peopleanalyst

Tools · Publishing

Publishing

Go from a research dump to a titled, positioned, chapter-by-chapter content plan.

How it works

Three chained-but-independent stages (ideas → concept → outline), each callable alone; status ladder idea→concept→outline; audience/market metadata carried through.

You bring

{ source: deep_research|notes, text, sourceId? } per stage, or the full pipeline

You get

BookIdeaCanonical[] → EnhancedBookConcept → BookOutline (parts/chapters/learning objectives)

Use it for

See it work

example output

Source: a deep-research dump on AI-assisted performance reviews; run through the full ideas → concept → outline pipeline.

Book Concept — The Reviewed Machine: Fair AI in Performance Reviews

Status: outline · Source: deep_research

The idea

A practical playbook for HR and people-analytics leaders deploying AI in performance reviews without importing bias or eroding trust. Pairs the measurement discipline (validity, adverse-impact testing) with the change-management reality of telling a workforce an algorithm now touches their ratings.

Value proposition: the first review-process book written for the team that actually has to defend the model — not the vendor selling it.

Unique selling points

  • A four-test fairness checklist any people team can run before go-live
  • Real adverse-impact math, worked end to end
  • Scripts for the "we're using AI in your review" conversation

Target audience

  • Primary: people-analytics leads + HR directors at 500–5,000-person firms piloting AI in talent processes
  • Secondary: comp & total-rewards teams; employment-law counsel reviewing the rollout

Market opportunity

  • Timing: regulatory pressure (NYC LL144-style audits) is making "we can't explain the model" a liability
  • Trend: every major HRIS is shipping AI review features faster than buyers can govern them

Positioning

Not another "AI will transform HR" book — a field manual for keeping the review process fair when a model is in the loop.

Differentiators: methodology-first, defensible, written for the buyer not the seller.

Outline

Part I — Before You Switch It On

  1. What a review is actually measuringObjectives: name the construct; distinguish behavior from outcome; spot the proxy trap.
  2. The four fairness testsObjectives: run validity, reliability, adverse-impact, and explainability checks on a model.

Part II — Running It in the Open

  1. The disclosure conversationObjectives: script the employee-facing message; pre-empt the trust collapse.
  2. Calibration with a model in the roomObjectives: keep human judgment accountable; audit drift quarterly.

Part III — Defending It

  1. The audit you'll be asked forObjectives: assemble the evidence file; survive a regulator or a grievance.

Two parts, five chapters drafted at outline stage — ready to advance to writing.

Run it on your data

Call it on your own inputs — over the API, or hand it to your AI agent via MCP. Discovery is open; running it is metered.

REST  POST /api/bicycle/publishing/{ideas|concept|outline|pipeline}
MCP   publishing_extract_ideas, publishing_formulate_concept, publishing_generate_outline, publishing_run_pipeline

← All tools