Detailed Summary: AI Explored Podcast

Episode: Beyond Rigid Automation: How Custom GPTs Add Flexibility to Your Workflows
Host: Michael Stelzner
Guest: Isar Matis, AI Strategist and Founder of Multiply
Date: December 9, 2025

Episode Overview

This episode focuses on how integrating custom GPTs (AI agents) into workflow automation tools can transform rigid, rule-based automations into highly flexible, intelligent systems for marketers, creators, and business owners. Michael Stelzner and guest Isar Matis break down the misconceptions of AI automation, illustrate practical use cases, and offer step-by-step advice on implementing AI assistants with vector stores for dynamic, scalable automation. The discussion is peppered with actionable examples and advice for both beginners and advanced users.

Key Discussion Points & Insights

1. Debunking Workflow Automation Myths ([01:57]–[03:29])

• Misconception 1: You must be a software engineer to automate workflows
  • “These tools are becoming more and more user friendly and you can do amazing things with them with zero knowledge in computers just by understanding the business processes.” – Isar Matis [02:15]
• Misconception 2: Automations are inflexible and can only run static, rule-based tasks
  • With AI, automations can now handle nuanced decision-making previously requiring human intervention.

2. Evolution of Automation Tools ([03:45]–[04:41])

• Early tools like Zapier were data movers—taking info from one app to another.
• Now AI can perform contextual analysis (e.g., categorizing, summarizing, qualifying leads), transforming static processes into intelligent ones.
  • “The ability now to actually know what the data is and act based on that data… these are the magical things that you can do now with AI.” – Isar Matis [04:05]

3. Real-World Example: Shopify B2B Lead Qualification ([05:00]–[08:50])

• Use Case: Previously, a staffer manually sifted through Shopify B2C orders to identify hidden B2B opportunities—rarely completed due to workload.
• Automated Solution:
  • AI automatically analyzes orders, researches companies and contacts, determines B2B fit, then either:
    • Sends enriched info via Slack to staff, or
    • Drafts targeted outreach emails for the relevant salesperson.
  • “Now you don’t need the human in those steps.” – Isar Matis [08:28]

4. Cost-Effectiveness of AI Automations ([08:50]–[11:19])

• AI-powered automation reduces staff time and costs.
• Critical tip: Use AI 'assistants' with built-in instructions for efficiency—send only necessary data, not full instruction sets—to minimize token (API cost) usage.

5. Automation Tools Breakdown ([14:27]–[16:14])

• Recommended Tools:
  • Make.com: Easiest for beginners; graphical, visual builder. Great for most standard automations.
  • n8n: Open-source, more advanced; best for complex needs or when scaling lots of automations cost-effectively (flat hosting fee).
    • “If you’re good and it does everything you need, just stay with Make. The only reason to switch… is for more sophistication or cost.” – Isar Matis [15:28]
  • Most tools already have prebuilt connectors for all major apps (CRMs, email, etc.).

6. How AI Custom GPTs Fit In ([22:18]–[25:17])

• Custom GPTs: Repeatable AI agents for standardized tasks (e.g., sales-marketing report comparison, hook generation, proposal writing).
• Each custom GPT includes: an input, step-by-step instructions, and a defined output.
• Problem: Custom GPTs normally operate in a silo (manual copy-paste required).

7. OpenAI Assistants & Vector Stores Explained ([26:04]–[39:24])

• OpenAI Assistants: Backend (API-based) versions of custom GPTs that can connect to other apps—ideal for automated workflows.
• Knowledge Base: Data/files added to assistants to enable contextually rich outputs.
  • Can include templates, style guides, case studies, etc.
• Vector Store: Special OpenAI folder/database to dynamically update the knowledge base via API during automations (bridge between automation tools and AI agent).
  • “The vector store has the ability to connect to third party processes… I can dynamically change the knowledge base and/or the inputs.” – Isar Matis [33:56]
• Example Workflow:
  1. File drops into Google Drive → triggers Make to upload file to Vector Store
  2. Assistant receives new knowledge → generates quiz questions or repurposes content
  3. Automation saves AI output in a new location or routes it for approval

8. Marketing Example: Dynamic Content Repurposing ([39:43]–[42:36])

• Workflow: Blog post or podcast is published → automation grabs content → vector store inputs to assistant → AI writes LinkedIn posts, scripts, etc.
• Human “approval” step can be built in to ensure output quality before publishing.

9. Conditional Outputs & AI in the Loop ([42:36]–[45:17])

• AI assistants can return variables, scores, or branching results to drive logic in workflows (e.g., lead qualification, content scoring).
• Automations can include “if” branches for customized action paths based on AI output.

10. Ensuring Reliability and Quality ([45:17]–[47:08])

• Best practices:
  • Insert manual “human in the loop” approvals at critical workflow points.
  • Rigorously test and fine-tune prompting—instructions are key.
  • Use AI as a QA checker for other AIs if desired.
  • Develop instructions collaboratively with AI for optimal clarity.

Notable Quotes & Memorable Moments

• On Automation Accessibility:
  “These tools are becoming more and more user friendly and you can do amazing things with them with zero knowledge in computers just by understanding the business processes.” ‒ Isar Matis [02:15]

• On the New Flexibility:
  “Every time you needed a human to evaluate the data, do some calculation… now you don’t need the human in those steps.” – Isar Matis [08:28]

• On Cost:
  “The biggest bill you’ve ever had? Less than 20 bucks.” – Michael Stelzner [27:50]

• On Dynamic Knowledge:
  “I can take any file or any information from my workflow automation… and upload it into the vector store, which means now I can dynamically change the knowledge base.” – Isar Matis [33:56]

• On Workflow Approvals:
  “You don’t need the human in the process, but you want the human in the process… Add a step that connects it to your regular task management system.” – Isar Matis [41:08]

Timeline of Important Segments

| Timestamp | Segment | |-----------|--------------------------------------------------------------------------------------| | 01:57 | Myths about workflow automation (not just for programmers, not always rigid) | | 05:00 | Shopify B2B lead automation case study | | 11:19 | Basic workflow automation explained for non-techies | | 14:27 | Key tools: Make.com vs. n8n (pros, cons, costs) | | 22:18 | Transition to AI custom GPTs and backend assistants | | 26:09 | How to create OpenAI assistants for workflow automation | | 32:00 | Vector store explained; dynamic vs. static knowledge in AI automations | | 33:56 | Dynamic knowledge base automation example (quiz generation) | | 39:43 | Marketing example: content repurposing workflows using assistants/vector store | | 42:36 | Conditional branching and smart outputs in AI workflows | | 45:17 | Ensuring prompt compliance and using AIs for QA or human check-points | | 47:17 | Where to find Isar Matis’ courses, podcast, info |

Resources & Where to Find the Guest

• Podcast: Leveraging AI (search in your podcast player)
• Courses: Learn more at the links in the Social Media Examiner show notes; use code AIXPLORED for a discount
• LinkedIn: Isar Matis – "the only Isar Matis on LinkedIn!"

Final Takeaway

This episode provides a roadmap for moving beyond rigid, rule-based automations with the latest in custom AI assistants powered by OpenAI and other LLM platforms. By mastering tools like Make, n8n, and OpenAI assistants with vector stores, marketers and business owners can delegate decisions, analysis, content creation, and more to AI — safely, reliably, and at minimal cost. Flexibility and scale are in reach for anyone willing to implement and experiment.