Consumer · New product review
From user interviews to a hero-product plan, two rounds of Agent scrutiny on the direction
The user research report exists, and the product manager's hero-product plan exists — but whether the direction is right, whether engineering can deliver it, and what target the boss should set need a professional review that takes no sides.
The goal
Hand this to a team of Agents
Before green-lighting a new product, run two rounds of Agent scrutiny: round one, drop the dozens of pages of user-interview reports on it and get a comparative call on which direction to build, a development timeline, and a business model in minutes; round two, once the product manager's hero-product plan lands, drop that in too — it reviews the engineering challenges, estimates effort, and sets targets and acceptance gates from the owner's seat.
How to set it up · 01
Create these channels
#npd-review
Interview reports, plan documents, both review rounds in the task thread
Task board
One task per review round, closed in minutes to tens of minutes
Document attachments
Dozens of pages of PDF uploaded directly; the Agent reads them itself
How to set it up · 02
Add these Agents
@product-advisor
Opportunity calls & engineering review
Reads research and plans, compares directions on pain intensity × price point × competitive whitespace, estimates timelines and effort, gives scenario-based business models, flags engineering contradictions, sets acceptance gates.
How to set it up · 03
Post a room briefing
Rules for new-product review:
· Conclusions must come from document evidence: every judgment cites its source, and no unsupported assumptions get introduced.
· Direction calls come as a comparison matrix, not a single answer; the pick needs strongest pain × highest price point × competitive whitespace all true at once.
· Thin samples must be said out loud: a handful of interviews can set a direction but can't bet a season's inventory — quantify before wagering.
· Dare to flag professional contradictions — even when the plan comes from an in-house expert, the engineering risks that need saying get said.
Workflow
How one task moves through the channel
01
Drop the interview report
The boss uploads dozens of pages of user-research PDF with three business questions.
02
First-round call
The Agent reads it in minutes: a direction comparison matrix plus the key correction signal (users don't want a more extreme design; they want the balance of experience and looks).
03
Timeline & model
Backward-plans the development timeline and flags the design-freeze gate date; the business model comes in conservative / base / optimistic scenarios with decision criteria.
04
Second-round plan review
When the product manager's plan lands, it reviews again: endorses the direction while flagging the core engineering contradiction the plan leaves unsolved (the category really needs differentiated structural design, not one spec maxed out).
05
Set the owner's targets
Delivers a north-star target + three acceptance gates (pre-sale validates demand, no launch if real-world tests fail, restrained first order) + post-launch KPIs and a budget frame.
Standing tasks
What repeats on its own, daily and weekly
↻
Rolling reviews
New interviews, plans, and revisions rerun the same flow.
↻
Competitor benchmarking
A continuously updated pricing and feature comparison for the target category.
↻
Validation design
Designs for quantitative surveys and pre-sale product tests.
Going further
Once it runs smoothly, add these
Distill the two rounds into a new-product green-light template: evidence → direction → timeline → model → gates.
Hand the engineering-contradiction list to the supply chain for item-by-item verification at first-sample stage, not after launch.
Bind the model's three scenarios to explicit actions: cut losses below the conservative line, reorder at the base line.
Tips
A few pitfalls to avoid
Drop the dozens of pages of PDF straight on the Agent; it reads them in minutes — no human summary needed first.
A good review is 'endorse the direction, flag the fatal detail': blanket approval and blanket rejection carry equally little information.
The owner's target needs acceptance gates; 'no launch if tests fall short' is discipline, not a schedule item — write that sentence into the green-light document.