For years, AI assistants have been good at answering questions.

Now they are starting to do something different:

Taking action.

Instead of asking AI what hotel to book, what email to send, or what deal to negotiate, users can build agents that actually handle those tasks.

One experiment showed where this is heading.

An AI agent was given one simple mission:

Call hotels.

Ask for availability.

Negotiate a better price.

And do it without a human ever picking up the phone.

The agent was given basic instructions:

Travel dates. Budget preferences. Questions to ask. A goal to negotiate.

Then it started calling hotels across the US.

During one conversation, something unexpected happened.

The hotel employee interrupted and asked:

“Are you an AI?”

The agent had been discovered.

But instead of failing, it continued the conversation.

It confirmed the booking details, asked about the price, and negotiated.

The original rate:

$108 per night.

The final offer:

$103.

A small discount.

But the important part was not the $5.

It was that an AI agent successfully handled a real-world conversation where the outcome depended on persuasion.

The system was created using AI development tools and voice-agent platforms.

The setup included:

• A simple interface for travel details

• A location tool to find hotels and contact information

• A voice AI system to make calls

• A language model to improve conversation quality

The first version took only a few days to build.

The bigger challenge was refinement.

The AI had to learn how to sound natural.

Early versions were too formal.

Instead of saying:

“I would like to inquire whether there are available discounts.”

The improved version sounded more human:

“I was wondering if there’s any flexibility on that rate?”

Small prompt changes created big improvements.

The biggest challenge was behavior.

The agent had to learn:

When to negotiate.

When to stop.

How aggressively to push.

How to handle unexpected responses.

Early versions talked too much, asked for discounts too quickly, and struggled when negotiations failed.

The lesson:

Building AI agents is not just about giving them access to tools.

It is about designing the right decision-making process.

Traditional AI:

Answer a question.

AI agents:

Complete a task.

The difference is huge.

A chatbot might tell you:

“Here are some hotels under your budget.”

An agent can:

Search options. Call businesses. Compare prices. Negotiate. Report back.

The AI moves from information provider to digital worker.

Saving $5 on a hotel room is not the breakthrough.

The breakthrough is that many small tasks that humans avoid because they are annoying can now be delegated.

Calling customer support.

Negotiating bills.

Scheduling appointments.

Researching options.

Comparing services.

These are not difficult tasks.

They are just time-consuming.

And that is exactly where AI agents are starting to fit.

The interesting part is that you no longer need to be a traditional programmer to create these systems.

With tools like AI coding assistants and agent platforms, people can describe what they want and gradually build working applications.

The barrier is shifting.

Before:

You needed to know how to code.

Now:

You need to know what problem to solve and how to guide AI.

The next phase of AI is not just smarter models.

It is AI that can interact with the real world.

The biggest change will happen when people stop asking:

“What can AI tell me?”

And start asking:

“What can I delegate to AI?”

Because once AI agents can call, negotiate, buy, schedule, and execute tasks on our behalf, they stop being tools.

They become digital employees.

Perplexity is moving AI processing closer to users’ devices with a new hybrid system for its Personal Computer AI agent. Starting in July, the system will automatically decide which parts of a task should run locally on a user’s device and which should be handled by larger cloud-based AI models. This allows sensitive information like personal files or financial data to stay on-device while more complex tasks use powerful servers.

The goal is to create a smarter balance between privacy, speed, and cost. Perplexity says the technology can turn personal devices into mini AI data centers, reducing reliance on expensive cloud computing while still delivering advanced AI capabilities. The system is currently available through the Mac app, with Windows support and compatibility with hardware like Intel and Nvidia platforms coming as the company expands the technology.

A hospital in Milan is testing Alter-Ego, an AI-powered robot designed to support healthcare workers by handling basic tasks and interacting with patients. The 1.2-meter robot can deliver items, guide patients, collect health information, and act as a remote presence for doctors. During trials with ALS patients, the robot has been used to monitor pain levels and send updates directly to nurses.

The goal is not to replace medical staff but to reduce repetitive workloads and give caregivers more time for human interaction. Researchers believe robots like Alter-Ego could eventually assist patients at home, although tasks involving medical decisions, such as administering medication, will remain under human control. The project highlights how AI robotics is moving from labs into real-world healthcare environments.

Unscreen removes backgrounds from videos without needing a green screen. It is useful for creators, teachers, marketers, and social media editors.

You can use it for short videos, presentations, reels, and creative projects.

​Try Unscreen

That’s it for today.

AI is moving fast - models are getting better, tools are getting cheaper, and the gap between “people who use AI” and “people who don’t” keeps widening.

The only real advantage left is speed of learning.

Until next time: stay AI smart, stay ahead, and keep building with the future instead of reacting to it.