Prompting at Scale: How to Deploy AI Across Your Company

Wish there was an easier way to get your staff to embrace AI? Looking for an interchangeable prompting system everyone can use?

In this article, you'll discover how to scale AI prompting across a company.

Why Most Organizations Struggle With AI Implementation

If you've experimented with artificial intelligence (AI) tools like ChatGPT or Claude, you've likely experienced the initial excitement of quick, impressive results. But as you've tried to expand AI adoption across your organization, you've probably encountered a familiar pattern: inconsistent results, frustrated team members, and endless prompt refinements.

This challenge multiplies across your organization as different team members develop their approaches to prompting and AI projects, leading to inconsistent results and inefficient workflows.

The solution isn't just better prompts—it's a systematic AI framework that scales.

To implement this structured approach, you need a framework everyone in your organization can understand and use.

The AI Strategy Canvas from John Munsell provides precisely that by breaking down effective prompts into nine essential blocks that work together to produce better AI-generated content across your business.

John's journey from banker to AI strategist offers valuable insights into the development of scalable AI systems.

His transition to AI strategy wasn't immediate. After selling his agency's client portfolio, he launched the CXO AI Roundtable, a weekly forum where he taught AI concepts to business leaders.

“Every Friday, I would teach people something about AI,” he recalls. “Sometimes it was investigating new tools, evaluating different solutions for specific problems.”

These sessions revealed two critical market needs:

1. A way to distill and make sense of rapidly evolving AI capabilities
2. A systematic approach to scaling AI usage across organizations

This real-world testing ground helped refine what would become the AI Strategy Canvas, a proven approach across various industries and use cases from customer service to grant writing.

#1: The AI Strategy Canvas: Your Blueprint for Scalable AI Prompt Engineering

Before exploring specific techniques, it's essential to understand how scalable prompt engineering differs from traditional approaches. Most people start by asking generative AI tools what they want.

Everyone in your company uses their own language, which means that the results differ wildly, even if they're asking for the same thing.

Instead, you can embrace a system that breaks your prompts into structured blocks that can be repurposed, reused, and combined with other blocks to generate an ideal output.

Think of it as building with LEGO blocks instead of sculpting with clay. Each block serves a specific purpose and can be combined with other blocks to create more complex structures. This modular approach makes it easier to:

• Create consistent results across different teams
• Quickly adapt prompts for new uses
• Troubleshoot when something isn't working
• Share effective techniques across departments

First, Standardize Your Prompt Block Formatting

This approach builds on a two-part foundation of prompt formatting and dynamic variables that can be quickly dropped into a prompt.

You need to standardize how a block is constructed to ensure everyone in your organization understands what it is for and how they can repurpose it for specialized use. This is easily achieved by using delimiters to open and close a prompt block.

Imagine you want to write an email using Claude. Start your block with all caps and a colon:

AUDIENCE:

Then, enter the name of a target audience persona you've already fully defined and stored in a library where your AI can retrieve it. This is your variable:

The audience is: ${target_audience_a}

Using dynamic variables that pull standardized details from a library ensures that everyone can easily include the correct information in their requests because there is a single source of truth. You can store these variables in Notion or similar tools your AI can connect to, allowing quick substitution (AKA hot-swapping) for different use cases.

Close your block with a forward slash:

/AUDIENCE

For example, this prompt tells the AI that you want it to generate content for prospects who fall into your Target Audience A without you having to copy and paste a lengthy persona detail.

AUDIENCE:

The audience is: ${target_audience_a}

/AUDIENCE

Building Block 1: Audience Details

The foundation of any good prompt begins with a clear understanding of who will receive the output. This goes beyond simple demographics. Think about who you're creating value for. Is it customers? Prospects? Employees? Using psychographics, demographics, economics, and sociographics, build a complete profile for each group.

Give each persona a name–paying particular attention to taxonomy, and store them in your prompt library for hot-swapping.

Pro Tip: Ask your AI tool to analyze customer data and suggest key characteristics you might have missed.

Building Block 2: Company Context Integration

Your AI tools need to understand your organization to provide relevant outputs. However, many companies need help with how much information they should provide.

At a base level, this block should tell the AI who you are, what your company stands for, and why your target audience should care about you. However, the key is creating a layered company context that anyone can use flexibly, depending on the task.

“Think of your company context like layers of an onion,” Munsell explains. “The outer layer might be your basic ‘About Us' content, but deeper layers contain everything from your mission and values to detailed historical information and testimonials.”

Create multiple versions of your company context, from a one-paragraph summary to comprehensive documentation to use at various stages of the customer journey. 

Name and store these company variables in your prompt library for easy access to ensure consistency across your organization.

Building Block 3: Products and Services

How you present your products and services to AI tools significantly impacts the quality of their output. You should maintain two distinct versions of your product information to detail what value you deliver to customers:

1. A concise, promotional description for basic tasks
2. A comprehensive document detailing features, benefits, and technical specifications

You can use a variable with basic product information if you're writing something that just needs to mention a product. For more specialized writing tasks, you'll need variables that contain extensive product documentation.

Name and store these product variables in your prompt library—for example, crm-product-brief and crm-product-detail.

Building Block 4: Dynamic Context Integration

This flexible component allows you to provide situation-specific information that shapes the AI's understanding of your current needs. Think of it as a situation-specific catch-all block.

For example, when analyzing a customer support call, your dynamic context might include:

• Key points discussed in the review meeting
• Specific challenges your team identified
• Any customer feedback received during the call
• Current business objectives related to the discussion

As you document and save successful dynamic context blocks in your prompt library over time, you'll develop a collection of variables that can be adapted for similar situations.

Building Block 5: Role Definition

The role you assign to your AI tool significantly influences its output. Many role-related prompts fail because they don't clearly define the AI's role in addressing your request.

“It's not just about saying ‘be an expert,'” Munsell cautions. “You need to specify what kind of expert and what perspective the AI should take. A chemical engineer will write very differently from a marketing strategist, even about the same topic.”

Be very clear about the roles your team will need and how you want the AI to behave within the scope of each role. For example:

• You are a data analyst 
• You are a customer service analyst
• You are a world-class copywriter

Save each role to your prompt library.

About Prompt Conflict

A prompt conflict can arise between Blocks 5 and 6 if you give the AI too much detail about the role you've told it to take on, meaning that the role may supersede any style prompts. For example, if you tell the AI to act as a chemical engineer and write in a friendly voice, the AI will struggle, and the output will likely sound very clinical rather than warm and friendly. If you have trouble getting the desired output, scale back your role instructions.

Building Block 6: Style and Voice Calibration

Here is where you and your teams will discover the true power of scalable prompting. Instead of everyone entering their own version of vague instructions like “make it professional,” you can develop specific parameters to control everything from tone to sentence length in AI outputs.

Munsell maintains an exclusions rule that gives the AI a list of words and phrases his company doesn't want to see in its content–words like ‘delve' and ‘skyrocket' or phrases like ‘I hope this email finds you well.' By using this rule, the AI learns to avoid these clichés and produces more original content.

While he says identifying and implementing variables to control these facets is more art than science, he has developed a surprisingly simple but systematic approach.

“I literally just ask the AI, ” Munsell explains. “If I notice something I want to control, like repetitive sentence structure, sentence length, or technical complexity, I ask…”

What's a variable I can use and a numeric I can adjust to fine-tune sentence length?

This direct questioning has led to identifying 56 different style parameters that can be adjusted using numerical scales. Here are four basic parameters with scales:

• Humor Level: Scale of 0-10
• Sentence Length Variation: Short (1) to Complex (10)
• Technical Language: Basic (1) to Expert (10)
• Formality: Casual (1) to Formal (10)

When working with different AI models, note that some require more complex numeric values and contextual descriptions. For instance, ChatGPT might work with just the numeric value, while Claude may require you to use:

The tone should be: ${humor = moderate humor, 6/10}

Create several style and voice blocks so that you're using the best fit for every need, such as:

• Internal emails
• Prospect emails
• Customer emails

Save each style and tone variable in your prompt library.

Building Block 7: Resource Integration

This block connects your AI tools with the information and tools they need to complete your request. These resources might include:

• Meeting transcripts
• Reference documents
• External URLs
• API data feeds
• Advanced data analysis
• Image generation

With these resources in place, you can create template prompts for any analysis you commonly perform. For example, for every meeting you hold, you can use a tool like Fathom to record and transcribe your meetings and simultaneously analyze meetings from different angles. 

Munsell uses AI prompts to generate six different analyses from a single transcript:

1. Action Item Extraction: An email summarizing key tasks and responsibilities.
2. Learning Analysis: A list of ‘aha moments' and areas where concepts clicked for participants.
3. CRM Documentation: Detailed notes for customer relationship management.
4. Strategic Insights: Analysis of patterns and opportunities that can be improved.
5. Process Refinement: A list of areas where teaching or communication could improve.
6. Follow-up Planning: A structured follow-up schedule for content that must be created.

Building Block 8: Rule Setting

Clear rules prevent typical AI outputs that don't align with your needs. Beyond basic guidelines from Block 6, these are detailed rule sets that put guardrails in place for things such as data security and copyright infringement.

Building Block 9: Request Formulation

While this is the final block, it's where most people start—and precisely why they struggle. Your actual request for output should build upon all previous blocks to create a clear, actionable instruction for your AI tool–what you want the final output to look like.

For example, if you want a comprehensive article, use the word ‘Write” instead of “Outline.” You must also consider any word-count limits and consider whether you need responses broken into sections.

#2: Implementing the AI Canvas Strategy Across Your Organization

Once you compile multiple variables for each block, the next step is to implement them across your organization by creating custom GPTs in ChatGPT or Claude Projects for frequently repeated tasks. These AI models can access your organization's knowledge base and provide more consistent results.

AI Tool Selection and Integration

Different AI tools excel at different tasks. Understanding these strengths allows you to choose the right tool for each job. Here are three tool stacks to consider:

AI Tool Stack for Meeting Documentation

• Fathom for recording and transcription
• Custom GPTs for analysis
• Claude for narrative development

AI Tool Stack for Content Creation

• ChatGPT for strategic planning
• Claude for copywriting
• Custom GPTs for specialized tasks

AI Tool Stack for Knowledge Management

• Notion for prompt libraries
• Dropbox for detailed documentation
• Custom GPTs for specific subject matter domains

Document Management and Organization

Successful organizations use tools like Notion to maintain their prompt libraries. Create a structured system that includes:

• Consistent taxonomy
• Complete prompts for everyday tasks
• Reusable context blocks for hot-swapping variables
• Style guides and rule sets
• Example outputs and use cases

Train Your Team

Invest time in proper training. Ensure everyone understands:

• The building blocks of effective prompts
• When to use different AI tools
• How to access and use your prompt library
• Troubleshooting techniques

Monitor and Refine

Regularly review and update your AI processes:

• Refine your style parameters
• Collect feedback from team members
• Document successful prompt patterns
• Update your exclusions list

Other Notes From This Episode

• Connect with Michael Stelzner @Stelzner on Facebook and @Mike_Stelzner on X.
• Watch this interview and other exclusive content from Social Media Examiner on YouTube.

Where to subscribe: Apple Podcasts | Spotify | YouTube Music | YouTube | Amazon Music | RSS

✋🏽 If you enjoyed this episode of the AI Explored podcast, please head over to Apple Podcasts, leave a rating, write a review, and subscribe.

---

Stay Up-to-Date: Get New Marketing Articles Delivered to You!

Don't miss out on upcoming social media marketing insights and strategies! Sign up to receive notifications when we publish new articles on Social Media Examiner. Our expertly crafted content will help you stay ahead of the curve and drive results for your business. Click the link below to sign up now and receive our annual report!