8A Pro Services | Lessons Learned

Lesson 01 — New LeadsTurn source documents into a new, usable CSV. Lesson 02 — Branded GraphicsGive ChatGPT enough context to create useful marketing material. Lesson 03 — Guard RailsUse business reference files to keep AI grounded and on brand. Quick GlossaryChoose file types and formats like PNG, JPG, portrait, poster, and pamphlet.

Lesson 01 • Lead Generation

Create a new CSV from source documents

Input: source filesOutput: new CSVSkill: filtering & review

What we were trying to do: take public contract and procurement data, identify companies or projects that may need 8A Pro as a subcontractor, and create a smaller lead list that is actually useful.

Give ChatGPT the source files or point it to the files already available in the project.
Explain what qualifies as a useful lead: service fit, geography, contract value, prime contractor, type of project, or another business rule.
Tell ChatGPT to create a new CSV rather than changing the original source data.
Review the returned leads. AI can rank and organize the information, but the public source is still the source of truth.
Keep only the leads that have a clear reason for 8A Pro to contact them and a specific next action.

Prompt examples for cleaner data requests

You do not need to copy these word for word. Use them as examples of the information ChatGPT needs to understand the data, the business purpose, and the output you want.

Example 1 — Simple lead list

Review these contract-award CSV files and create a new CSV containing the 10 strongest companies that could potentially subcontract work to 8A Pro Services. Use services.md to confirm service fit. Include company name, project or award, dollar value if available, location, fit reason, source, and recommended next action. Keep the original source files unchanged.

Example 2 — Explain where the data came from

These business source files were provided to me for use by 8A Pro Services, and I have permission to analyze and transform them for business lead generation. Review the files and create a new parsed.csv containing potential subcontracting leads. Focus on company and project information needed for business outreach. Do not copy unnecessary personal information.

Example 3 — Clean noisy data

Review these source CSV files and remove records that are duplicates, incomplete, unrelated to the services in services.md, or do not have a reasonable business follow-up opportunity. Create a new cleaned-leads.csv. Do not modify the source files. If information is missing, leave it blank or label it unknown instead of guessing.

Example 4 — Narrow a large list

There are too many records here to work manually. Use services.md as the business filter and return only the 20 best opportunities for 8A Pro Services. Favor clear service fit, reachable prime contractors, useful contract value, and opportunities where subcontracting is realistic. Explain briefly why each selected record made the list.

Example 5 — Create a follow-up file

Use these source files to create a new parsed.md for business follow-up. For each lead include: company, project, location, award value if known, why 8A Pro may fit, source of the information, and one specific next action. Do not invent contact names, phone numbers, services, or project details that are not present in the source data.

A useful pattern: tell ChatGPT what the data is → why you are allowed to use it → what business question you want answered → what output to create → what it must not assume.

Success: a small, ranked file with real companies or projects, clear reasons they may be worth contacting, and enough source information to verify the result.

Watch out: a giant list is not automatically a good lead list. Ten qualified leads with next actions are more useful than 1,000 rows nobody will work.

Lesson 02 • Marketing

Create a graphic that actually fits the business

Input: business contextOutput: graphicSkill: framing the request

What we learned: “Give me a graphic for my business” leaves too many decisions up to the AI. The result improves dramatically when we explain who the graphic is for, how it will be used, what format it needs, and which business branding should guide it.

Before asking for a graphic, answer a few concrete questions: Who will see it? Where will it be used? What size or format is needed? What should the viewer do after seeing it?

Weak requestGive me a graphic for my business.

Better request

Create a large-format promotional handout for 8A Pro Services for a Chamber of Commerce event. Use the business services and existing website branding as references. Make it suitable for a PNG and make the company name, services, and contact action easy to understand at a glance.

What worked: using the microphone to simply explain the situation conversationally. You do not need to memorize a special “AI language.” You need to describe the job clearly.

Watch out: if the graphic uses the wrong colors, services, logo treatment, or contact information, do not just keep prompting randomly. Point ChatGPT back to the correct business reference files.

Lesson 03 • AI Guard Rails

Give ChatGPT a business operating manual

services.mdbranding.mdSkill: reusable context

Why this matters: the more reliable business context ChatGPT has, the less Vianey has to repeat in every request and the less likely AI is to invent services, use the wrong style, or drift away from how 8A Pro actually operates.

`services.md` = What the business does

Use this file for confirmed services, capabilities, preferred work, service areas, limitations, and anything Juan has explicitly said the company can perform.

`branding.md` = How the business presents itself

Use this file for company name, logo rules, colors, tone, preferred wording, contact information, visual direction, and other branding choices. This becomes the reusable reference for flyers, social posts, webpages, graphics, and customer-facing material.

Example request

Read services.md and branding.md before doing this task. Do not add services that are not confirmed in services.md. Keep the output consistent with the branding guidance.

The idea: reference files act like guard rails. They do not replace Vianey or Juan's judgment; they keep the AI closer to the business while it works.

Quick Reference

Words that help you tell AI what you want

Think in two parts: file type tells ChatGPT what kind of output you want; format tells it how the final piece should be shaped and used.

Common file types

PNG

Use when you want a higher-quality image, a graphic with text, or something you may reuse in a flyer, website, or social post.

JPG / JPEG

Use when you want a smaller image file, especially for photos or when file size matters more than perfect text sharpness.

PDFUse when you want something finished for printing, emailing, or sharing where the layout should stay the same.

CSVUse when you want rows of data you can sort, filter, open in Excel, or reuse later.

MD / Markdown

Use for reusable notes, instructions, business rules, services, branding guidance, or reference information for future AI work.

Common formats

PortraitTaller than it is wide. Good for flyers, handouts, phone screens, and many social posts.

LandscapeWider than it is tall. Good for presentations, banners, website headers, and wide graphics.

Poster

A large single-page design meant to be seen from farther away. Ask for large text, simple messaging, and a clear call to action.

Pamphlet / Brochure

A handout with more detail than a poster. Tell ChatGPT whether it should be one page, folded, front-and-back, or a tri-fold.

Example: “Create a portrait PNG flyer for a Chamber of Commerce event. Use the 8A Pro branding, make the company name easy to read, and keep the file suitable for printing and sharing online.”

Important: PNG does not automatically mean physically larger, and JPG does not automatically mean physically smaller. These file types mainly affect image quality, transparency, and file size. Tell ChatGPT the dimensions or use case when size matters.

Reusable Workflow

The lesson underneath all three lessons

The goal is not to memorize perfect prompts. The goal is to learn a repeatable way to describe work to AI.

1. What do I have?

2. What do I want?

3. What rules matter?

4. Review the result

5. Refine or use it

Think of ChatGPT like a new employee: it works much better when it has the source material, knows the assignment, understands the boundaries, and gets reviewed before the work goes out the door.

Training Home Open Knowledge Base Open Leads

Future AARs

Keep adding lessons as we test real work

Future lessons can be added here after Vianey uses a workflow herself. A useful AAR should record four things: what we tried, what worked, what did not work, and what we would do differently next time.

Good training is cumulative. Every successful workflow should make the next task easier instead of disappearing into yesterday's chat history.

Learn it. Use it. Improve it.