Procedure Writing
Getting Started Guide to Prompt Engineering for SOPs
Apr
Recently, I’ve been exploring how Artificial Intelligence (AI), specifically Large Language Models (LLMs) like Google’s Gemini or OpenAI’s ChatGPT, can ‘re-imagine’ how we create and manage Standard Operating Procedures.
I’ve learned a ton trying to get these to work for us, and thought to share some of the lessons learned. If you’re new to AI and trying to find a way to weave it into your daily SOP writing, this set of tutorials should help.
As mentioned, I’ve been running some internal training sessions helping colleagues at a small manufacturing firm streamline their SOP development, and the results have been genuinely impressive. It’s lead to faster drafting, more consistency, easier updates. So, I thought I’d package up the key learnings into a practical, multi-part tutorial series specifically for fellow SOP writers.
Where to start with AI?
Think of it as a super-powered assistant. It can’t replace your subject matter expertise (SME) or understanding of your organization’s specific needs (Sean – it can. We’ll have more on this in future tutorials), but it can handle a lot of the heavy lifting involved in research, drafting, and polishing. The key is learning how to ask the right questions – or in AI terms, learning how to write effective prompts.
What Makes a Good Prompt for SOPs?
A prompt is an instruction.
That’s all. Based on best practices, like those outlined by Google in their prompt engineering guides, a prompt generally considers four key elements.
Getting these right tells the AI exactly what you need, leading to much better results. Think of it like giving clear instructions to a new team member:
- Persona: Who should the AI act as? Defining a role helps the AI adopt the right tone, style, and knowledge base.
- SOP Example: “Act as an experienced pharmaceutical quality assurance specialist…” or “You are a technical writer specializing in safety procedures for manufacturing…”
- Task: What specific action do you want the AI to perform? Be clear and concise. Use action verbs.
- SOP Example: “…draft the initial steps for…” or “…rewrite the following paragraph for clarity…” or “…identify potential safety hazards related to…”
- Context: What background information does the AI need? This is crucial. Provide relevant details about the process, the audience, the industry, existing documents, or specific constraints. The more relevant context, the better the output.
- SOP Example: “…for the process of receiving raw materials in a food production facility.” or “…based on the attached notes from the subject matter expert.” or “…keeping in mind that the end-users are new hires with minimal technical background.”
- Format: How should the AI present the information? Specify the desired output structure.
- SOP Example: “…in a numbered list format.” or “…as a bulleted summary.” or “…in a table with columns for ‘Step,’ ‘Responsibility,’ and ‘Notes’.”
Putting it all together (Example):
Let’s adapt the Google Workspace example for an SOP task:
Persona: You are a senior safety officer in the construction industry.
Task: Draft the initial safety checklist steps
Context: for operating a concrete mixer on a residential building site, based on standard OSHA guidelines and common site hazards. Assume the operator has basic training but needs clear reminders.
Format: Present this as a numbered list, using simple, direct language.
See how specifying each element gives the AI clear direction? This is far more effective than just asking, “Write safety steps for a concrete mixer.”
In this series, we’ll break down how to apply this PTCF (Persona, Task, Context, Format) framework to different stages of the SOP lifecycle. We’ll start with the often-challenging initial research and outlining phase.
Ready to make your SOP writing process smarter and faster? Let’s look at the first tutorial.
