal framework and guidance on personal data processing and automated decisions.
- Incidents: security failures, data leaks or incorrect answers that affected customers, including those at other companies in your sector.
- Team capacity: how many people can review, adjust and operate the workflows without depending on a single vendor.
A signal matters only if it changes a decision. If your company would not know what to do when a signal shifts, it does not belong on the dashboard.
How to build a signal dashboard in five steps
A simple dashboard stops endless debates about the future and focuses the conversation on what can be decided.
- List the decisions that depend on some expectation about AI, such as expanding an agent, switching vendors or hiring for a new function.
- For each decision, write down which signal would change the answer and in which direction.
- Set a threshold before looking at the data. For example: if the correction rate exceeds the limit agreed with operations, the workflow returns to human review.
- Schedule a recurring review on the calendar, unrelated to forecasts. The cadence is a management choice, not a bet.
- Record each decision with the date, the observed signal, the reason and who approved it. This history is worth more than any projection.
Illustrative scenario: a clinic network decides without a date
Illustrative scenario: a network of clinics wants to use an AI agent on WhatsApp to confirm appointments and answer questions about preparation instructions. Leadership debates signing a long, multi-year contract because an executive read that the technology will change a lot. Instead of picking a date, the operations team proposes starting with a single queue, appointment confirmations, with a log of every conversation, handoff to a human attendant for any clinical question, and a correction threshold approved before launch. After a review period, the decision depends on what was observed: keep it as is, expand to rescheduling, or stop. The long contract waits until there is their own history to justify it.
How Juan Carlo's method applies
When working with companies, the path usually follows four stages: map the current process and measure the baseline, limit the data and actions the agent can use, pilot a real workflow with logs and a success metric agreed in advance, and then transfer the operation and decide whether the project will be expanded, adjusted or closed. This sequence does not depend on any forecast. It places the decision at the point where the company has its own data, which is exactly why it works well when the future is uncertain.
If your leadership team still discusses this topic in the abstract, the piece on what a founder should decide alone and what to take to the team helps separate the decisions that need consensus from those that can be tested on a small scale.
Risks that exist today, with or without superintelligence
The most concrete risks of an AI agent are in the present: incorrect answers delivered with an appearance of certainty, malicious instructions hidden in customer messages, excessive access to internal systems, and processing of personal data without a clear legal basis. Frameworks such as the NIST AI Risk Management Framework and the OECD AI Principles address governance, transparency and human oversight, and they are useful references for any company building its controls. These controls do not depend on knowing when the next generation of models will appear.
Frequently asked questions
When will superintelligence happen? There is no reliable answer to that question, and the available predictions depend on assumptions that change over time. For a company, the useful question is different: which decisions today depend on some expectation, and which signals would show that the expectation needs to be revised.
Should I wait for superintelligence before investing in AI? You do not need to wait. Systems available today can already automate parts of customer support, triage and administrative routines, as long as the process is mapped, limited and monitored. Waiting usually means only losing the learning about your own business.
How do I follow AI progress without becoming a specialist? Choose a few signals tied to your decisions, such as performance in your workflows, cost, applicable rules and incidents. Review them at fixed intervals and record your decisions. To understand the terminology without getting lost in debates, the glossary for executives is a good starting point.
What should I do if a vendor says superintelligence is close? Treat the claim as a sales opinion and ask for what matters to your decision: how the product behaves in your process, which data it accesses, how errors are logged, and what the exit conditions of the contract are. A promise about the future does not replace a pilot with an agreed metric and human review.