Production-ready Custom GPT instruction sets, knowledge base templates, and frameworks for MSP owners. Built for the real world, not theory.
This repository contains the exclusive materials from the "Developing Custom GPTs with OpenAI's ChatGPT" workshop, designed specifically for MSP and copier dealership leaders.
Problem it solves: Manual workload tracking and client portfolio balancing across your VCIO team
What you get:
- Complete Custom GPT instruction set
- Sample client contract templates (CSV)
- Time tracking templates (CSV)
- Utilization targets and thresholds (TXT)
- Example prompts and expected outputs
Build time: ~90 minutes
ROI: 2-3 hours saved per VCIO per week
Problem it solves: Inconsistent, time-consuming new client onboarding
What you get:
- Starter instruction set (customize for your MSP)
- Service tier definitions template
- Contact and escalation procedures template
- Brand voice guidelines template
- Sample prompts
Build time: ~30 minutes to customize
ROI: 3-4 hours saved per new client onboarding
Problem it solves: Help desk overwhelmed with unsorted tickets, critical issues buried in noise
What you get:
- Starter instruction set (customize for your MSP)
- Priority matrix template
- Common issues quick reference guide
- Sample ticket analysis outputs
Build time: ~30 minutes to customize
ROI: 10+ minutes saved per ticket × 50 tickets/day = massive efficiency gains
Problem it solves: Proposal creation takes 2-4 hours, inconsistent messaging
What you get:
- Starter instruction set (customize for your MSP)
- Service packages and pricing template
- Industry-specific value propositions
- Objection response frameworks
- Sample proposal output
Build time: ~30 minutes to customize
ROI: 2-3 hours saved per proposal
Pick the use case that solves your biggest pain point right now. Don't try to build all four at once.
Every instruction set includes [MSP_NAME] and [CUSTOMIZE_THIS] placeholders. Replace these with your actual company info, service tiers, processes, and terminology.
Critical: Generic instructions produce generic results. The more you customize for YOUR MSP, the better your Custom GPT will perform.
Each scenario includes template files (CSV, TXT, etc.). These are starting points:
- Use the same file formats (CSV for structured data, TXT for context/guidelines)
- Keep column names consistent in CSVs
- Don't overcomplicate - simpler is better
- Go to ChatGPT (Plus or Team/Enterprise account required)
- Create a new Custom GPT
- Paste in your customized instructions
- Upload your knowledge base files
- Test with real scenarios from your business
- Iterate based on results
Custom GPTs aren't "set it and forget it." As you use them:
- Note what works and what doesn't
- Update instructions to fix issues
- Add new knowledge base content as needed
- Test changes before rolling out to your team
✔️ Best formats:
.txtfor guidelines, procedures, context.csvfor structured data (clients, tickets, pricing).docxfor longer documents (if needed)
❌ Avoid:
- Excel workbooks with multiple tabs (GPT struggles with these)
- PDFs with complex formatting
- Images of text (GPT can't reliably extract)
These Custom GPTs use OpenAI's infrastructure. Consider:
- Don't upload real client data for testing (use anonymized or fake data)
- Review OpenAI's data usage policies for your account type
- For highly sensitive use cases, consider alternatives like Claude Projects (we'll discuss in the workshop)
Custom GPTs are powerful but not universal solutions:
- Don't use them for: Real-time automation, API integrations, complex multi-step workflows
- Do use them for: Content generation, analysis, triage, document creation, knowledge base Q&A
If you need actual automation (not just generation), you need AI Agents or workflow tools.
Continue the conversation, share what you're building, and get help from the community:
[Discord invite link - provided separately]
Flag me down while I'm circulating during hands-on time.
- Discord is your main resource
- I'll share additional instruction sets and updates monthly
- Watch for early warnings about OpenAI changes that might break your GPTs
- Pick one use case to start with
- Customize the instruction set for your MSP
- Create at least one knowledge base file
- Build the Custom GPT in ChatGPT
- Test with a real scenario
- Deploy to your team (if it works)
- Share your experience in Discord
Workshop developed and delivered by Jim Haney
CMO, Doceo
MIT Professional Certificate: AI and Digital Transformation
Connect with Jim: jimhaney.link
Remember: The difference between a Custom GPT that works and one that fails comes down to specific instructions, clean data formats, and testing. Don't rush the customization step.
Now go build something useful.