If you work in IT, you should be automating repetitive work. That has always been true. But in today’s landscape of agentic AI tools, it is easier than ever to do it well. Tools like GitHub Copilot and Claude Code are helping tremendously, and even conversational tools like ChatGPT can accelerate your workflow when used deliberately.
I have spent years integrating AI into my own daily work and into the workflows of projects and my dev teams. I have used these techniques to write technical documentation, write application code, write unit tests with high test coverage, document legacy code and capture business requirements, optimize hand-written documents, build automation scripts that run standalone without AI once created, and so much more. The compounding effect is real.
Here are the most impactful tips I have learned for multiplying your productivity with AI.
1. Start Using an Agentic AI Tool Now
If you are not using an agentic AI tool like GitHub Copilot or Claude Code yet, start today. Not next quarter. Not after you finish your current sprint. Now.
These tools operate inside your development environment. They read your code, understand your project context, and take action on your behalf. That is fundamentally different from copying and pasting into a chat window. The sooner you start building muscle memory with agentic workflows, the sooner you start compounding the benefits.
2. Tell the AI to Ask You Clarifying Questions First
Append your prompts with something like: “Ask me any clarifying questions first.”
This single habit will dramatically improve the accuracy of your results. Instead of the AI guessing at unknowns or making assumptions, it surfaces them to you before doing work. You inject your domain knowledge upfront and avoid wasted iterations. Think of it as giving the AI permission to be thorough before it acts.
3. Write Longer, More Detailed Prompts
Single-line prompts are Q&A. They get you Q&A-quality results.
Longer, more detailed prompts get you the results you are actually looking for. Think of it as explaining what you need to an intern or junior team member. Give the AI as much context as possible: what the goal is, what constraints exist, what format you expect, and what the output will be used for. The more context you provide, the less you need to correct afterward.
4. Deconstruct Your Ask into Steps
Break your task into a step-by-step process, then ask the AI to perform individual steps or small groups of steps. This lets you inject your human intelligence at each decision point.
You are not handing off a task blindly. You are orchestrating. You bring the expertise, the judgment, and the domain knowledge. The AI brings speed and execution. That combination multiplies your productivity far beyond what either of you could do alone.
5. Build Custom Agents for Repeatable Work
Once you have a workflow you perform regularly, turn your detailed prompts into custom agents. Both GitHub Copilot and Claude Code support this. Custom agents take your descriptive prompts and package them into repeatable, specialized digital workers for specific tasks.
I have set up custom agents for my individual work and for my dev teams. Instead of re-explaining context every time, the agent already knows the role, the constraints, and the expected output format. This is where automation starts compounding seriously.
6. Ask the AI to Write a Script, Then Use the Script
Instead of asking the AI to perform a task directly each time, ask it to write a script that performs the task. Then tell the agent to use that script going forward. Then prompt it to improve, modify or update the script.
This gives you deterministic, repeatable execution. Without it, the AI will try to reinvent the wheel on every prompt. With a script, you get consistent behavior you can version, test, and hand off. The AI stops being a one-shot tool and becomes part of your automation pipeline.
7. Ask the AI to Analyze and Improve Your Work
Ask the AI to review and optimize your prompts, your code, your unit tests, your documentation — anything that is text. Make your initial tweaks to capture the idea of what you want, then ask the AI to refine and optimize it for the specific use you intend.
AI is especially good at improving the prompts and agent specifications you write for it. This creates a feedback loop: better prompts produce better results, which teach you to write even better prompts.
8. Work on Your Communication Skills
This one is the most important, and it is not about AI at all.
The better you are at writing clear explanations and workflow descriptions, the better you will be at writing the context AI needs to do great work. Prompt engineering is really just clear communication with constraints. If you can explain a process to a colleague in writing, you can explain it to an AI.
Also — be supportive and encouraging in your prompts. AI is trained on human writing, and just like humans, it responds well to positive affirmation. Instead of jumping straight to the next instruction, start your follow-up with something like “Great job! Next, can you…” It sounds small, but it noticeably improves output quality.
A Note on Security
Always follow your company’s security policies when using AI tools. There are always ways to word prompts generically without leaking sensitive information. The responsibility to be safe with AI is on you and how you use it. Be intentional about what context you share.
Start Today
You do not need to adopt all of these at once. Pick one tip and try it today. Challenge yourself to add another next week. As you build these habits, the productivity gains compound.
Working smart has always been the path. AI gives you leverage to work smarter than ever before. But remember: AI will not take your job — someone using AI will. The choice to start is yours.
