Let me walk you through a project I am working on right now: writing a book.

My actual workflow

I am completing a non-fiction book that is nearing publication. Here is my actual workflow:

Building the reference database. I used Manus and Perplexity to research my subject area and compile a structured database of sources, case studies, and evidence. Manus handled the autonomous multi-site research. Perplexity provided sourced answers I could verify and cross-reference. The result is a comprehensive reference library that I can query at any time.

Drafting content. I write the first draft myself. AI does not write my book for me. I write it.

Refining and polishing. I feed my draft to Claude (inside a dedicated Project that has my style guide, reference material, and chapter structure). Claude suggests refinements: structural improvements, clearer phrasing, gaps in the argument. I take what is useful, ignore what is not, and write a second draft.

Iterative checking. I run my polished draft back against my reference database to verify claims, check consistency, and make sure I have not drifted from the evidence. This loop (draft, refine with Claude, check against references) repeats until the chapter is solid.

Why tools work better together

No single tool does all of this. Perplexity finds information. Manus gathers it at scale. Claude helps me think and write better. My brain makes the decisions. That combination is what makes AI tools genuinely powerful: not any one of them alone, but the workflow you build across them.