@Marigold Henshaw i I had previously built projects in Claude chat and cowork (which were task specific) and in my haste to get things operational (as I also need to get the work done🤪), I think I have certain tasks which are not single exclusive tasks and overlap with others. Therefore it feels clunky to refine the md files. It’s probably just jumping from Claude cowork/chat into code and not really understanding the benefit of keeping tasks simple and exclusive but that I can have more of them - which I can then route. Its all very much a learning curve and hopefully I’ve got the right idea now.

@Marigold Henshaw I provide expert witness reports for personal injury and clinical negligence cases—typically lengthy documents. The main parts of my workflow involves: 1. Documentation Review – I OCR PDFs and extract annotations (formatted to conventions) to create a structured evidence base. 2. Interview – I record conversations, transcribe them, and distil key sections. 3. Examination – Physical assessment with photo documentation for insertion into the report. 4. Evaluation & Recommendations – Synthesise findings and formulate recommendations with costings, illustrations, and clinical rationale. I’ve been using an AI RAG database (Pinecone with Gemini) to use semantic search to find precedent-matching cases for context, and previously used AI Projects to handle customisation of the recommendations based on the current cases presentation. However, the constant file handling and prompt-switching created friction. I need better orchestration across the entire pipeline. Current bottleneck: Formulating recommendations is where time gets lost. That should probably be my immediate focus before optimising upstream processes

It all makes perfect sense and feels like the direction of travel of my workload. Now I need to develop my skills to build a closed-loop! 😁

@David Vogel I really am a beginner with AI, but even with my limited IT systems experience and getting an understanding from here, I seem to be able to get a working system in place that is far better than conventional processes and document management already. Still much improvement needed (from me and my promoting) but working between Claude code and chat (via vscode) already seems like a more natural environment. Excited to see where I can take this👍

Great work and interesting video. I love videos/tutorials that focus on the foundational principles - helps me to retrace back to them when getting distracted by shiny new features! Thanks for your efforts👍

Hey, I’m Scott. 1. I work with medical devices and I provide Medicolegal services in personal injury and clinical negligence cases. 2. I’m interested in Clief notes as when looking for guidance on how to improve my ai workflows, these systems make sense to me despite my non-technical background. 3. Looking to develop my workflows and systems. Currently, I have bits and bobs all over the place and it’s not a cohesive system. I’m looking to start pulling it all together so it’s a more effective and efficient system that grows with my business.

@Jake Van Clief preparing the first draft of the opinion and rationale for the recommendations. I have my reports stored in a vector db (pinecone) and I want to first rationalise my recommendations (brief headers) have Claude go off and find similar recommendations from my existing knowledge base in pinecone and then whilst referencing the context of the findings detailed in the active report, have it interview me to refine and customise the recommendations. The recommendations should come out in a fairly standard format. I’m kind of doing it at the moment but I’d exceptionally clunky and I need to really refine and speed it up