HOW TO BUILD AN AGENTIC AI
It looks like good news for content writers and specialists.
Because you're going to help build agentic AI.
HERE’S THE BASIC THREE STEP PROCESS
Building an agentic AI can be broken down into a three-step process.
Who’s involved?
Humans.
Usually a collaboration of subject matter experts, content specialists, and prompt engineers.1
An agentic doesn’t get built from scratch.
It’s something with a lot of moving parts.
You put them together, and train a system to use (and ‘learn’ from) the different components.
Here’s a simple example using a human resources (HR) agent designed to screen job candidates:
One: Define the Purpose
First, decide what you want it to go. Give the AI a clear, specific mission.
Start with the end in mind, as they say.
Know your business rationale, goals, and parameters.1
Let’s say, your mission is to "find the three most qualified candidates for an open marketing position."
Two: Assemble the "Engine" and "Tools"
To begin, work out which core "reasoning engine," your AI will be built on.
You can use a large language model (LLM) like GPT or Claude.2 This model is the digital brain.
It’ll analyse information and use logic to make decisions.
Next, think about which "tools" to work with.
This is where it gets interesting as far as privacy and data leaks are concerned. Because the tools are connections to other software or databases.2
In this instance, your HR agent would need tools such as:
access to the company's job application database
the email system
a scheduling app.1
Three: Tuning and Training
Now it gets serious.
You need a human. A subject matter expert, like a senior recruiter. This person works with the system to "tune" and train it.
It’s up to them to feed the system’s ‘brain’ with examples of good resumes and bad ones.1
So the agent can:
Perceive: The AI uses its "perception" to read and analyse the contents of all the job applications. It should search for keywords and qualifications.2
Plan: After that, it should plan a series of actions based on its mission. First - sort candidates by resume quality. Second - score the top ones. Third - send out interview requests.1
Act: Using the tools it has access to, the AI goes into action. It could send automated rejection emails to candidates who don't meet the criteria. And follow that up with an email to the top three candidates to schedule an interview.1
When all the moving parts are working together it becomes an AI agent.
The term "agentic" just means it doesn’t need a human to prompt it for each step.
Once you set the goal, it just completes the project.
internal logic to move from one action to the next, just like an assistant would on a project.3
After the interviews, a human does more fine tuning. They give feedback and the AI agent improves performance based on this..2
HOW THE CONTENT SPECIALIST HELPS BUILD AN AGENTIC AI
Your job kicks in during the tuning and training stage.
There are three distinct areas for you to work on.
Content strategy
This isn’t just about implementing the company voice and style guide.
You choose which types of information the AI needs to use.
In our HR AI agent example this means things like quality resumes, job descriptions, and outlining the scoring criteria based on info from the recruiting team.
And you’ll be part of the feedback system.
Content will need constant refining. It needs to be effective and targeted and you might use split testing and other methods to rewrite to a specific audience.
In other words, you get to be creative while the machine chops wood and carries water.
This is how, according to the latest reports, you can become part of the new Agentic AI-Human workforce 5.
References
1 Codeo. "What is a Neural Network? Breaking Down the Backbone of AI and Agentic Systems."
2 Codeo. "What is a Neural Network? Breaking Down the Backbone of AI and Agentic Systems."
3 IBM. "AI Agent Learning."
https://www.ibm.com/think/topics/ai-agent-learning
4 Monetizely. "How Are Graph Neural Networks Revolutionizing Agentic AI Through Relationship-Based Intelligence?"
5McKinsey. “The future of work is Agentic”
Amar, J., Weddle, B., Hancock, B., & Rahilly, L. (2025, June 3). The future of work is agentic [Podcast transcript]. McKinsey Talks Talent. Retrieved from https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-future-of-work-is-agentic
