This as-told-to essay is based on a conversation with Justin Parnell, 39, an AI consultant based in San Francisco. This piece has been edited for length and clarity.
As an AI consultant, I automate cumbersome parts of my workflow to save me hours and make me far more efficient.
Through my business, Justin GPT, I build custom AI agents for clients who want to do the same.
I advise every client to follow these five simple steps for using AI agents to lighten their workload. (My advice is model-agnostic. It doesn’t matter whether you use ChatGPT, Gemini, or CoPilot).
1) Understand how AI works
Keep in mind what AI actually does as you think about the job you’re trying to get done. If you understand AI’s data, objectives, and constraints, you understand how it works.
AI aligns the context you give it with the information it has been trained on for next-token prediction.
Instead of viewing AI agents as autonomous digital workers, break it down. First principles thinking says: “An AI agent is simply a model that receives input, follows defined goals and rules, makes step-by-step decisions, and uses tools to take actions.”
2) Break down the job into its atomic parts
Separate the task into the smaller steps you perform within it, then ask yourself whether AI can handle them. If the answer is yes, consider how valuable it would be to have an AI agent take over those tasks.
For example, if it’s admin work, ask questions like: what steps are involved in the new client intake process? What goes into building a proposal? What steps are required to send a contract for signature and officially onboard a new client?
Breaking those processes into atomic parts might involve, for example, logging into your accounting software, adding a new client, and sending an invoice.
3) Prioritize the atomic tasks
Ask yourself: is this task high or low impact? High or low effort?
Prioritize high-impact, low-effort tasks, then high-impact, high-effort tasks, followed by low-impact, low-effort tasks. Low-impact, high-effort is the lowest priority.
As you tick those tasks off in order of prioritization, your workflow quickly becomes AI-empowered.
4) Put yourself in the workflow
There will always be points in the workflow where a human needs to step in. Before an invoice is sent, I review it. Before a proposal is approved, I get a Slack notification to check it to see if it needs editing.
Design yourself into the workflow to make sure AI hasn’t done anything you wouldn’t do.
Explaining to AI that it has made a mistake and how not to repeat the error helps it improve.
I recommend building this step into any end-to-end workflow.
5) Get going!
Once you’ve completed the first four steps, you are ready to start integrating AI agents into your workflow.
You’re off to the races, and you can start building and seeing real efficiency gains.
