From Co-Pilot to Auto-Pilot: AI Agents Are Taking Over

An article in Business Insider, which reports that 26% of leaders are seriously looking into AI Agents, made me quickly jump to our Lead with AI Community chat. 

“If we want to be ahead of the curve and lead others, implementing agentic workflows is a must. It’s so easy to get started with Tasks and Deep Research now embedded in every ChatGPT plan,” I shared with the group.

Even for free newsletter readers, you’re among a select group of leaders taking AI seriously. 

But reading, learning, and talking are not enough – we must move to action. 

AI Agents: The Insane Implications

AI agents are AI-powered software that can perform tasks autonomously or with minimal supervision.

Unlike simple chatbots, which only respond to queries, modern agents can act, make decisions, and coordinate steps to achieve goals. 

For example, a traditional AI assistant might help draft an email, but an agent could draft the email and then send it or even book meetings discussed within – all without constant user input.

Just look at what Salesforce’s Einstein SDR is already capable of doing:

I see it as a shift from having AI work with you, to making AI work for you.

In the Business Insider article, ServiceNow reports that AI agents resolve 80% of their customer support cases without human help and help humans close complex cases 52% faster. 

They do this by breaking down complex objectives into sub-tasks, iterating on solutions, and adjusting based on feedback or new information, a significant improvement over earlier AI systems.

And not just within the four walls of ChatGPT’s chatbox. 

For example, Cohere’s North combines language models with multimodal search, allowing an AI agent to extract information from images, PDFs, and spreadsheets.

And after we were wowed by ChatGPT Operator, this week’s ‘talk of the town’ is Chinese AI startup Manus AI, which can break out of the chatbox and unzip files, read PDFs, and create reports.

But this is just the start.

As Sam Altman said in a recent blog post:

“Imagine [AI] as a real-but-relatively-junior virtual coworker. Now imagine 1,000 of them. Or 1 million of them. Now imagine such agents in every field of knowledge work.

Anyone in 2035 should be able to marshall the intellectual capacity equivalent to everyone in 2025; everyone should have access to unlimited genius to direct however they can imagine.”

All that’s left is figuring out where to “point” all the genius Sam Altman mentions. 

What will your AI organization do? 

The One-Person Unicorn, Revisited 

I wrote in my 2024 AI predictions that we may soon see a one-person unicorn.

Looking at the references above, this should start to feel less and less hypothetical.

If an AI can autonomously figure out how to best do something, work with other agents, and direct itself, then scaled to 1,000s of them, a simple idea could turn into a billion-dollar company. 

But not just anyone will create that unicorn, I was reminded when I heard Loveable founder Anton Osika on Lenny’s Podcast. 

Loveable lets a user describe an app in plain language and have the AI instantly generate a working web application – effectively coding, testing, and deploying on the fly. 

On the podcast, the pair discussed that it’s becoming less about how to build something, more about what to build and why. In other words, the role of “Product Manager” may be a better lens to look at the future of work. 

‍

The emphasis is on guiding resources rather than personally executing every step and becoming a strategic director for hybrid teams of humans and AI agents.

To tap into the full benefits of AI agents, we all need to become AI orchestrators and set up the proper prompts, tools, and checkpoints. 

We need to know what they’re capable of and get them going, occasionally providing the strategic context that AI lacks. 

As The AI Daily Brief’s Nathaniel Whittemore explained in a terrific video essay, this will eventually evolve into the equivalent of Dr. Strange scanning through many multiverses.

Steering an AI agent organization like this, with thousands or millions of agents each better than the foremost human expert in their field – the one-person unicorn doesn’t seem so futuristic anymore. 

Getting on the Train Now

These developments and the far-reaching implications of truly agentic organizations create even more urgency for us all to get on the train.

Even if you don’t have any ambition to be the founder of a one-person billion dollar unicorn, you’ll want to be at least a SuperWorker, an “an individual who uses AI to dramatically enhance their productivity, performance, and creativity.”

In other words, someone who takes that ‘genius laser’ Sam Altman describes and points it at every objective and challenge in their role.

Here’s what you can do today:

• Determine which elements of your role are General, Error-Friendly, Digital, Recurring, and Toil. These “GED-RT” tasks are prime for AI to take over.
• Implement AI tools that work as a co-pilot or autopilot, with you or for you. From off-the-shelf products like Copilot for PowerPoint or Gamma AI to self-built AI automations and AI agents like Deep Research, Loveable, Operator or Manus. (See Build Your AI Team if you need help or check out these other AI courses.)
• Personalize and fine-tune each tool. Gamma and other AI websites are great, but works even better with your brand assets. ChatGPT and other AI models are excellent, but works way better when you program custom-trained GPTs. 

As Loveable’s founder said: “being extremely proficient with AI tools is now a superpower.”  Being in the top 1% or even 10% of AI tool users will dramatically set you apart. 

And that’s how you get to lead in the future of work.

Go!