From Brains to Bodies: The Rise of Agentic AI

It has been a few busy weeks since my last AI post, but I figured it was time to dive into what I personally believe will be fundamentally changing the way we work and leverage AI in 2025 - what is often referred to as "agentic" AI.

Let me start with a formal definition (bear with me for a moment ??):

Agentic AI refers to artificial intelligence systems that exhibit agency, meaning they can act autonomously to achieve specific goals, make decisions, and take actions within a defined environment or context. This involves a combination of capabilities, such as perceiving the environment, reasoning about it, and executing actions to fulfil objectives without continuous human intervention.

In plain English? We're talking about AI that doesn't just think - it does.

In my previous posts, we explored the rapid evolution of AI "brains" (Large Language Models) and the crucial role of context in unlocking their potential. With the brains and proper context in place, we are ready to explore what I believe is the most exciting development yet: AI's transformation from a powerful thinking engine into a system that can see, hear, speak, think, reason and—most importantly—take action in the real world!

First Came the Senses ?? ??

Just like humans need eyes and ears to perceive the world, AI has gained similar capabilities:

1. Vision: Modern AI can now:

• Analyze and understand images and video in remarkable detail
• Process documents, diagrams, and screenshots
• Understand and respond to visual context in real-time

2. Hearing: AI's audio capabilities have become incredibly sophisticated:

• Natural voice conversations that feel human-like
• Real-time speech recognition with unprecedented accuracy
• Voice synthesis that's nearly indistinguishable from human speech
• Multi-language processing and translation

So now we have:

• A powerful brain, equipped with the right knowledge and context ?
• Sensory capabilities so we can interact through text, voice, images and even video ?

But here's where it gets really interesting—and why I'm particularly excited about what's happening right now...

The Game-Changer: Taking Action ??

In the past 12 months, we've seen an explosion in AI's ability to interact with the world. First came the "tools" - APIs that allowed AI to search for information, generate code, and create visualizations. This opened up fascinating new possibilities, but last month's announcement from Anthropic took things to a whole new level: their AI model Claude can now "use computers." This might sound simple, but it is really changing the game completely. It means that AI systems are now gaining the ability to "see" your computer screen, understand what's happening, and take appropriate actions using the same tools we use every day - by moving the mouse on the screen, clicking and "typing" on the keyboard.

Let's draw a parallel to self-driving cars. Tesla's Autopilot system uses cameras (and to some extent sensors) to create a real-time representation of its surroundings, which then informs decisions about steering, braking, and acceleration. Sensing the environment, reasoning in lightning speed and the taking action in the real world!

What Can AI Actually Do Now? ??

Let's get concrete. Think about your typical workday and break down the process. Traditional automation works well for repetitive, structured tasks, but struggles with anything unstructured or creative. In our day-to-day we go through a number of "micro-tasks" and the outcome of one task may lead to a new scenario that we could not pre-determine. We need to think and adapt to the changing environment we operate in.

Even though the brains (LLMs) are trained on "old" data - we can equip them with tools to access the same information we humans can access today. Here are some examples from the top of my mind where AI shines in executing tasks today:

1. Research and Analysis

• Search through multiple web sources
• Read and summarize documents
• Extract key information from emails
• Compile findings into a structured report

2. Data Processing

• Access databases and spreadsheets
• Clean and analyze data
• Create visualizations
• Generate insights and recommendations

3. Communication and Coordination

• Draft and send emails
• Schedule meetings
• Follow up on pending items
• Coordinate with team members

4. System Integration

• Work across multiple applications
• Transfer data between systems
• Update records in your business systems
• Generate and validate reports

And rather than just executing one single item, it can now take the outcome of one to achieve the next - with no or limited human intervention...

The Rise of Agentic AI ??

So in a nutshell - "Agentic AI" means AI systems that can autonomously pursue goals through a combination of:

• Reasoning (the brain)
• Perception (the senses)
• Action (the limbs)

Think of it as giving AI a complete "body" to complement its "brain." These agents can:

1. Understand tasks through natural conversation
2. Break down complex goals into steps
3. Use appropriate tools and resources
4. Execute actions
5. Learn from feedback
6. Adapt to changing conditions

And here's what's really exciting: they can collaborate with other agents, each "trained" for specific tasks or responsibilities—much like how we organize our companies with specialists and different roles! Multi-agent systems will be a topic for a future post where I will also share concrete examples of frameworks and approaches that you can use today to get started!

Putting it all together ??

Here is an actual workflow that you can achieve today:

• Read a big report for me while I drink my morning coffee...
• Summarize the key findings
• Put it into a presentation format with nice visuals that is easy to digest
• Create a podcast episode of the same content for me to listen to while driving into the office
• Schedule a team meeting to discuss the findings
• Send an email to the relevant stakeholders with the summary as a pre-read to our scheduled meeting
• Update my CRM system with a draft newsletter based on this information
• Do not send it out until I have approved it
• Find a nice restaurant downtown and book a table for 2 and send a text message to my wife with the time an place

-> Human-in the loop for key decisions and approvals - 10x your own productivity!

The Path Forward ??

While these capabilities are exciting, they also raise important questions about:

• Security and access control
• Quality assurance and oversight
• Human-AI collaboration models
• Ethical considerations

But, I will leave it there for the now - until next time, I encourage you to think about:

1. What repetitive tasks in your work could benefit from AI that can actually execute, not just advise?
2. How might your team's workflow change with AI assistants that can take action?
3. What safeguards would you want to see in place before giving AI more autonomous capabilities?

Let's continue this conversation in the comments. I'm particularly interested in hearing about specific use cases where you see the potential for agentic AI in your industry.

#AgenticAI #FutureOfWork #AIEvolution #GenerativeAI #TechInnovation