Using multiple AI models as a personal advisory board

Executive overview

Most people rely on a single AI tool, missing out on the distinct strengths each model brings. Different models absorb and process knowledge differently, so querying more than one returns genuinely different perspectives — on ideas, structure, and even phrasing.

Treat AI like a think tank, not a single assistant. The core insight: run your hardest problems through two models, then synthesise the best of both outputs.

Why one AI tool limits you

• Each model has unique strengths; no single one is best at everything
• Multiple models return different opinions, not just different wording
• Language differences alone can surface better phrasing or framing

Strengths of major AI models

• ChatGPT — everyday workhorse; strong at analysis, logic, and maths (especially O1/O3 models)
• Claude — best for writing; outputs sound more human; strong at coding and safety
• Gemini — rapidly improving; currently free; worth experimenting with
• Copilot — well-suited to enterprise environments with access controls

The two-model feedback loop

• Work deeply with one model on a problem, then run the same prompt on a second
• Use the loop for: complex analysis, document structure, pitch decks, pros/cons thinking
• Writing alone rarely warrants two models — reserve it for ideas and structure
• Synthesise the two outputs yourself into a final version

Real-world example: writing a sales proposal

• Start with ChatGPT: discuss goals, customer context, and ask for a structure
• Run the same process in Claude; note where the structures diverge
• Combine the best structural elements from both before drafting
• Repeat for writing: generate paragraphs in each, then select the stronger output

When to stop

• Not every problem needs two AIs — apply the loop only to deep or ambiguous problems
• There is a point of diminishing returns; most users need it only occasionally
• Avoid over-engineering simple writing tasks with multiple-model passes

Getting started

• Sign up for Claude, ChatGPT, and Gemini (all have free tiers)
• Run the same problem through two models and compare the outputs
• Over time, identify which model fits which type of task for your work