The 5-Step System That Turns AI Into Your Graduate Assistant (Instead of a Time Waster)
Stop Treating AI Like Magic
About an hour into the call, I had to stop everything.
We were building a consulting report together. My consultant friend and I, working through 80 pages of field notes from his $30K project.
We were using AI to help organize and analyze it all, and ChatGPT had just drafted a section about the four field offices he’d visited. It looked good. Professional. Well-written.
Except it was wrong.
“Where are you getting this information?” I asked ChatGPT. “Why didn’t you use the field office names? I only want you to use what I gave you in the draft report and project documents. Be specific.”
Rob looked confused. “It seemed fine to me.”
That’s the problem. It seemed fine. That’s why most people end up with AI outputs they have to completely rewrite.
The AI had invented generic placeholder content instead of using the actual data we’d given it. If we’d kept going, the next six sections would have built on invented information. We’d be rewriting the entire report.
This is what happens when you treat AI like a magic content generator instead of what it actually is: a very capable, very literal junior consultant who needs constant validation.
A short while later, we had a 27-page client-ready report.
Rob’s exact words: “It’s like having a graduate assistant.”
You Don’t Have a Prompting Problem
Instead of worrying about better prompts, we used a better process.
Most people use AI the same way:
Open ChatGPT. Maybe dump in notes.
“Write me a report.” Or “Write me a sales page.” Or “Write me a course outline.”
Get back generic content that needs complete rewriting.
The problem isn’t the AI.
It’s treating it like magic instead of treating it like what you’d actually do if you hired a junior consultant.
If you hired someone smart but new to your work, you wouldn’t hand them 80 pages of notes and say “write me a report.”
You’d set up their workspace. Give them clear instructions about the audience and format. Build an outline together. Work section by section. Validate as you go.
That’s exactly how you need to work with AI.
The 5-Step System
This is the process we used. Same process works whether you’re creating consulting reports, sales pages, course content, video scripts, or long-form newsletters.
STEP 1: PREPARE – Set up the environment before asking for content
STEP 2: PLAN – Create structure first, content second
STEP 3: EXECUTE – Build section by section with constant validation
STEP 4: ITERATE – Refine complex sections through multiple approaches
STEP 5: COMPLETE – Write summaries last, not first
Here’s exactly what each step looks like.
STEP 1: PREPARE – Set Up the Environment
Most people skip this. They open ChatGPT, paste notes, ask for content.
Don’t.
What to do:
Create a dedicated project (ChatGPT Projects or Claude Projects).
Write system instructions that tell AI its role, the audience, output style, and objective.
Upload SOURCE documents to project files. Keep them separate from your draft notes.
For consultants: source documents are research, field notes, client briefs, background reports.
For content creators: source documents are interview transcripts, testimonials, product details, past successful content.
Avoid this mistake:
Uploading draft notes as project files. That’s raw material to organize and you don’t want the AI to see it as authoritative.
Exmple System Prompt Template:
You are my AI co-consultant on this project: [PROJECT NAME]
Your job is to help me:
- Cut through noise and organize information clearly
- Write quickly without sacrificing quality
- Produce [TYPE] that serves [AUDIENCE]
Context and Style:
- Audience: [specific role, needs, concerns]
- Tone: [Professional/Conversational/Technical/Persuasive]
- Output Style: [format expectations]
Our objective: [PASTE ACTUAL PROJECT OBJECTIVE]
Critical Rules:
- Use ONLY source documents and my draft
- Never invent or extrapolate beyond what I provide
- When information missing, flag it - don't fill gaps
- Ground all analysis in actual data providedExamples:
Consulting report: “You are my AI co-consultant helping me create a consulting report for field office staff who already know the background. Write clearly and succinctly.”
Course content: “You are my AI co-consultant helping me create a course for solopreneurs who want to build email lists. Write conversationally and practically.”
Sales page: “You are my AI co-consultant helping me create a sales page for consultants trapped in delivery bottleneck. Write persuasively using direct response principles.”
Test it:
Ask AI: “What’s your role in this project?” If it tells you back the audience, tone, and constraints, you’re good.
Time: 5 minutes
STEP 2: PLAN – Build the Structure First
This is where most people fail.
They ask AI to “write a report.” AI writes it. Structure is wrong for the audience. Includes sections that don’t matter.
Complete rewrite.
What to do:
Upload draft notes to the conversation (NOT project files).
Prompt: “Based on this draft and the source documents, suggest a detailed outline.”
Give examples, but let it think:
For reports: “Consider sections like executive summary, methodology, findings, recommendations, but use your judgment.”
For sales pages: “Consider sections like problem, solution, how it works, social proof, offer, but structure for maximum conversion.”
For courses: “Consider modules that build sequentially, but organize based on the actual learning journey.”
Review. Adjust. Get it right.
Copy final outline to a separate document for reference.
The mistake:
Telling AI exactly what sections to create without letting it think.
You want: “Consider sections LIKE...” (suggestions, not commands).
Our example:
AI suggested consulting engagement structure. We modified to include summaries of the four field offices. Locked it in. Referenced it throughout.
Time: 10-15 minutes
STEP 3: EXECUTE – Section by Section with Validation
This is where AI becomes your co-pilot or a liability.
The difference: constant validation.
What to do:
Start with body sections, not intros or summaries.
Work section by section:
“Draft [section name]. Use draft notes and source documents. Ground everything in actual material.”
Paste in the full section outline you’d copied to a seprate document. Doing this keeps the AI “honest” and prevents it from forgetting the outline and inventing new stuff.
Read what it produces
Catch errors or drift immediately
Add missing context
Let the AI know what the final version looks like so it has the context for the next sections.
Move to next section
The critical skill:
Catching drift the moment it happens.
That field office section looked fine. But AI had invented placeholder content instead of using actual names and data.
I stopped immediately: “Where are you getting this? I only want you to use what I gave you in the draft and project documents. Be specific. Name the field offices. Ensure everything is correct.”
AI regenerated with actual details from source material.
If you let AI drift in section 3, sections 4-7 build on invented information. You’re rewriting everything.
Validation technique:
As you read each section, ask: “Could I have written this from my notes and sources?”
If yes, keep going. If no, stop and re-ground.
For content creators, this is critical for:
Client testimonials (don’t let AI embellish)
Product features (stick to what you actually offer)
Results and outcomes (only real numbers)
Case studies (ground in actual stories, not invented scenarios)
Speed technique:
Use voice-to-text to make corrections as you read:
“Section 5.1 needs more about ABC. Section 5.2 looks good except 5.2.2 should say ‘downstream engagement.’ Section 5.2.3 needs detail about D.”
Fix as you go, not later.
Re-Grounding Prompt:
“Where are you getting this information? I only want you to use what’s in the draft and project documents. Be specific. Ensure everything is grounded in actual source material.”
Time: 15-90 minutes (depending on length & complexity)
STEP 4: ITERATE – Try Multiple Approaches
Not every section works on first try.
Some sections are complex. Multiple ways to organize them.
What to do:
For complex sections, don’t accept the first version.
Prompt: “Give me 2-3 different ways to structure this section.”
For reports: “Try chronological, thematic, and priority-based approaches.”
For sales pages: “Try problem-solution, before-after-bridge, and story-based approaches.”
For courses: “Try skill-building, problem-solving, and case-study-based sequences.”
Review options. Pick the best or combine elements.
Our example:
For the Recommendations section, Rob suggested: “Could we structure this thematically? Sustainability, system strengthening, proximity, value added, cost savings.”
We asked AI to re-outline thematically. Compared both versions. Chose the better one.
The mistake:
Accepting mediocre first drafts because you’re tired.
Best sections often come from iteration 2 or 3.
Time: 10-20 minutes per complex section
STEP 5: COMPLETE – Write Summaries Last
Don’t start with summaries or intros.
What to do:
Draft all body content first. Get it right.
Then ask AI to create summary pieces based on actual final content:
Now summaries reflect what you actually wrote, not what you planned.
The mistake:
Writing summaries first, then revising them 5 times as body content evolves.
Our example:
We wrote all body sections first. Validated each. Then:
“Draft the conclusion based on actual content.”
“Draft the executive summary based on the report we’ve created.”
Perfect match. No revisions needed.
Time: 5-15 minutes
What “90% Ready” Means
When he saw what we were able to do, my friend saw it was like having a graduate assistant to do the work.
The 27-page report needed:
Minor word choice adjustments
Reordering a few points
Adding one or two examples I knew but hadn’t mentioned
Formatting for submission
What it DIDN’T need:
Complete rewrites
Fact-checking everything (validated as we went)
Reorganizing sections (built structure first)
Fixing invented information (caught drift immediately)
That’s 90% ready. Just needs some polish, not reconstruction.
Start With Your Next Project
Don’t wait for perfect.
Try this for your next consulting report, sales page, course module, strategy document, newsletter, etc..
Don’t jump to “write this for me.”
Take a moment to set up the environment. A few more minutes to build structure. Then execute section by section, validating as you go.
You’re not working harder. You’re working systematically.
The difference between 60% drafts you rewrite and 90% drafts you polish.
Hit reply or leave a comment and tell me: What’s your process for getting good outputs with AI?