Nothing to hand? Load the — a tightly scoped European heat-pump study that should come back ready to collect — the , where leadership wants “a brand analysis” and nothing else is decided, or the , which asks for a report that proves a conclusion it has already reached.
Paste the brief — the scan is free
The browser scores your brief against eight analytical domains using weighted lexicons and reports the margin between the top two, because a brief that reads equally as industry research and investment diligence is a brief that has not decided what it is. It extracts the time range (explicit ranges, relative windows like “the last three years” resolved against today, or bare years), the geographies, the subject candidates, the currency and whether a decision and an audience are actually named. Anything missing becomes a numbered gap with a reason attached. It also ranks a 32-entry framework catalogue against your brief and picks a complementary shortlist — one framework per analytical altitude, never three competitor tools stacked. No upload, no AI, no cost.
The AI plans the study — this is the metered part
A research director's pass: what the subject actually is, which two to four frameworks the study will use and why they are complementary, the chapter skeleton with an objective and one to three falsifiable hypotheses per chapter, and for every chapter the data that must exist before it can be written — each requirement with a priority, a period, source categories and the queries a collector would type. Then a feasibility read on what your constraints make hard, and a position on the brief itself: ready to collect, scope first, or reframe the question. It plans the study; it never invents the answer. Any number you see is either quoted from your brief or labelled as a hypothesis to falsify. Pricing is honest too: a worst-case amount is reserved before the run and only what the run actually uses is charged — the meter next to the button shows both.
Hand it over, tighten it, re-run
Two artifacts leave this page, one for each person who reads a plan. The collectors get the data-requirements table as a CSV worksheet with empty owner, due, source and notes columns already in place and a P0-only cut for the first wave — P0 being the requirements without which a chapter cannot be written at all. The analyst who writes the prose gets the report outline: every chapter as a heading, every hypothesis as a check-box to confirm or refute, a placeholder where each chart goes and an evidence stub per requirement. The chart plan exports as CSV, the whole plan as Markdown or JSON, and any single chapter copies on its own for the ticket. Then the loop: whatever the plan could not settle comes back as a list of questions you answer in place, which fold into the brief as a scoping-decisions block, so the next run plans against your answers instead of asking again. Past plans are saved to your account, not just this browser, and a re-run is compared against your earlier run of the same brief — so when it comes back from the stakeholder review with a decision attached, the list tells you exactly how much the study just shrank.
A derived work of @bytedance/consulting-analysis, the open-source consulting-report skill from ByteDance's DeerFlow project (MIT).