Meet The Company
Geofund is redefining product traceability to operate at a global scale, ensuring the qualities of new innovations like Sustainable Aviation Fuel (SAF) are traceable, verifiable, and tradable.
Founded in 2024 with presence in both US and Europe, Geofund built a platform that solves a fundamental problem that is at the center of the aviation industry’s race to decarbonise.
As airlines face increasing regulatory pressure to use sustainable aviation fuel (SAF), they need a way to prove where their fuel comes from, track its sustainability credentials and manage the complex web of 70+ global certificate programs and regulations that prove compliance.
Think of Geofund as the traceability ledger for aviation fuel. While there are 25 different types of certificate standards that could be tied to aviation fuel, Geofund creates a single source of interoperable truth, connecting fuel suppliers, airlines, and regulators through automated certificate processing and verification and issuance.
The Challenge: Stuck in AI Experimentation Mode
Logan wasn’t new to AI. Like most founders, he had been experimenting and using ChatGPT and Claude daily since launch. But the AI promise for completely taking work off your plate always seemed a little bit out of reach.
While AI was helping Logan increase the quality of his own work, he was still spending as much time cleaning up and restructuring the outputs to align with the business standards and practices he wanted to uphold.
Our Approach: The AI Implementation Hackathon
When Logan reached out, we knew that this could be a perfect opportunity to showcase our new approach to making Notion Agents actually good.
So we blocked out a focused hackathon session where we could work through his real workflows, show him our methodology in action, and give him immediately actionable frameworks.
The goal: Transform Logan from casual AI user to someone who could leverage structured prompting for production work.
Teaching Through Doing
Rather than abstract theory, we worked on Logan’s actual tasks:
- Product Requirements and Issue creation for new Geofund features
- Help documentation based on recent customer onboarding transcripts
- Master Prompt setup for his workspace
- Playbook development for repeatable processes
Each task became a teaching moment for the principles that help turn Notion from a copypaste thing into a true co-pilot.
What We Built Together
1. Structured Prompting Methodology
The problem: Most prompts start as casual one-liners. AI would either ask obvious questions or make wrong assumptions.
What we did instead:
The six-principle framework from our Mastering Notion AI guide:
- Use Claude 4.5 Sonnet (not Auto) for complex reasoning
- Request checklists and plans first before execution
- Create progress tracking pages so AI can recover from interruptions
- Define step-by-step processes to eliminate ambiguity
- Request autonomous completion to avoid unnecessary back-and-forth
- Set clear success criteria so AI knows what “done” means
Real example from the session:
Instead of: “Create a PRD for this feature”
We structured it as:
“Since this is a complex task, I need you to first create a checklist and write yourself a plan on how you want to proceed. Feel free to create a separate tracking page and update it as you work. Interview me if you have questions. Once done, verify your work against your checklist before reporting completion.”
The difference: AI went from producing generic drafts to creating structured, context-aware PRDs that Logan’s engineers could actually use.
2. Voice-First Workflow
The insight: Logan was still typing complex prompts. This is why most people give up on structured prompting, because it just feels too slow.
What we showed him:
Using a tool like Whisper or Monologue by Every means you can speak a 300-word structured prompt in 10 seconds.
3. Help Documentation from Transcripts
The challenge: Logan had detailed customer onboarding call transcripts sitting in Notion, but no systematic way to turn them into help docs.
What we built together:
A workflow to transform a 60-minute transcript into 8 structured help documents:
- Feed the transcript to Claude with clear structure requirements
- Ask AI to identify gaps (“I can extract content for X, but not Y”)
- Generate first drafts with placeholders for screenshots
- Iterate with Loom videos for missing sections
- Store the workflow as a Playbook for the team
Working your way through a bigger task like this will take more time on the first go, but by using the ACDC Framework you can then turn it into a repeatable process.
Assess → Start with real work, not abstract planning
Collaborate → Use the same chat to build context and correct assumptions
Draft → Let AI create the first version
Certify → Review, critique, and capture learnings in a Playbook
The pattern:
“Actually, all my PRDs should follow this structure. Please review this example and update the Playbook to ensure you capture these requirements next time.”
This transforms every task into a training opportunity. The Playbooks get smarter. The outputs get more consistent. And eventually, Logan can delegate these workflows to his team with confidence.
4. Master Prompt + Playbooks System
Breaking down complex workflows into prompts is helpful, but it’s only part of the challenge. There are also plenty of everyday tasks that Notion Agent could handle but where writing a thorough prompt often takes too much time.
Enter a combination of a Master Prompt together with dedicated playbooks for specific tasks. Here’s how this looks like in practice:
Master Prompt → Global instructions for Logan’s AI agent that include:
- Geofund’s workspace structure and databases
- Default behaviors (“never create a ticket without linking to a customer”)
- Tone and formatting preferences
- Common guardrails (“if customer not specified, ask before proceeding”)
Playbooks → Reusable workflows for specific tasks:
- PRD creation with engineering task breakdown
- Help doc generation from transcripts
- Feature specification formatting
Now when Logan asks AI to “create a PRD,” it already knows:
- The expected structure
- Which databases to link
- What information to ask for if missing
- The level of technical detail his engineers expect
Why This Approach Works
Most companies approach AI adoption backwards.
They wait for the “perfect moment” to invest in training. They try to build comprehensive systems before using them. They separate “learning” from “doing.”
We compress the learning curve by teaching through real work:
- Start with urgent tasks → Use what you actually need to get done as training material
- Learn by doing → See structured prompting work on your problems in real-time
- Capture as you go → Turn every correction into a reusable Playbook
- Scale immediately → Apply the same methodology to the next task
This is how you go from AI experimentation to production-ready workflows in hours, not months.
Our AI Implementation Methodology
The Geofund hackathon is a compressed version of what we do in our full 8-Week Notion Transformations.
In larger engagements, we:
- Design AI-ready data architecture from the ground up
- Build Master Prompts tailored to your workspace and workflows
- Create Playbooks for your team’s most time-intensive processes
- Train your team on structured prompting principles
- Implement automated workflows and agent triggers
But the core insight is the same:
AI adoption isn’t about the technology. It’s about teaching your team how to structure their thinking in ways that let AI be genuinely autonomous.
That’s what separates teams who struggle with AI from those who use it to save hundreds of hours.
The Future of Work is Already Here
Logan’s situation wasn’t unique.
Every technical founder, chief of staff, and operations leader we work with faces the same challenge:
They know AI should help.
They’ve tried the casual approach.
But they’re stuck between experimentation and production.
The companies that win won’t be the ones who wait for AI to “get better.”
They’ll be the ones who learn to prompt it properly—right now, with the tools that already exist.
In six months, your competitors will be working completely differently.
They’ll be:
- Making decisions faster because AI surfaces the right context immediately
- Documenting systematically because voice input makes it effortless
- Scaling without proportional headcount because AI handles the operational load
The question isn’t whether this will happen.
It’s whether you’ll be ready.
