Gridlines: Making expert knowledge accessible
When everything depends on one person, the business becomes harder to run.
Client: Gridlines
Challenge:
The business depended on one person answering the same questions repeatedly, creating constant interruption and pressure.
What we did:
Designed a structured knowledge system and built an AI-supported mentor that reflects the instructor’s method and tone.
Outcome:
Fewer interruptions, reduced dependency, and expertise that scales without the founder needing to be constantly available.
The broader pattern this reveals
When access to knowledge depends on asking the right person, work slows down and the outcome is predictable:
the same questions surface again and again
experienced people spend time answering instead of moving the work forward
progress depends on availability, not clarity
This is one example — but the same pattern appears anywhere knowledge exists without a system that makes it usable.
Why the handbook wasn’t the problem
Kenny, a respected financial modelling instructor, had written a comprehensive handbook.
Four hundred pages of structured methodology, worked examples, and hard-won expertise.
The problem wasn’t the content. It was access.
Students would ask questions that were answered in Chapter 7. Or Chapter 12. Or across three different sections. Kenny found himself answering the same queries repeatedly, pointing people back to material they’d already bought but couldn’t navigate efficiently.
The handbook was valuable. But the knowledge was trapped in a static format.
The real problem (not "we need AI")
Kenny’s first instinct was a common one: "Can we just add a chatbot?"
The real problems were:
Fragmented knowledge retrieval
Students couldn’t find relevant sections when they needed them. Search relied on exact phrases.
Context collapse
Questions often spanned multiple chapters. A chatbot trained on raw text would either miss connections or hallucinate them.
Tone mismatch
Generic AI responses feel generic. Kenny’s teaching style — methodical, encouraging, grounded — was the product. Losing it would undermine the value.
No feedback loop
Kenny had no visibility into what students struggled with, where explanations broke down, or which parts of the handbook needed improvement.
Before building anything, we needed to understand how the handbook actually worked and how students actually learned from it.
How we approached it
We didn’t start by building AI. We started by understanding how the knowledge actually worked.
Mapped the knowledge structure
We analysed the handbook’s architecture: chapters, sections, dependencies between concepts, worked examples and the specific terminology Kenny uses. This wasn’t about "training an AI". It was about understanding what makes the knowledge coherent.
Designed the retrieval system
Rather than dumping everything into a chatbot, we designed a structured knowledge base. Content was broken down so related concepts stayed connected, and each answer retained its context — where it came from and how it related to the wider method.
Built the mentor interface
The system responds in Kenny’s voice, cites specific sections, and guides students back to the material. It behaves like a mentor who knows the handbook — not just a place to look things up.
Added the resource layer
This is where it became usable. The handbook and the Excel models were brought into the same system. Models could be browsed by chapter, difficulty, and type — and each one linked directly back to the relevant explanation.
What it looks like in practice
A student asks a question. The system retrieves the most relevant sections from the handbook and responds in Kenny’s voice - with citations back to the source material.
The response connects concepts, references the broader methodology and maintains the teaching tone throughout.
Students ask questions in natural language and receive structured answers with source citations.
The Excel models aren’t separate from the learning experience. They’re integrated and directly linked to the relevant handbook sections.
Excel models are organised by chapter and linked back to the relevant handbook sections.
The entire system is designed to build independence. Students can verify their own work, find their own answers, and progress without needing constant intervention.
A clear, unified view that brings methodology, mentorship, and resources together.
What changed
Excel models now searchable and linked to handbook content
Access to answers — not just when Kenny is available
Fewer repeated questions Kenny needs to answer manually
More importantly, the handbook is now a living system. Kenny can see what students ask, identify gaps in the material, and improve the content based on real usage patterns. The AI didn’t replace Kenny’s expertise.
It made that expertise available without requiring Kenny to be constantly present.
Why this worked
We didn’t start with "let’s add AI."
We started by asking what valuable knowledge already existed and why the system wasn’t making it usable.
The answer was structural: fragmented access, lost context and no way to connect concepts across a large body of material.
AI only became useful once the system was designed to support it.
This isn’t a chatbot bolted onto a PDF. It’s a knowledge system that happens to use AI under the hood.
Facing a similar challenge?
If important knowledge in your business lives in people’s heads (long documents, or scattered systems) and that creates dependency, interruptions, or friction, we should talk.
We’ll start by understanding how your business actually works,
not by assuming AI is the answer.