Jorge InfoBits 20260908: Stop Asking How Smart AI

Subscribe to JorgeTechBits newsletter

Disclaimer: I create this content entirely on my own time, and the views expressed here are mine alone (not my employer’s). Because I love leveraging new tech, I use AI tools like Gemini, ChatGPT, Claude, Perplexity and others as a “digital team” to help research and polish these articles so I can share the best possible insights with you!

September 8, 2026

Stop Asking How Smart AI Is

By Jorge Pereira

AI Adoption & Technology Strategy | Helping people and organizations turn AI, automation, and emerging technologies into practical business outcomes | Workplace Transformation

The chatbot was only our introduction. The real transformation begins when AI becomes a partner in our everyday work.

For the past several years, nearly every major AI conversation has revolved around intelligence.

Which model is the smartest?
Which one achieved the highest benchmark score?
How large is its context window?
How quickly can it reason?

Those questions still matter, but I increasingly believe they are no longer the most important ones.

The more interesting question is:

How do we move beyond occasionally asking AI questions and begin using it as a practical partner in our everyday lives and work?

Most people still experience AI through a chatbot. We open ChatGPT, Copilot, Claude, Gemini, or another application. We ask a question, receive an answer, and then decide what to do next.

That is helpful, but it is only the beginning.

AI has already moved beyond the chat window. It can research information, work with our documents, organize files, operate applications, coordinate tasks, monitor changes, and complete multi-step workflows.

Instead of asking:

“Can you help me write this?”

we will increasingly say:

“Handle this process, keep me informed, and come back when you need my judgment.”

That is a fundamentally different relationship.

AI becomes a working partner for the individual, a digital teammate for a group, and an operational layer for an organization.

Over the past few weeks, I have written about five developments that may initially appear unrelated. One concerns OpenAI’s newest models. Another is about gigawatts and data centers. The others explore agentic operating systems, identities for AI agents, and AI watermarking.

Taken together, I think they point toward the same larger story:

AI is no longer simply a tool we visit. It is becoming an active participant in how we work.

AI as an Everyday Partner

The most significant change may not be that AI is getting smarter.

It is that AI is beginning to participate directly in the work.

A practical AI partner might help someone:

Search years of personal notes and documents
Find connections among previously separate ideas
Research a topic across multiple sources
Organize information for a new project
Monitor incoming messages and identify what needs attention
Prepare a daily summary
Maintain a personal knowledge base
Create a first draft based on previous work
Operate applications and complete routine tasks
Coordinate a workflow across several tools
Return to the person when a decision or approval is required

In The Operating System Is Becoming Agentic, I explored how the operating system itself is becoming far more flexible and responsive to the individual.

For most of computing history, we have adapted ourselves to the operating system. We learned its menus, applications, folders, interfaces, and workflows. We decided which program to open, moved information between applications, and manually performed each step.

Agentic operating environments begin to reverse that relationship.

Instead of adapting ourselves to a fixed collection of applications and predefined workflows, we may increasingly be able to shape the computing environment around what we need to accomplish. We describe the task or desired outcome, and the environment can bring together the appropriate models, agents, tools, applications, files, and computing resources to help us perform it.

The operating system becomes more than a passive platform on which applications run. It becomes an active and adaptable participant in the work.

For the individual, this could mean creating a personalized working environment that understands our information, preferences, tools, privacy requirements, and way of working. Different agents might research, organize, analyze, create, or operate applications while sharing the same environment and returning to us when direction, judgment, or approval is needed.

The important transition is not simply from generating content to completing tasks. It is from people adapting themselves to computers toward computing environments adapting themselves to the people and the work they need to accomplish.

This does not remove the human from the process. It changes the human’s role.

Instead of manually performing every step, we define the objective, provide context, establish boundaries, review the work, and apply judgment where it matters.

Working effectively with AI will therefore require more than learning how to write clever prompts.

We will need to learn how to delegate and how to triage the things that do not work as planned.

An AI partner will not always interpret the objective correctly. A tool may fail. An application may behave unexpectedly. The available information may be incomplete. An agent may encounter a situation outside its permissions or confidence level.

The human will need to identify the exception, determine what went wrong, decide whether to correct or redirect the work, and recognize when the AI should stop altogether.

In many ways, using AI effectively will begin to resemble managing a capable but imperfect digital coworker.

Harvesting the Knowledge We Already Have

This is one of the areas that interests me most.

Many of us have accumulated years, sometimes decades, of useful knowledge.

It is scattered across many forms of digital media, formats (Documents, Notes, Email, Presentations, Project folders, Meeting records, Bookmarks, ,Photos, Screenshots, Blog posts) and personal experiences,

We own all this information, but much of it is effectively trapped.

We may remember that we wrote something relevant five years ago but not where we saved it. Ideas created at different times remain disconnected. Useful research disappears into folders that we rarely open again.

AI gives us an opportunity to harvest that knowledge.

It can help us search, summarize, organize, compare, and connect information that would otherwise remain scattered. It can identify patterns across documents, combine ideas developed years apart, and help us reuse our own work instead of constantly starting over.

This could lead to much more than better search.

When AI helps us connect information that was previously isolated, it can reveal:

New ideas
Unrecognized patterns
Reusable intellectual property
Better decisions
New content
New products or services
New personal and professional opportunities

One of the most valuable uses of AI may not be generating something completely new.

It may be helping us rediscover and make better use of what we already know.

Side note: This is something organizations also need to harvest, manage and govern!

The Hidden Multiplier: Token Consumption

There is a significant consequence to this transition.

A chatbot normally responds after someone sends it a prompt. An agentic system may continue working after receiving the initial objective.

It might break the objective into multiple tasks, search several sources, analyze documents, call tools, communicate with specialized agents, validate its work, and return to the original model for additional reasoning.

What appears to the person as one request could generate hundreds or even thousands of model interactions behind the scenes.

The multiplier begins to look something like this:

More people using AI × more workflows per person × more agents per workflow × more interactions per agent

As AI moves from occasional questions to persistent workflows, token consumption could grow exponentially.

That connects the individual use of AI directly to The Economics of Gigawatts: Blew Mind-Blowing!.

Every AI interaction ultimately depends on physical infrastructure: chips, data centers, networking, cooling systems, electricity, land, and enormous amounts of capital.

AI may feel virtual, but its economics and constraints are very physical.

This is also why model selection matters.

In Understanding OpenAI’s Model GPT-6 Astra and GPT-5.6, I looked at the growing number of choices involving model capability, reasoning effort, speed, and cost.

A simple task may need a small, fast model. A complicated project may require deeper reasoning. Sensitive information may belong with a local or private model. An agent operating applications may require stronger planning and tool-use capabilities.

Using AI well will include learning how to use the appropriate amount of intelligence for the task.

Not every request requires the largest, most capable, or most expensive model.

Privacy Becomes Personal

If AI is going to work with our accumulated knowledge, privacy cannot be an afterthought.

Our information may include financial records, health information, private correspondence, family documents, business ideas, customer information, intellectual property, and years of personal notes.

We should not automatically send all of it to a public cloud service simply because doing so is convenient.

We need to ask:

What information am I sharing?
Where is it being processed?
How long is it retained?
Is it used to train a model?
Who else might have access?
Does this task need the cloud at all?
Could some or all of it be handled locally or within a private environment?

These are not questions only for large organizations or security teams.

They are questions every individual using AI will increasingly need to consider.

Local, Private, and Cloud AI

I do not believe the future will consist of one enormous cloud model doing everything for everyone.

The more practical future is hybrid.

Local, private, and cloud AI are not mutually exclusive choices. They can participate in the same workflow. The task, the information involved, the required capability, and the acceptable level of risk should determine which environment or combination of environments is used.

Cloud AI may provide broad knowledge, advanced reasoning, large-scale research, convenience, and access to the newest capabilities.

Private AI may be appropriate for organizational information, governed business processes, controlled data, and workflows with specific security or compliance requirements.

Local AI is especially interesting for individuals, both personally and within organizations. It offers a way to work with private documents, sensitive information, and proprietary knowledge while maintaining greater control over where that information is processed.

It can also make AI more economically viable.

Not every classification, summary, extraction, search, or routine task needs to be sent to a large cloud model and billed by the token. Smaller local models may be able to handle many recurring tasks without creating a new cloud charge every time the workflow runs.

A hybrid process could:

Use a local model to search and summarize private documents
Keep sensitive or proprietary information on the device
Send only the necessary non-sensitive context to a cloud model
Use the cloud model when more advanced reasoning or outside knowledge is required
Return the result to a local system for storage or additional processing

The goal is not to declare that local AI is always better than cloud AI.

The goal is to understand the choices and match the technology to the task.

With Greater Capability Comes Greater Authority

Once an AI partner can act, intelligence and privacy are not enough.

It also needs defined authority.

An agent may eventually have access to our calendars, email, files, applications, subscriptions, business systems, and online accounts.

That could be incredibly valuable, but it also introduces risk.

We need to understand:

Which agent is acting?
Who owns or controls it?
What information can it access?
What actions can it perform?
What requires human approval?
What happened while it was working?
What should happen when something fails?
Can its access be removed?

This is why Copilot Studio Moves to Entra Agent ID caught my attention.

Giving agents governed identities shows that they are no longer being treated merely as experimental features.

They are becoming active participants in our digital environments.

When AI only suggests an answer, identity may not seem particularly important.

When AI can take action, identity, ownership, permissions, monitoring, and lifecycle management become foundational.

Trust Must Become Part of the System

We will also need to evaluate what our AI partners find and create.

AI can already generate realistic text, images, audio, and video. As those capabilities improve, simply looking at something will not always tell us where it came from or whether it has been modified.

Was it created by a person?

Was AI involved?

Was the original altered?

Who published it?

Can its history be verified?

In The New Operational Reality: AI Watermarking, I examined why watermarking and content provenance are becoming operational concerns.

Watermarking will not solve every problem. Content can be edited, transformed, captured, or deliberately stripped of identifying information.

But the larger idea remains important.

People will need stronger digital literacy and an appropriate degree of skepticism. Technology platforms will need better methods for communicating how content was created and whether its history can be trusted.

Trust cannot depend entirely on whether something looks real.

Looking Beyond the Hype

These developments are connected.

More capable models give AI the intelligence to perform increasingly complex work.

Agentic computing gives that intelligence somewhere to operate.

Our documents and data provide the context that makes an AI partner personally useful.

Local, private, and cloud environments determine where different parts of the task should be processed.

Identity and permissions establish what an agent is allowed to do.

Watermarking and provenance help us evaluate what those systems produce.

And beneath everything are the models, tokens, data centers, energy, chips, and capital required to keep it running.

Put those pieces together and AI begins to look less like another application and more like a new working environment:

Intelligence + knowledge + compute + tools + identity + permissions + privacy + trust

This is an organizational transformation, but it is also a deeply personal one.

People will need to adapt to working with AI well beyond the chatbot. We will need to learn how to collaborate with it, delegate everyday tasks, triage exceptions, connect our knowledge, protect our information, supervise agents, and apply human judgment where it matters most.

My Practical InfoBit

The chatbot was only our introduction to AI.

The more important transformation begins when AI becomes a practical partner that can work across our knowledge, applications, and everyday workflows.

The next phase will not be defined only by who builds the smartest model. It will be defined by how effectively people learn to work with AI while managing the compute, cost, privacy, authority, and trust that partnership requires.

Perhaps we should stop asking only:

“How smart is the AI?”

And begin asking:

What can it help me accomplish? What information should it access? Where should each part of the task be processed? What is the AI authorized to do? What happens when something goes wrong? And can I trust what it produces?

The real opportunity is not simply having better conversations with AI.

It is learning how to build better ways of working with it.

As AI becomes an active partner across our knowledge, workflows, and decisions, how should we prepare ourselves for that change?

Would love to hear your thoughts!

— Jorge

👉 About Jorge | 👉 Follow with me on LinkedIn | 👉 Send me a note

Keep learning. Keep experimenting. Keep connecting the dots.

Articles and Further Reading

The Operating System Is Becoming Agentic – What caught my attention is the movement from AI assistants that respond to individual prompts toward agentic environments that can operate applications, use tools, coordinate tasks, and continue working. It matters because people will increasingly need to direct outcomes, delegate work, establish boundaries, and triage exceptions instead of manually performing every step.
Understanding OpenAI’s Model GPT-6 Astra and GPT-5.6 – What caught my attention is that selecting an AI is no longer simply a matter of choosing the “best” model. Different combinations of capability, reasoning, speed, and cost are appropriate for different workloads. It matters because individuals and organizations will need to match intelligence to the task instead of using the most powerful and expensive option for everything.
The Economics of Gigawatts: Blew Mind-Blowing! – What caught my attention is the direct relationship among AI revenue, computing capacity, and energy. It matters because persistent agents and multi-step workflows could consume far more tokens than occasional chatbot conversations. The economics of chips, electricity, data centers, and compute will influence which AI capabilities remain affordable and widely available.
Copilot Studio Moves to Entra Agent ID – What caught my attention is that AI agents are receiving identities designed specifically for them. It matters because once agents access information and take action, we need to know who or what acted, what it was permitted to access, who owns it, and how its activity and lifecycle are governed.
The New Operational Reality: AI Watermarking – What caught my attention is the growing effort to establish the origin and history of AI-generated content. It matters because people cannot rely only on their eyes and ears to determine what is authentic. Provenance, disclosure, verification, and digital literacy will become essential parts of working responsibly with AI.

About This Edition

Beyond the AI Hype – Short, practical, and sometimes opinionated insights about using AI as a working partner—connecting knowledge, improving everyday workflows, protecting privacy, and exploring what becomes possible with local, private, and cloud AI.

About Jorge

I’m Jorge Pereira — a technology strategist, advisor, mentor, and lifelong technologist focused on AI, technology transformation, and helping people and organizations understand what’s next and put it to practical use.

👉 About Jorge | 👉 Connect with me on LinkedIn | 👉 Send me a note