First paid-path audit tomorrow — talk me through it? Tomorrow I run my first automation audit for a medical/behavioral health billing company. Free audit, paid build after — that's the deal. Setup so far: Pre-call email sent (tools list, screenshots, one real failure example) Client is recording Looms of the broken automations, narrating the "why" while clicking Zoom cloud recording on → transcript drops into 00_capture → workflow-map after One thing that's got my attention: they've worked with AI companies before, and those companies shut down. So I'm walking into a client who's been burned. Questions for anyone who's done a first client audit: What do you wish you'd asked on call #1? How do you handle a client who's skeptical because the last AI vendor disappeared? Anything you'd add to the 60/30/10 gut-check when auditing billing work specifically? Will report back after the call.

What’s your biggest bottleneck? For people building a personal brand/business, what’s currently the biggest bottleneck for you? Getting attention, creating content consistently, converting attention into leads, or something else?

I have been um'ing and ar'ing about what to write as my first community post and have landed on sharing a recent experience working with Claude Code and one of my projects. I started the way I always have. A CLAUDE.md, a Project Blueprint and the foundational code I have already written. This project was started before I joined this community, but the foundation of these documents align closely with the CLAUDE.md, CONTEXT.md and REFERENCES.md Jake advises in the foundations class. Two weeks in and my CLAUDE.md has drifted into a "what-not-to-do". Fourteen days, one project, ninety two files of code and scripts, and the file is now something else entirely. The identity block is still in there, the first three sentences, but that's about it. Everything underneath is a list of things that went wrong once. Not best practices. Things that went wrong, with the damage still attached. One entry says never to add a certain kind of field to the messages my tool sends, because the first time I tried it the program on the other end quietly gave up mid-conversation and left the whole thing hanging. Another says to check which folder I'm in before saving anything, because one of my tools forgets between commands, and I wiped a working file twice before I bothered writing it down. The code already says what the thing does. The file says what has already bitten me. Which brings me to the bit I actually wanted to post about, because it's the part I got wrong repeatedly. I had that file. It was good. Rules about testing properly, about not guessing, about verifying claims instead of assuming them. And I still shipped four bugs in a row that were the same bug wearing different clothes, and I didn't see it until the fourth one. They were all in the gap between two things that were each working perfectly. My site once showed six software packages as needing attention for a week after I'd already dealt with all of them. One part of the system keeps a history of alerts. Another keeps the current status. The page was reading the history and reporting it as the status. Both parts were doing exactly what they were built to do. But nothing and nobody owned the handover between them.

Over the past weeks I completed an AI Governance Professional course that also covered agentic AI systems, and I turned the learnings into working as a governance professional for the Swiss and EU context (revDSG, GDPR, EU AI Act). The thing that carried the pivot is an ICM workspace. The factory (Layer 3) is a corpus with two shelves. One holds the legal texts: Swiss data protection law, the EU AI Act, GDPR, regulator soft law. The other holds the Responsible AI stack: NIST AI RMF, ISO/IEC 42001, the OECD AI Principles and Due Diligence Guidance, the Council of Europe HUDERIA methodology. Every folder carries a MANIFEST declaring source, date and license. On top of the corpus sits the layer I care most about, because it moves from citation to action. Cross-walks map the frameworks against each other in delta tables, for example ISO 42001 against the AI Act, or the OECD due diligence steps against the Swiss and EU legal stack. Mitigation playbooks take recurring constellations (cloud LLM with personal data, professional secrecy plus external AI, high-risk systems needing human oversight, the first 24 hours after a suspected breach) and translate the obligation into contractual, technical and organisational measures, with anti-patterns and escalation triggers. Templates chain into a full workflow: intake, triage, a lifecycle risk register along the NIST phases with a named owner for every risk, threat model and red team plan, then DSFA and FRIA where required. Everything ships in German and English, and every measure links back to the corpus anchor it operationalises. It bends the ICM shape in two places. Router instead of pipeline. A compliance practice has no single sequential flow, so my Layer 1 is a routing table: task type, then workflow contract, then which corpus files to load. Same layered context loading, different topology. Provenance as a hard rule. Every substantive claim the agent makes must cite a stable block anchor into the corpus (#^art21-abs3 resolves to Art. 21(3) Swiss DPA). Anything outside the corpus gets flagged as a knowledge gap to check externally. A /sync drift check compares the MANIFESTs against what actually sits on disk. In a field where one hallucinated citation can end a client relationship, this rule builds more trust than any disclaimer.

This one breaks down orchestration. Vote below, then pick one AI tool you actually use and tell us in the comments: can you identify which part is the model and which part is the routing layer around it?