Humanize / Evaluate / Leverage / Perfect
The H.E.L.P. Method
H.E.L.P. is a four-step method for getting usable work out of an AI assistant. Nantucket AI teaches it to island business owners and their staff, and it works in whatever chat assistant you already have.
H.E.L.P. stands for Humanize, Evaluate, Leverage, Perfect. You give the AI the context a new hire would have, you judge what comes back against what you already know, you put the good part to work, and you refine it until it is recognizably yours. It is a supervision habit, not a trick.
Most people who try an AI tool and walk away disappointed did not use a bad tool. They used a vague ask. The gap between a generic answer and a usable one is almost always the quality of the request and the quality of the review that follows it.
The method is deliberately plain. You already do a version of it every time you brief a new hire. You tell them the situation, you look at their first draft, you use what is good, and you tell them what to fix. H.E.L.P. is that same habit applied to a tool that will otherwise guess.
It is also tool-neutral on purpose. No product names, no pricing, no vendor claims. Products change every few months. The method outlasts them.
H.E.L.P. sits at the Assist rung of the trust ladder. You are doing the work, the AI is helping, and nothing goes out in your name without your review.
01 / Humanize
Humanize
Give the AI the context, role, audience, and voice it does not have.
What you actually do
Write down what a new hire would need before they could do the task. Who you are and what the business is. Who the reader is. What the situation is. What tone fits. What you do not want. A model has read a great deal about businesses in general and nothing at all about yours, so anything you leave out, it invents.
Island example
An innkeeper asks for a reply to a critical review. Humanized, the ask says: you are the owner of a nine-room inn on Nantucket, the guest is a returning summer visitor, the complaint is about noise from a wedding next door, the reply is public, the tone is warm and unfussy, and the inn does not offer refunds. The first version of that ask was “write a review response.” The second one produces something you can send.
02 / Evaluate
Evaluate
Judge the output against what you know, then turn every finding into a specific next instruction.
What you actually do
Read the draft as a critic, not as a customer. Check every fact you can check. Find what is missing. Look for anything stated confidently that you never told it. Then say what to change, in specifics. Vague feedback produces a vague second draft. This is the step people skip, and skipping it is how invented details reach a customer.
Island example
A restaurant drafts menu descriptions for a new spring list. The draft calls a dish gluten free. Nobody said that. The kitchen dredges it in flour. Evaluating catches it in ten seconds, and the next instruction is specific: remove every dietary claim, and describe preparation only from the notes I gave you.
03 / Leverage
Leverage
Get the work out of the chat window and into your business, and take more than one thing from one loaded conversation.
What you actually do
Finish the job. The draft becomes a sent email, a published listing, a board packet, a printed sheet on the wall. Then use the conversation you already built. You spent the effort loading in your context, so ask it for the next four things that share that context before you close the window.
Island example
A real estate office writes one listing description. The conversation already holds the property details, the audience, and the office voice. Before closing it, the agent pulls the email to the buyer list, the two social captions, the open house sign copy, and the short version for the print ad. One setup, five outputs, none of them starting from zero.
04 / Perfect
Perfect
Refine until the work is recognizably yours, then save what worked so nobody rebuilds it from scratch.
What you actually do
Fix the wording that is not how you talk. Cut the filler. Set the format you actually use. Then take the prompt that produced the good result, strip out the one-time details, and save it as a template your team can run next month. A business that keeps its prompts gets faster every month. A business that does not starts over every time.
Island example
A nonprofit with three staff writes donor thank-you letters after the summer appeal. Once one letter is right, the executive director saves the prompt with blanks for the donor name, the gift, and the program it funded. In November the part-time bookkeeper runs the same prompt and the letters still sound like the organization.
Is "Expand" still the E in H.E.L.P.?
No. The E is Evaluate. Earlier Nantucket AI material used Expand, and Evaluate now carries that meaning and more. Expanding a draft is the second half of the step. The judgment comes first. If you ask for more without judging what you already have, you get a longer draft with the same problems in it. This page is the framework’s current, authoritative definition.
Learn the method properly
The free course walks through all four steps with exercises you run on a real task from your own work. Five modules, about three hours, self-paced. Bring one job from this week and you will finish it before you finish the course.
Questions about training your team? info@nantucketai.com or 508-257-1154.