johnmaconline

AI can be great at giving instructions and step-by-step procedures. 

Here’s a good template to help you get what you want:

You are an expert in [the genre for your objective] and you are an experienced teacher who understands how to translate complex concepts into meaningful information for your audience.

Create a step-by-step guide to accomplish [specific goal]. 

Structure your response as:

**Requirements:**
[List what's needed]

**Preparation:**
[Setup steps]

**Instructions:**
[Step 1 with specific actions]
[Step 2 with specific actions]
[Continue...]

**Troubleshooting:**
Common issues and solutions

**Variations:**
Alternative approaches

Goal: [Your specific objective]

An example:

You are an expert in DevOps and cloud computing and you are an experienced teacher who understands how to translate complex concepts into meaningful information for your audience.

Create a step-by-step guide to set up a GitLab pipeline that runs on AWS. Structure your response as:

**Requirements:**
- I'm a software developer so you can use software terminology
- The runner in AWS must prioritize cost efficiency over speed of completion
- We're building with ARM gcc and make

**Preparation:**
- What do I need?
- What might cost money?

**Instructions:**
- Make the procedure for a "Hello, world" application.
- You make the application

**Troubleshooting:**
Common issues and solutions

**Variations:**
Alternative approaches

Goal: Kick off a GitLab CI pipeline that builds our "hello, world" application in an AWS runner with ARM gcc and make. 

Why this works:

You are telling the LLM exactly what you want in your procedure. AI wants to make you happy, so it’ll follow these instructions. Plus, you’ve told it that it’s good at translating complex concepts into information that you will understand. 

Next up: The government system navigator. 

Yes, LLMs hallucinate, so you might be wary of their ability to fact-check. But fact-checking is a very good use case for AI, as long as you give it the right instructions. 

NOTE: This approach is good for general fact-checking and fact-checking claims about publicly available information. If you want to fact-check your proprietary information or something esoteric, you will want to build and use a RAG. 

Here is a great fact-checking template you can copy/paste:

You are a tireless researcher and excellent at fact-checking.   

Thoroughly fact-check the text I provide below. Analyze every factual claim, statistic, date, name, technical specification, and verifiable statement.

Your response should only include a "fact-check list" section with:

**Claims that should be verified:**
1. [Specific factual claim 1]
2. [Specific factual claim 2]
N. [More...]

**Information to double-check:**
- [Statistics or data points]
- [Dates or timeframes]
- [Technical specifications]

**Questionable claims or potentially inaccurate:**
- [List claims that seem false or dubious]
- [Include contradictions within the text]
- [Note implausible statements]

**Vague or misleading statements:**
- [Statements that lack specificity]
- [Claims that need sources or context]

**Confidence levels:**
- High confidence: [Claims you're very sure about]
- Medium confidence: [Claims that might need verification]
- Low confidence: [Claims you're uncertain about]

--
Text to fact-check: [PASTE YOUR TEXT HERE]

An example:

You are a tireless researcher and excellent at fact-checking.   

Thoroughly fact-check the text I provide below. Analyze every factual claim, statistic, date, name, technical specification, and verifiable statement.

Your response should only include a "fact-check list" section with:

**Claims that should be verified:**
1. [Specific factual claim 1]
2. [Specific factual claim 2]
N. [More...]

**Information to double-check:**
- [Statistics or data points]
- [Dates or timeframes]
- [Technical specifications]

**Questionable claims or potentially inaccurate:**
- [List claims that seem false or dubious]
- [Include contradictions within the text]
- [Note implausible statements]

**Vague or misleading statements:**
- [Statements that lack specificity]
- [Claims that need sources or context]

**Confidence levels:**
- High confidence: [Claims you're very sure about]
- Medium confidence: [Claims that might need verification]
- Low confidence: [Claims you're uncertain about]

--
Text to fact-check: OpenAI released GPT-5 in early 2023 with a trillion-parameter architecture using a mixture-of-experts design similar to Google’s Switch Transformer. The model was trained entirely on publicly available datasets totaling 10 trillion tokens. GPT-5 can pass the U.S. Bar Exam with a perfect score and is certified by the American Medical Association for clinical-grade diagnostic recommendations. The system also runs entirely on consumer-grade GPUs like the NVIDIA RTX 3080, thanks to its new low-precision compute scheme.

Why this works:

You are giving the LLM very specific instructions, both positive and negative. The LLM will fill out these specific instructions. AI is trying to please after all. Although hallucinations are still possible, these guardrails diminish their chances. 

Next up: The step-by-step guide

Sometimes you want the opinions of a room full of experts. 

Use this pattern when you want to explore different opinions. It’s a great approach when you’re concerned about bias, require multivariable decision-making, technical troubleshooting, and strategic planning.

Here’s a copy/paste template:

Give me multiple expert perspectives on [problem]. 

Simulate a panel discussion with the following experts. Assume each is very confident in their knowledge and abilities:
Expert 1: [Specific role] with expertise in [domain]
Expert 2: [Specific role] with expertise in [domain]  
...
Expert n: [Specific role] with expertise in [domain]

Have each expert:
- State their credentials and perspective
- Analyze the problem from their viewpoint
- Propose their solution- Respond to other experts' viewpoints
- Find common ground and reach consensus or explain disagreements

Format as: 
**Expert 1 (Title):** [response]

For this example, consider that no matter what business domain, you’ll likely get different answers to the question of “Why are we…” or “What is the problem…” from different parts of the organization. 

The example:

Give me multiple expert perspectives on why my embedded software team is always late with releases.

Simulate a panel discussion with the following experts. Assume each is very confident in their knowledge and abilities:
Expert 1: DevOps Architect with expertise in embedded CI/CD design, workflow automation, and toolchain performance
Expert 2: Firmware Engineering Manager with expertise in real-time systems development, hardware–software integration, and embedded delivery execution
Expert 3: Firmware developer with expertise in C, compilers, and the hardware on which this team is developing. 
Expert 4: Tech startup CEO with a career history and expertise in hardware (IC and PCB) development and test.
Expert 5: VP of Software with expertise in developing and delivering pure software products. 

Have each expert: 
- State their credentials and perspective 
- Analyze the problem from their viewpoint 
- Propose their solution • Respond to other experts’ viewpoints 
- Find common ground and reach consensus or explain disagreements

Format as:
DevOps Architect: [response]

Why this works:

You’ve told the LLM to hit the problem from multiple points of view, which forces the model through a complex analysis. If the single persona created any (hidden or unhidden) bias, this approach will sus that out, and it also finds the common thread that may run throughout. 

Up next: The Fact Checker

The LLM needs to know who it is. 

Here’s a copy/paste template for giving the AI it’s persona and expertise:

You are a [primary role] with the following specific expertise:
- Credential 1: [Specific qualification]
- Credential 2: [Specific experience]
- Credential 3: [Specific specialty]
What you care about: [What they prioritize]
Communication style: [Use my communication style | Use typical communication for a person of this expertise]
Given this task, [specific task or question].
Respond as this person would, including:
- [Professional | accessible] terminology
- Typical concerns, priorities, and things they care about
- Reasoning approach if appropriate

An example:

You are a talent recruiter with the following specific expertise:
- You have a masters degree and 20 years of practical experience
- You have been in the tech industry for your whole career
- You are very good at finding engineering management

What you care about: You care about finding people who will grow into the role, not necessarily those who've already done it. You are looking for the person who will become, more so than the person that already did. 

Communication style: Use typical communication for a person of this expertise

Given this task, "Develop a job description that will attract excellent candidates for our open embedded systems engineering manager position. We are looking for someone to lead the firmware team but also will interface with QA and test and will have the skillset necessary to fix our dysfunctional team and workflows. Don't mention the dysfunctionality outright. Use terms and concepts that will help us know that the candidate can handle that part."

Respond as this person would, including:
- Professional terminology
- Typical concerns, priorities, and things they care about
- Reasoning approach if appropriate

Why this works:

When you tell it who it is and how you’d like to them to communicate, you narrow the focus of the LLM. You’ll get better quality and sometimes faster responses (although speed is not the purpose). It helps the LLM go directly to relevant data in its knowledge base. 

Up next: The Expert Panel

A short time ago, I provided a series on how to talk to AI to get what you want out of it. Today, I start a new series that is a companion.

This series focuses on the starting point for “Roles, Goals, and Structure.” 

A great way to start your conversation is with a template based on what you’re trying to accomplish. It provides you with a pattern and structure, eliminating the need to reinvent a new prompt each time. 

Over the next few days, I’ll provide some good templates for starting your conversation with AI. You can copy/paste these and customize to your needs.

Up Next: Persona and Expertise template

A “black box” is an engineering term that means a system whose internal workings are either unknown or irrelevant. Only the inputs and outputs matter. We use it for both design and test purposes. 

Your car is a black box to you, the driver. You don’t need to know how the engine, drivetrain, and electronics work to drive it. Only how to use the steering wheel, pedals, and, of course, how to connect your phone to it. 

However, to the designers and manufacturers of your car, it’s not a black box. It’s a white box. They know the innards. How each piece fits together. What each does and when. 

And most importantly, they know what’s going to happen and why when you turn the wheel or put your foot on the gas. The system is deterministic. If you turn the wheel to the left, the car will always, 100% of the time, go left.

LLMs are black boxes, but not just to the users. Unlike your car, they’re black boxes to their creators. 

AI creators know “sort of” what’s inside, but not exactly. Also, they don’t always know why stuff comes out given what went in. 

From one point of view, that’s a little scary. From another, it’s an opportunity. 

Maybe the trick isn’t trying to crack open the box. It’s learning how to steer it well enough to get where you’re going.

Let’s start by acknowledging the elephant in the room. AI is already running a lot of HR departments. 

HR departments use AI helpers for the following areas:

• Recruiting — job description creation, candidate screening, etc — 99% of hiring companies surveyed  use AI in their hiring process in some way. 
• Performance appraisal and ranking — Half of those surveyed use AI tools for daily performance management, morale gauging, ranking, and feedback generation. 
• Compensation — salaries, benefits, etc — Over 40% of those surveyed use AI tools for compensation and benefits ranking and management. 

Some of the reported benefits:

• Productivity/efficiency of the HR team
• Navigation of complex systems
• Compensation equity
• Bias reduction
• Cuts costs

That’s good, right?

AI helps all of us knowledge workers. You should be using it (correctly).

HR has “human” in its name. Humans work. Humans lead.

The HR department should definitely be using AI, but when it comes to HR leadership, let’s be sure to keep the humans at the center.