Source before memory
The supplied export, not chat memory or unrelated documents, governed what could be extracted.
AI workflow case study
Knowledge Extraction
A human-governed AI workflow for turning long-form career conversations into structured, traceable knowledge.
I turned one exported chat into a knowledge archive without reducing the original work to a single summary.
01 / Problem
Important career knowledge was buried in conversation history.
Long-running conversations can hold far more than final answers. They also contain the reasoning behind decisions, alternatives that were rejected, methods that changed over time, and questions that were never fully resolved.
Reconstructing the conversation from memory creates a different risk: treating recollection as evidence. A conventional summary can make the material easier to read while diluting the very context needed to use it responsibly.
The goal wasn't to make the conversation shorter.
It was to make its durable knowledge findable, traceable, and usable.
02 / Approach
The source comes before conclusions.
I treated the conversation as an extraction problem before treating it as a writing problem.
Coverage before compression
A lightweight map and focused extraction made it less likely that important reasoning would be lost before the material was distilled.
Extraction before synthesis
I organized and extracted the material before synthesizing it, so later conclusions rested on recovered material rather than an early impression.
03 / Operating model
Human judgment stayed in the loop.
I directed the workflow and made the final decision about how its output could be used.
Owner
Defined the problem and shaped the method.
- Defined the task as extraction, not summary.
- Set the source before extraction began.
- Defined what the source could support.
- Required extraction before synthesis.
- Reviewed and approved the final synthesis.
AI agent
Analyzed the source and drafted outputs.
- Proposed the workflow structure.
- Mapped coverage across the source.
- Extracted material by knowledge class.
- Produced extraction artifacts and a draft synthesis.
- Flagged limitations and unresolved gaps.
04 / Workflow
From source export to reusable knowledge.
The workflow created a deliberate chain from a verified source to an approved synthesis.
- 01
Verify and preserve
Establish the supplied transcript's identity and metadata before processing it.
- 02
Build a coverage map
Create a lightweight outline so the source can be navigated and reviewed as a whole.
- 03
Extract by knowledge class
Use focused passes to separate decisions, methodology, prompts, and philosophy or lessons.
- 04
Synthesize after extraction
Consolidate only after the coverage-oriented passes are complete and their limitations are visible.
Each stage required my approval, from the source through later reuse.
05 / Practical use
The conversation stays useful after it ends.
Together, the outputs preserved decisions and reasoning for later work without needing to reread the full conversation.
Coverage map
A navigable outline of the supplied conversation.
Decision record
Decisions, alternatives, reversals, and open questions.
Methodology evolution record
Methods, changes, rationale, and implementation limits.
Prompt evolution record
Reusable prompt patterns and the reasoning behind them.
Philosophy and lessons record
Principles, caveats, and experience-specific learning.
Final synthesis
Consolidation and routing into later work.
06 / Example
From conversation to reusable record.
This simplified illustration shows how the workflow preserves a decision and its reasoning.
Conversation
User:
How should I capture the perspectives of different reviewers?
Agent:
We could merge them into one conclusion.
User:
But disagreement can reveal hidden assumptions. Let’s capture all of it.
Reusable record
- Decision
- Capture the complete review, including each perspective, agreement, and disagreement.
- Rejected alternative
- Merge all feedback into one conclusion.
- Why it changed
- Disagreement can reveal hidden assumptions.
- Reusable result
- Carry the full review into later work without reconstructing each perspective from memory.
07 / Value
The next question did not start from scratch.
Later, I asked an agent to check the extracted material for gaps in my interview preparation. It drew on different reviewers’ perspectives to identify what to build next and what still needed development.
Instead of starting over, I could build the next stage from what had already been recovered.
08 / Evidence and limits
Evidence and limits belong together.
The project demonstrates a completed workflow cycle.
It does not establish full automation, independent validation, or a universal method.
Demonstrated in this project
- One conversation export was processed.
- I set the direction while AI drafted outputs.
- Five documents captured the source; synthesis brought them together.
- I reviewed and approved the final synthesis.
- The resulting archive supported later work.
Not established
- How completely the process captured the source.
- Repeatability across materially different sources.
- Independent validation of the method.
- A replacement for human judgment.
- External adoption or deployment.
09 / Capabilities
What this work demonstrates.
I used an AI-assisted workflow to preserve decisions and reasoning from a long-form conversation for later use.