AI and Prompt Engineering for MSPs
You're missing out on productivity if you're not using AI now. Soon, if you don’t use AI and don’t scale your operations using AI, you will start losing customers to competitors that will be able to serve more customers at a lower cost. Cost-conscious SMB owners will gladly switch to MSPs that offer services at a lower price, especially if they are not delivering value beyond traditional infrastructure management and helpdesk services.
Many MSPs have implemented AI in their daily operations. This article will examine easy-to-implement scenarios and best practices for AI prompt engineering that do not require development and can be achieved using public large language models (LLMs) like ChatGPT.
Scenarios for automation
Number of scenario MSPs outsource to AI – creating content for marketing purposes. Typically, MSPs lack dedicated personnel for marketing, and they must rely on part-time marketers and agencies to produce the content themselves. AI helps to significantly reduce the time required to create marketing assets, and, most importantly, review and update them as assets tend to age.
Sales and marketing scenarios:
- Website pages – maintaining product catalog, service descriptions, blog content and SEO optimization.
- News and social media – writing announcements, responding to market news (for instance, guiding customers to defend against a new cyberattack).
- Case studies and customer stories – writing case studies for the website and using them in the sales process based on projects with customers.
- Adjusting proposals – modifying standard proposals based on a specific customer’s needs.
The second most popular scenario is automation for customer communications. Preparing responses and creating various reports consumes a significant amount of the technicians' productive time. AI can simplify and accelerate the development of documents, especially when templates are developed and a straightforward process is in place to adjust them to a specific case.
Customer communication scenarios:
- Email templates – responses to popular requests, generic guides for step-by-step issue resolution, explanations of various problems and situations, notifications, and announcements. AI is especially useful for preparing security advisory notes in a language the customers will understand, as technical people tend to overcomplicate the explanation of the issues.
- User guides and onboarding materials – documentation for customers to enable them to use self-service to resolve the most frequent issues. A good guide allows for offloading a significant volume of simple problems to the customer.
- Customer-specific FAQ documents – a solid addition to the user guide, answering frequent questions for customers on how to enable specific capabilities, use tools, request new software and hardware, how to prepare and file a ticket with HelpDesk, and so on. Customer infrastructure, needs, and tools vary; AI helps to adjust the content and simplify the language to make the FAQs more useful.
- Knowledge base articles, standard operating procedure (SOP), and implementation guide documents – documenting cases and standard procedures for internal and external use and sharing with customers, vendors and contractors.
- Reports – weekly, monthly, quarterly reports, presentations for business reviews, summaries of the work done, and improvement proposals.
Last but not least, automation for internal operations is needed. Usually, it is the hardest part to automate due to a lack of trust in AI to perform a good job, as it is an essential part of MSP operations. However, AI is useful for expanding on the documents and scripts prepared by the technical expert, validating them, and identifying gaps.
Internal operations scenarios:
- Infrastructure documentation – internal MSP infrastructure and customer infrastructure.
- Process documentation – internal procedures and best practices, essential for the onboarding of new technicians.
- Troubleshooting and automation scripts – writing, debugging and validating scripts for automation in test and production environments.
Prompt engineering technique
Now that we have defined scenarios, it is time to examine the effective prompting technique that has proven useful in achieving consistently high-quality results for the scenarios listed above.
1. Create a prompt: provide context and examples
Begin building the prompt by providing as much context as possible to ensure a successful execution of the task. For example, if you need to create an instruction for end customers about the upcoming operating system update that you will roll out to all machines, provide the AI with details about the customer's infrastructure, the version of the operating system, and any other relevant information. Don’t forget to instruct the AI on the style to use for writing the text, the actions to document, and the steps to execute. If you have an example of a communication, attach it to the chat and tell AI to review it and use it as an example.
Example:
You are an MSP technician writing instructions for non-technical users about the upcoming operating system upgrade. Please provide clear, straightforward instructions, and answer questions you anticipate users may have.
1. Begin with a brief, non-technical explanation of why this update is important (security improvements, performance enhancements)
2. Include clear pre-update preparation steps, including: not shutting down systems at the end of the day, not interrupting the update.
3. Explain the approximate duration and what users can expect from the systems in the process.
4. Explain what users can do to verify that all of their applications works and they can access all systems and data they use.
5. Provide a troubleshooting guide for the common issues with details on how to escalate the issue to the helpdesk if they can’t fix the issue using the guide.
If I attached a file with an example of a communication, please review it and use it as an example to create the new communication.
2. Improve the prompt: evaluate and refine
Before executing the prompt, try asking AI to review and improve it. In the process, you may find out if you are missing anything. There is a simple prompt you can use to achieve that.
Example:
You are an AI prompt engineer. Please review the prompt I provide below that I want to use to create a communication to my customers, non-technical users, about the upcoming system upgrade. Please suggest improvements, and let me know if the prompt effectively achieves the result.
AI will provide recommendations on improving the prompt and the revised prompt.
Example of the AI output:
Add specificity about the update details
- Include a parameter for the specific OS being upgraded (Windows 10 to 11, macOS version, etc.)
- Add a placeholder for the actual scheduled date/time of the update
Enhance the communication style guidance
- Specify that instructions should use numbered steps with screenshots where helpful
- Request bullet points for key warnings or cautions
- Suggest using bold text for critical information
Add personalization elements
- Request a friendly, reassuring tone to reduce user anxiety
- Include a closing section with company contact information
- Add a clear subject line recommendation for email communications
Incorporate user preparation specifics
- Request specific instructions about backing up critical files
- Include guidance on applications to close before leaving for the day
- Mention what users should do with open documents
Add post-update verification details
- Specify common applications users should test
- Include guidance on what "normal" looks like post-update
Example of the improved prompt:
You are an MSP technician writing instructions for non-technical users about the upcoming [SPECIFIC OS] upgrade scheduled for [DATE/TIME]. Create a clear, friendly communication that reduces anxiety while ensuring users know exactly what to do. Please provide straightforward instructions and answer questions you anticipate users may have.
1. Begin with a brief, non-technical explanation of why this update is important (security improvements, performance enhancements) and reassure users about the process.
2. Include clear pre-update preparation steps, with numbered instructions and screenshots where helpful:
- Specific steps for backing up critical files
- Instructions to save and close all documents before leaving
- Direction to leave computers powered on at the end of the day
- Bold warnings about not interrupting the update process
- A clear timeline of when users should stop working on their systems
3. Explain the approximate duration and what users can expect:
- How long the update will take
- What they might see on their screens when they return
- Any potential temporary performance impacts after the update
4. Provide specific post-update verification steps:
- How to confirm the update completed successfully
- A checklist of common applications to test
- How to verify access to critical systems and data
- What "normal" operation looks like after the update
5. Include a troubleshooting guide for common post-update issues:
- Solutions for 3-5 frequently encountered problems
- Clear indicators of when to stop trying to fix it themselves
- Specific instructions for contacting the helpdesk (phone, email, ticket system)
- Information to provide when reporting an issue
6. End with a friendly closing that includes:
- Your MSP company name and contact information
- Helpdesk hours and response times
- A thank you for their cooperation
If I've attached a file with an example communication, please review it and use its style, formatting, and tone while enhancing it with the above requirements.
Suggested email subject line: "Important: Your Computer Update on [DATE] - Action Required"
3. Simplify
Run the prompt and review the result. Continue to refine the prompt to enhance it and achieve the desired outcome. Focus on simplifying the prompt, as it will allow you to maintain it more easily. Producing results takes less time, and, most likely, the output quality will be much better if the prompt has only the necessary context and instructions.
Before editing the final communication, invest time in enhancing the prompt. You can create a library of prompts and use them with slight modifications. Your prompt library will be your knowledge base for future use. You can reuse elements of prompts for other documents and adjust them as needed for future use.
Treat the process of creating the prompt as training a new employee. You invest time in building high-quality prompts, and then it will pay off with a dramatic increase in productivity.
Conclusions
Prompt engineering is becoming a natural part of life. Simple prompts lead to simple results. Complex prompts with context, tailored for the task, can produce outstanding results and can be reused in the future.
Invest in an AI prompt library, train your team to use AI effectively, establish a knowledge exchange, and you will gain a competitive edge now and stay relevant in the future.