Notes from the studio.
- Finance
Worked example: catching out-of-policy expenses
An illustrative worked example of AI expense review that flags out-of-policy spend before approval. The firm and numbers are assumptions, not a client result.
- Operations
GPT-6 astra for operations teams: scope it, don't unleash it
OpenAI shipped GPT-6 astra on 3 Sept 2026. It runs multi-step back-office work on its own. Here is how an ops team scopes one workflow before the switch flips.
- Sales
AI for rfp responses: draft from your own answer library
AI for rfp responses drafts each answer from your own answer library. Your bid lead edits and signs. Nothing is submitted to the buyer without a person.
- Operations
Worked example: prior-auth prep with human sign-off
An illustrative worked example of prior-auth packet prep with a human approving each submission. The group and numbers are assumptions, not a client result.
- Sales
What claudeforce means for revops pipeline hygiene
Salesforce and Anthropic announced claudeforce on aug 26, 2026. For a revops lead: sellers update the pipeline from inside Claude, so the CRM stays current.
- Education
AI for personalized learning paths that lift completion
Adaptive learning paths re-sequence lessons around what each student got wrong, so more of a cohort finishes and passes. No new instructors, no lms rebuild.
- Marketing
Worked example: segment personalization for lifecycle email
An illustrative worked example of segment-level email personalization for a d2c brand. The brand and numbers are assumptions, not a client result.
- Compliance
Claude's memory is on by default: what compliance must review
Aug 25, 2026: Anthropic turned Claude's persistent memory on by default in chat and cowork. For a compliance lead, that is a retention decision, not a feature.
- Finance
AI for bank reconciliation: clear the breaks before close
A controller's guide to AI bank reconciliation. It pairs the routine lines, flags the exceptions, and posts nothing to the ledger without a human sign-off.
- Customer support
Worked example: agent-side retrieval for a support team
An illustrative worked example of agent-side retrieval for a support team. The team and numbers are assumptions, not a client result. No customer-facing bot.
- Product
Claude's watermark is your EU AI act disclosure layer
August 2026: Anthropic now watermarks all Claude output for the EU AI act's article 50 rule. What that hands a product lead shipping a feature into the EU.
- Marketing
AI for competitive intelligence: a weekly rival digest
Stop getting blindsided by a competitor's price or positioning change. A claude-based system that tracks rival moves and ships your team a ranked weekly digest.