When AI Goes to Work

Imagine you're planning a long weekend away.

Today, you might open an AI assistant and ask: “What are some good hotels in Montreal for a family of five?”

Within seconds, you'll probably get a useful list. You can ask follow-up questions about neighborhoods, restaurants, attractions, or whether it's worth paying more to stay downtown.

Helpful? Absolutely. But you're still doing the work.

You're opening hotel websites. Comparing prices. Checking flights. Looking at your calendar. Making reservations. Adding confirmation emails to your itinerary.

Now imagine a slightly different conversation.

You tell your AI assistant:

“Plan a three-day trip to Montreal in October. Keep the hotel under $300 a night, use my airline miles if they're worthwhile, find a few restaurants my family would enjoy, and make sure we don't have anything important on the calendar.”

The AI checks your schedule. Searches flights and hotels. Compares using points with paying cash. Builds an itinerary. Finds restaurants near your hotel.

Then it asks: “I found a good option for $1,480. Would you like me to book it?”

You say yes. And it does.

That's a very different kind of artificial intelligence.

For the first several years of the consumer AI revolution, we've mostly interacted with AI by asking it questions. The next phase may be about giving it jobs.

Welcome to the age of the AI agent.

From Chatbots to Agents

The distinction between today's familiar chatbots and emerging AI agents can sound more complicated than it really is.

Think about it this way:

• A chatbot answers.

• An assistant helps.

• An agent acts.

Ask a chatbot how to organize a trip and it can give you instructions.

An AI assistant might help you research flights, compare hotels, and build an itinerary.

An AI agent could potentially take the next step: interact with other services and actually make the arrangements.

That's important because much of everyday life doesn't consist of answering questions. It consists of completing chains of small tasks.

Consider something as ordinary as scheduling a doctor's appointment.

You need to find the doctor's number or website, determine when you're available, find an appointment, make the booking, put it on your calendar, perhaps arrange transportation, and remember to complete any paperwork beforehand.

None of those tasks is particularly difficult. Together, however, they consume time and attention. An AI agent could be given the objective—schedule my annual physical—and work through those steps on your behalf.

That's the promise of agentic AI: rather than requiring us to direct every individual step, we increasingly tell the technology what outcome we want.

The AI figures out how to get there.

Your New Digital Assistant

If this works, the most important AI product in your life may eventually be less like Google and more like a highly capable personal assistant.

With your permission, imagine an AI that can access portions of your calendar, email, travel accounts, shopping history, subscriptions, contacts, documents, and financial information.

It wouldn't simply know that you asked about a trip to Montreal.

It might know that your children have school on Friday, that you prefer nonstop flights, that you usually choose Marriott hotels, that you have 70,000 airline miles available, and that your passport expires next year.

That sounds extraordinarily useful.

It could notice that a flight was canceled and begin searching for alternatives before you even reach the airport.

It could identify three subscriptions you haven't used in six months and ask whether you'd like to cancel them.

It could compare your family's schedules and find a Saturday afternoon when everyone is free.

It could prepare your grocery order based on the meals you've planned for the week.

It could notice that your car is due for service, find nearby appointments, compare them with your calendar, and offer you three choices.

Much of this is technically less glamorous than asking an AI to write a novel or create a movie. But it could prove far more consequential. The real AI revolution may arrive when artificial intelligence begins removing thousands of tiny administrative burdens from everyday life.

When AI Goes to Work

The same transition is beginning in the workplace.

This is often framed in dramatic terms: Will AI take my job?

That's an understandable question, but it may not be the best place to start. A job isn't one activity. Most jobs are bundles of dozens—or hundreds—of different tasks.

A teacher teaches, but also creates lesson plans, answers emails, grades assignments, prepares presentations, fills out forms, schedules meetings, communicates with parents, and searches for classroom materials.

A lawyer advises clients, but also reviews documents, performs research, drafts correspondence, summarizes cases, manages files, and tracks deadlines.

A manager makes decisions, but also sits through meetings, reads reports, prepares presentations, answers routine questions, organizes projects, and searches for information.

AI doesn't need to replace the teacher, lawyer, or manager to dramatically change those jobs. It only needs to become capable of doing some of the tasks inside them.

That creates a different question: Which parts of your job actually require you?

The answer may become increasingly important.

If AI handles more of the routine research, formatting, scheduling, summarizing, drafting, and administrative work, people may spend more time on the parts requiring judgment, relationships, leadership, creativity, negotiation, expertise, and responsibility.

Of course, that doesn't mean every worker or occupation will benefit equally. Some jobs contain more automatable tasks than others, and businesses may decide that AI allows them to operate with fewer people in certain roles.

There will be disruption. But "AI replaces jobs" is too simple a description of what may happen. In many cases, AI will first rearrange jobs.

When AI Gets a Wallet

There's another hurdle between today's AI assistant and a truly useful AI agent.

Money.

Imagine telling an AI: “Find me the best deal on a new dishwasher and have it delivered Saturday.”

The AI can research dishwashers. It can compare reviews. It can probably determine which models fit the dimensions of your kitchen. But to finish the job, it eventually needs the ability to buy something.

That raises a surprisingly complicated set of questions.

• Should your AI have access to your credit card?

• How much should it be allowed to spend?

• Can it choose a different retailer to save money?

• Could it negotiate a price?

• What happens if it buys the wrong dishwasher?

• Who is responsible if an automated purchase goes wrong?

And how does a merchant know that the software placing an order is genuinely authorized to act for you?

These questions are one reason the future of AI may intersect with technologies that, until now, have seemed unrelated to it.

Digital wallets. Digital identity. Programmable payments. Stablecoins. Authentication systems.

An autonomous software agent may eventually need its own version of a wallet—not necessarily a bank account in the traditional sense, but a secure way to transact within limits established by its owner.

You might tell your AI:

• Groceries: Up to $200 per week.

• Travel: Research freely, but ask before purchasing.

• Subscriptions: Renew anything under $25 per month.

• Unexpected purchases: Always ask.

Suddenly, programmable money doesn't sound quite so abstract. It becomes part of answering a practical question:

How do you safely give software permission to spend your money?

The Permission Problem

And that brings us to perhaps the biggest challenge facing useful AI agents.

For AI to become more helpful, it needs access. But access creates risk.

Tell an AI to "manage my travel" and think about what that seemingly simple instruction could require.

• Your calendar.

• Your email.

• Your passport information.

• Your credit card.

• Your airline accounts.

• Your hotel loyalty programs.

• Perhaps your location.

• Maybe even information about your family.

The more an AI knows—and the more systems it can access—the more useful it can become. But the consequences of mistakes, security failures, or misuse also become greater.

There's an enormous difference between an AI hallucinating the name of a restaurant and an AI incorrectly transferring $5,000. That means one of the most important technologies of the AI-agent era may not be the agent itself.

It may be the permission system surrounding it.

Humans Still Make the Rules

We already understand this concept in other parts of life.

You might give a teenager a credit card but set a spending limit.

An employee might be authorized to sign contracts up to a certain amount but need executive approval above it.

A financial adviser might manage investments within guidelines you've established without asking permission before every trade.

AI agents could operate similarly. Imagine giving your personal AI a set of standing instructions:

Go ahead: Add appointments to my calendar.

Go ahead: Reorder household products I've purchased before.

Go ahead: Cancel unused subscriptions under $20 per month.

Ask first: Spend more than $250.

Ask first: Share personal information with a new service.

Never: Make major financial decisions.

Never: Make medical decisions without my involvement.

The important concept here is bounded autonomy.

An AI doesn't necessarily have to be either completely controlled or completely autonomous. It can have freedom to act inside boundaries established by you.

As these systems become more capable, learning to establish those boundaries may become an important form of digital literacy.

The future may require us to become good managers of AI, not simply good users of it.

A Day in 2035

It's easy for conversations about artificial intelligence to drift into science fiction. So imagine a future that is considerably less dramatic.

It's an ordinary Tuesday in 2035.

At 6:45 in the morning, your AI assistant tells you that school will open two hours late because of snow. It has already identified the two morning meetings affected and drafted messages asking whether you'd like to reschedule them.

You approve both.

While you're getting breakfast ready, it gives you a three-minute summary of the emails and news relevant to your work. The other 47 messages can wait.

Later that morning, your car reports that routine maintenance is due. Your AI finds three appointments that fit next week's calendar and asks which one you'd prefer.

At lunch, it notices that the airfare for a trip you're taking next month has dropped substantially. Your ticket can be changed without a fee, so it asks whether you'd like to rebook and receive the credit.

You approve it.

At 4:00, you receive a message from your child's school about an event next Thursday. Your assistant notices a scheduling conflict and flags it.

On the way home, it reminds you that you're missing two ingredients for dinner and asks whether you'd like them added to the grocery delivery already scheduled for 5:30.

That evening, it notices an unfamiliar recurring charge on your credit card statement.

It doesn't cancel anything. It asks you what you'd like to do.

There are no humanoid robots. No sentient machines. No science-fiction supercomputer running your life.

Just dozens of small tasks that once required your attention quietly being organized by software operating within rules you've established.

That may be what living with advanced AI actually looks like.

What Happens to Us?

If machines become better at handling routine cognitive work, an obvious concern remains. What happens to human skills?

Some will undoubtedly become less valuable.

We no longer prize the ability to perform long division as highly as we once did because calculators can do it instantly. GPS changed how often we need to memorize directions. Search engines changed how we find information.

AI will do the same to other skills. But technological change can also increase the value of capabilities machines struggle to replicate.

Knowing facts may become less important than knowing which questions to ask.

Producing a first draft may become less important than recognizing whether it's good.

Completing routine analysis may become less valuable than knowing what the analysis actually means.

And as AI becomes more capable, qualities such as judgment, curiosity, leadership, empathy, taste, expertise, creativity, and responsibility may become more—not less—important.

AI can recommend an action. Someone still needs to decide whether it's the right one.

AI can generate possibilities. Someone still needs to choose among them.

AI can perform tasks. Someone still needs to decide which tasks should be delegated in the first place.

The future of work may therefore be defined not simply by what AI can do.

It may be defined by what humans decide not to hand over.

Living Alongside AI

Four weeks ago, we began this series with a simple question: What can AI actually do?

From there, we asked whether we can trust its answers.

Then we looked behind those answers to ask who controls the companies and infrastructure powering artificial intelligence.

Now we've reached perhaps the most consequential question of all.

What happens when AI begins acting on our behalf?

The answer won't be determined by technological capability alone.

It will depend on the rules we establish around these systems, the access we're willing to give them, the companies and institutions we trust, and the decisions we continue to reserve for ourselves.

AI will almost certainly become more capable. Our assistants will get better. Agents will perform more complicated tasks.

The boundary between asking software for information and asking software to accomplish something will continue to blur.

But greater capability doesn't require surrendering greater control.

The most useful future may be one in which artificial intelligence handles more of the work we don't particularly want to do while leaving us more time and attention for the things that require something distinctly human.

The challenge isn't simply learning how to use AI.

It's learning what to delegate, what to supervise, and what to keep for ourselves.

And that may be the most important AI skill of all.

Other Articles In This Issue:

• Living With AI: What It Can Do, What It Can't, and Why It Matters

• Can You Trust AI?: How to Separate Helpful Answers from Confident Mistakes

• Who Controls AI?: Why the Future of Artificial Intelligence Depends on More Than Technology

  CHECK OUT OUR PREVIOUS EDITIONS

Stay ahead of the curve with the latest in Web3 culture and innovation. Subscribe to Hashed Out for exclusive insights, case studies, and deep dives into the decentralized future.

Help Grow Hashed Out And Get Rewarded With Premium Content & Merchandise

If you believe in a more open, fair internet — help us build it, one reader at a time.

Web3 adoption starts with curiosity. Share Hashed Out with someone who’s ready to explore.

You’re not just sharing a newsletter — you’re inviting someone into the future of digital life.

Refer 3 friends and unlock premium content. The more you share the more rewards you unlock, including Hashed Out mugs or tote bags, and exclusive community memberships.