Agentic AI: Shifting from Knowledge to Action

By now, you’ve likely heard the term “agentic AI” or “agentic AI workflows” or simply “AI agents,” which are all phrases that are referring to a type of AI system that’s able to work through complex, multi-step workflows autonomously.

Whereas current LLM capabilities are indeed stunning and game-changing in their own right, they’re limited in what they can do, because what they can do is contingent on what we tell them to do.

Right now, we’re prompting LLMs to do very specific things for very specific uses cases, but soon, it seems, we’ll be entering the early innings of AI agents.

Agentic systems have been around for years, but as McKinsey writes in a recent piece called “Why agents are the next frontier of generative AI,” they’ve been fairly “difficult to implement, requiring laborious, rule-based programming or highly specific training of machine-learning models.”

What they mean is that there were often so many variables in a particular workflow that it became difficult to write rules for every scenario.

You had to be able to see around corners, in a sense, and it was incredibly difficult to do.

However, now that LLMs have the ability to perform complex reasoning, and since frontier models have become increasingly powerful and prolific, agentic AI is now a real possibility.

“When agentic systems are built using foundation models (which have been trained on extremely large and varied unstructured data sets) rather than predefined rules,” the authors write, “they have the potential to adapt to different scenarios in the same way that LLMs can respond intelligibly to prompts on which they have not been explicitly trained.”

This is an incredible breakthrough with potentially industry-changing consequences, and it’s worth exploring a bit more comprehensively.

The potential of agentic AI

An easy way to understand how an agentic AI system might work is to view it through the lens of booking travel.

Imagine using a natural language prompt—“Check my schedule for September and October and book me a trip to Europe. Include dinner reservations at top restaurants and some cultural experiences.”— and watching while an AI assistant seamlessly integrates with your calendar, assesses your availability, and autonomously designs an itinerary.

From selecting destinations and booking flights to securing dinner reservations at renowned restaurants, the AI handles every detail. The agentic AI consults with you at critical junctures, too, ensuring that the proposed plans align with your preferences, and uses your feedback to make adjustments as needed.

It’s not hard to see how incredibly convenient and useful this would be for scenarios like this one, how an agentic AI focused on travel would not only simplify the planning process, but also elevate the overall experience, giving you a more personalized and hassle-free journey.

Now imagine what it could do for your business.

Whether you’re a corporate accountant or a marketer inside of an industrial company or a CEO, on a fundamental level, your job is a bundle of individual tasks that you complete each and every day.

Many of those tasks are repetitive and time-consuming and keeping you from engaging deeply with the work that truly matters, the work that moves both you and the organization forward.

But what if you had a virtual coworker that could complete complex workflows autonomously? What if you had an agentic AI that could handle a good portion of that less valuable work for you? What would it mean for your business if you could spend the vast majority of your time on strategic, high-value work?

My guess is it would unlock a series of insights that would set the direction for your organization’s future.

A marketing use case

McKinsey has already done the hard work of outlining three illustrative use cases for AI agents, so rather that reinvent the wheel, I’ve included the marketing use case study in its entirety here.

Keep in mind that while an agentic AI system is performing work autonomously in this example, there’s still a human-in-the-loop control mechanism to ensure success.

The idea here, even with agentic AI systems, is that humans retain the ability to set the initial direction, and change it when needed.

***

Use case 3: Online marketing campaign creation

Designing, launching, and running an online marketing campaign tends to involve an array of different software tools, applications, and platforms. And the workflow for an online marketing campaign is highly complex. Business objectives and market trends must be translated into creative campaign ideas. Written and visual material must be created and customized for different segments and geographies. Campaigns must be tested with user groups across various platforms. To accomplish these tasks, marketing teams often use different forms of software and must move outputs from one tool to another, which is often tedious and time-consuming.

Potential agent-based solution: Agents can help connect this digital marketing ecosystem. For example, a marketer could describe targeted users, initial ideas, intended channels, and other parameters in natural language. Then, an agent system—with assistance from marketing professionals—would help develop, test, and iterate different campaign ideas. A digital marketing strategy agent could tap online surveys, analytics from customer relationship management solutions, and other market research platforms aimed at gathering insights to craft strategies using multimodal foundation models. Agents for content marketing, copywriting, and design could then build tailored content, which a human evaluator would review for brand alignment. These agents would collaborate to iterate and refine outputs and align toward an approach that optimizes the campaign’s impact while minimizing brand risk.

***

Work has already changed

McKinsey’s example gives us a good sense of how an agentic AI system could work, but let’s take it even further.

Imagine an AI system that manages the entire heavy-equipment manufacturing supply chain, from raw material procurement to final product delivery. It could adapt to supplier issues, demand fluctuations, and even geopolitical events affecting the supply chain.

Or imagine an AI agent that orchestrates marketing campaigns across multiple channels (trade shows, industry publications, social media, email) for industrial clients. It could autonomously adjust messaging and budget allocation based on real-time performance metrics.

I’m sure with a little bit of thought, you could imagine how something like this could work in your organization.

Opportunities abound, and those opportunities will likely be here before we know it.

Sure, it’s still possible that we’re in an AI bubble and federal regulation or GPU production or energy costs will pose a barrier that we simply can’t cross, but even if that’s the case, AI—and specifically, LLMs—have already changed how we work in important and significant ways.

If you’ve been waiting to see how things will shake out with AI, it’s time to stop waiting and act. Create an AI Council, an AI policy, and an AI use case, and start experimenting.

You can proceed slowly, but you need to start building your AI muscle now, because we’re getting to the point (if we’re not already there) where organizations are going to start feeling the financial effects of their reticence to act.

It’s time to get to it.