Anyone that has recently watched Jensen Huang’s keynote at CES or at Nvidia’s GTC is aware that “Agentic AI” is the next phase of the AI revolution. We are currently in the GenAI phase, and the road is being paved for Agentic AI. In other words – Agents will be everywhere, disrupting the lives of 1 billion knowledge workers globally.
A stinging quote by Jensen was his assertion that “The IT Department is going to be the HR department of AI Agents in the future”. I heard that and I started to visualize his point more clearly. Remember, Nvidia is now arguably the most valuable technology company on earth, so we better take when they sketch out plans.
Video Summary
Subscribe to The Cyber PMM Podcast on YouTube, Spotify, and Apple Podcasts.
A few other Jensen quotes from his CES keynote:
- “There are 30 million software engineers around the world. Everybody is going to have a software assistant helping them code. If not, you’ll be way less productive.”
- “There are a billion knowledge workers around the world. AI Agents is a multi-trillion dollar opportunity.”
- “AI Agents are the new digital workforce.”
- “The age of Agentic AI is here, for every organization.”
I watched a few videos after Jensen’s CES keynote and made the connection. I downloaded Cursor, which is an AI coding platform, and realized that even I, a lowly Cyber PMM, has the ability to build an AI Agent. Sure, it’s harder than the video made it seem, but I’m giving it a try anyway. I just wanted to test out the theory and see if it really is possible to build software with zero coding experience. Not only is it possible, it’s actually pretty simple and doesn’t require a lot of time.
I first started dabbling with ChatGPTs last year when I deployed the Cyber PMM GPT. That’s when I realized that there is definitely a day coming when AI will automate my job’s more routine tasks, and eventually the more complex stuff.
AI Agents however would provide an ability to embed LLMs into your own software, which is different from GPTs. The truth is that I don’t even know how to code to build an AI agent. It’s all natural language based using Cursor to code everything for me. If you haven’t tested Cursor, go ahead and give it a try. I’m not a developer but I can see why so many developers are worried and why the job market has become so saturated for software engineers. It’s never been easier to build software or apps. And Cursor can pull in OpenAI’s developer tools easily, so you can build apps with GenAI capabilities easily.
And AI Agents will be the next wave of software used by businesses to expedite and automate key processes. Product Marketing will be no exception. The tasks Cyber PMMs execute could definitely be replaced by an AI Agent, there’s no question. Think about a typical marketing organization. Imagine a world where a single operator using AI Agents executes and automates entire marketing programs 24/7 without the need for massive headcount. To illustrate, look at this image from the CES keynote. Imagine building a company with only a few employees, and the rest of your team are AI Agents.
Looking at this image I began to think, why can’t Product Marketing be included? It will be.
My opinion is that we should master AI before it masters us. Let’s try to build a functional agent and see if it works. I would be the first to say, “yep, this is better than me – you don’t need me anymore.” And then I move on with my life to a new career or some other challenge to solve.
I have been amazed at how good GenAI is at executing tasks such as messaging, message testing, buyer journeys, strategic planning, and more. Why is it not possible to create an agent that fully replaces me yet? There are a few barriers I see today.
- First, there still needs to be a human in the loop to spot check for accuracy. Even if I just scan an output, I need to double check AI’s outputs. I’m not yet able to fully trust an AI agent to input a command, output a result, and publish that result to a website. It’s just too risky still. However, that risk is diminishing daily and I do think there will be a day why I can trust AI more completely.
- Second, many cyber vendors are prohibiting the use of LLMs on corporate networks due to security risks. Until AI Agents are developed by mainstream software providers and baked into existing SaaS apps, I don’t see this being adopted by PMM teams. It’s hard enough as it is to use technology or get budget, why would there suddenly be an opening for new AI Agents that aren’t tested?
So my prediction is that companies like SalesForce, Google and Microsoft will continue to improve their AI Agents within existing platforms, and that’s where we’ll see more adoption within Cyber vendors. Microsoft and Google will drive adoption by baking in or releasing new apps (look at Gemini and NotebookLM as an example) for enterprise use to drive adoption as well. And then as that becomes more common and less risky, PMMs will be able to first leverage AI on the job more often, increasing productivity. PMMs that don’t leverage those tools will be stuck in the manual-driven past, like riding bicycles in the early 1900’s as Ford’s started hitting the road.
And don’t sleep on Nvidia AI’s software ambitions. They have loads of cash, and are working on usability by building the ability to run an AI Assistant locally on your machine. So if the machine you buy comes with Nvidia GPUs, you will be able to leverage their software as well. Nvidia is expanding bigtime from just being known as a chip manufacturer to a leading AI software provider.
For the near more realistic term, I see AI agency happening by opening a Google Doc to start drafting a messaging brief. Suddenly you click on a new Gemini icon and begin prompting an LLM to support your task, fully supporting you as an agent within the content production process. This is where I see agents coming into reality at work. I expect that to happen this year. Google is in full focus on this effort, and Microsoft already has the OpenAI stake, so you can expect enterprises to begin feeling the impact soon.
And look at this example from Nvidia – they demoed the ability to upload a PDF and create a podcast. Are you telling me we’re far away from uploading an AI-generated messaging brief as a PMM, and generating an entire content waterfall?
Startup technologies are often slow to be integrated into enterprise environments. I hope that we can see more innovation in product marketing, and that one day I can have an AI Agent that I can task while on the job. I’m somewhat dreading the fact that this will be done through Microsoft or Google, because I’d like to see someone take on this challenge with a focus on Product Marketing.
So to answer my original question, Will AI Agents replace PMMs? I think the answer is yes and no.
- Yes, because we will see adoption of AI capabilities within our existing SaaS applications at work. We will be more capable of higher quality outputs such as text to video, making us less reliant on Creative and agency outsourcing, speeding up our production processes.
- No, because we still need to be the human in the loop and vet outputs and monitor activities. Nothing happens overnight. In five years, I can see Agentic AI being dominant, and a common factor that people say, “of course I use an agent!” In 2025, I’m thinking we see agents appear in existing SaaS applications on the job, vs. new brands being adopted.
This area is ripe for innovation though and I’ll be watching closely. I don’t think that Google or Microsoft will be releasing AI Agents focused on Product Marketing any time soon. Perhaps this is an opportunity to build a new future for Cyber PMMs. PMMs that embrace AI Agents will lead the way forward.
What do you think?
Stay tuned for future posts by subscribing to the newsletter on LinkedIn or Substack. You can also subscribe directly below. Please also follow Cyber PMM and/or me on LinkedIn.
