General Management / Strategy

A competitive analysis that took two weeks is framed in two days. Time shifts from collection to interpretation and recommendation.

Claude with web search, market research tools, Perplexity for rapid sourcing.

1. Frame the strategic question in writing before any research: decision to clarify, options, criteria. 2. Have the sourced status report produced, then verify each determining fact at the source. 3. Build scenarios with AI as a challenger: strengths, weaknesses, success conditions of each option.

Any data that changes the recommendation must be verified at the primary source.

Reports, schedules, and updates are generated instead of written. Management refocuses on risks, arbitrations, and people.

Project tool copilots (Asana, Monday, Notion), meeting transcription, Claude for committee summaries.

1. Transcribe project meetings and generate reports and actions immediately. 2. Have the weekly reporting produced from the project tool data. 3. Ask AI each week for the three rising risks of the project based on collected signals.

The generated schedule must be confronted with reality: not all real dependencies are in the tool.

The CEO recovers hours of synthesis and preparation each week. Decisions rely on continuous monitoring and quantified scenarios rather than the last conversation.

Claude or ChatGPT (dedicated project with company context), AI monitoring tool, Notion AI for company memory.

1. Create an AI project with permanent context: strategy, key figures, quarterly priorities. 2. Before each important decision, request three scenarios with risks and quantified assumptions. 3. Have each committee prepared by AI: agenda, document summary, questions to resolve.

AI informs the decision; it never makes it. If the answer is too reassuring, ask for the counter-argument.

The production of notes, reports, and follow-ups goes from half the time to a fraction. The role shifts towards anticipation and real coordination.

Claude (projects by folder), meeting transcription (Fireflies, Granola), decision tracking tool.

1. Record and transcribe each instance, generate decision and action logs. 2. Maintain a decision register in an AI project that can be queried at any time. 3. Prepare arbitrations by having AI confront the positions at stake.

Review each note before distribution: a nuance error at the management level is costly.

Research, benchmarking, and deliverable production are divided by three to five. Billable value shifts towards diagnosis, facilitation, and engaged recommendation.

Claude (one project per mission with client context), AI slide tools, interview transcription.

1. Open each mission with an AI project: brief, client documents, interview reports. 2. Have the first versions of deliverables produced, then elevate them through expertise. 3. Use AI as a sparring partner: counterarguments, blind spots, questions the client will ask.

The client pays for judgment, not the document. Never deliver a text that you wouldn't defend line by line.

Data collection is structured faster, specifications gain completeness, and data can be queried in natural language without waiting for a developer.

Claude for formalization, modeling tools, natural language querying on data.

1. Transcribe business workshops and have needs structured by theme and priority. 2. Have the first version of specifications generated, then challenge it case by case. 3. Validate with the business using generated mockups and examples rather than abstract text.

An unchallenged generated specification is a trap: edge cases determine quality.