The Rise of Agentic AI
How AI tools that perform tasks are changing everyday writing, document work, and file organization.
In this essay
A Different Kind of Assistant
For the past few years, working with AI usually meant asking a question and receiving an answer. The response could be useful, but I still had to move it into a document, organize the files, or carry out the next step myself.
The newer tools can handle more of that work within an authorized workspace.
They can open files, search, organize material, write documents, and connect several steps in a task. A request can now produce a saved draft or an updated file, with the work available for review.
Using these tools has changed what I expect from an assistant.
What Is Agentic AI?
An agentic assistant can use tools within a workspace you authorize, carry out a sequence of actions, and inspect the results before continuing.
For example, I can ask it to read existing blog posts, follow their formatting, and save a new draft in the same structure. It studies the files and produces something I can open and review, saving the transfer and formatting work that used to fall to me.
The useful part is that cycle of planning, acting, checking the result, and adjusting.
From Answers to Actions
A standard chat assistant usually leaves me with a response to interpret and put to use.
An assistant with access to tools can also:
- Read and write files on your computer
- Execute code and use the results to inform the next step
- Work through multi-step tasks without waiting for you to relay information between steps
- Use tools like web search, spreadsheets, or document editors as part of a single workflow
That reduces the distance between describing a task and having a usable result. The benefit becomes more apparent when a job involves several small, repetitive steps.
How I Use It
Most of my use has been writing and content work. A few examples stand out.
Blog post creation. I asked the assistant to study the frontmatter and conventions in existing posts, then create a new post in the same format. It read several files and saved the draft. Work that might have taken me an hour took a few minutes to prepare for review.
Document drafts. With reference material available, it can produce a structured draft in the format I need. Having the sources and the intended output in the same workspace makes the result more useful.
File organization. Renaming, reorganizing, and consolidating files are natural tasks for these tools. I already make a habit of keeping files organized, and this makes the routine maintenance easier.
I still review everything. Errors happen, but the amount of preparation I have to do myself has fallen considerably.
What This Changes
More knowledge work can now begin with a goal and proceed through several steps in real software. The output is something a person can inspect, revise, and use.
That has practical consequences for both individuals and organizations.
“AI won’t take your job. It’s somebody using AI who will take your job.”
— Steve Baldwin, economist and professor at the IMD Business School
For an individual, a smaller gap between a plan and a usable draft means more time to evaluate the work. People who learn to direct these tools well can move through familiar tasks much faster.
Organizations still need people who understand the purpose of the work and can judge its quality. Vision, strategy, and experience matter when deciding what to ask for and whether the result is any good.
The tools are developing quickly. I expect capabilities that feel experimental now to become an ordinary part of work.
Getting Started
Begin with a small task in a workspace you are comfortable authorizing. Read the documentation so you understand what the assistant can access and which controls are available.
Give it a specific task. Asking it to read three recent posts and draft another with matching frontmatter gives it a clearer assignment than asking for help with a blog.
Review the result before publishing or sending it. A capable tool can still misunderstand an instruction or make an error that looks convincing.
Ask it to explain what it did. A visible account of the work helps me catch misunderstandings and improve the next request.
Use it for a complete sequence of steps. Much of the value comes from carrying a task through the reading, drafting, formatting, and saving that surround the central work.
Reusable Instructions
I get more consistent results when I provide examples and refine the instructions as I learn what the tool needs. Conventions that matter to me are worth writing down.
- Separate reference material from the main instructions.
- Keep each instruction set focused on a specific task.
- Review generated instructions for accuracy and unnecessary detail.
- Test the workflow on a small example before relying on it.
Where This Is Going
Agentic tools are becoming a more substantial part of how people use AI. Their usefulness will depend on how well they fit the work and how clearly people can direct them.
For me, the best arrangement is a defined workspace, a clear assignment, and results I can inspect before they go further.
That gives me useful assistance while keeping the decisions in my hands.
This is the first version of the technology that has meaningfully changed how I get work done.