ManagerOS.

The problem

People managers in large enterprises are doing the hardest job they were never trained for, on top of their "real" technical or functional role. Notes from 1:1s end up scattered or never re-read; action items fall through the cracks (damaging manager credibility); review season triggers a panic of memory-based prose; reviews use vague language that fails calibration; and the manager has no view of which direct reports they've systematically under-discussed.

The downstream cost is board-level: manager effectiveness has become a post-pandemic priority — quiet quitting, retention costs, hybrid work complexity — and every CHRO knows the existing performance management tools are documentation-first, not effectiveness-first.

The solution

ManagerOS sits next to the manager as a Chief of Staff. After every 1:1, the manager pastes the notes; the AI extracts action items with owners and due dates, updates the contribution log, and prepares the briefing card for the next 1:1 — open ARs, suggested agenda items, contribution highlights, recent themes.

When review season hits, the same notes feed a Review Writer + Coach that drafts from accumulated evidence and gives quality feedback on the draft — flagging bias, gendered language, and vague phrases that fail calibration. The product never positions itself as a cheerleader: the system prompt frames the AI as a seasoned Chief of Staff who has seen what good and bad management look like.

How it works

Flow: Team Setup once → recurring 1:1 prep (briefing card generated before each meeting) → post-1:1 notes pasted → AI extracts ARs → ARs tracked across employees → at review season, accumulated notes power the Review Drafter + Coach → bias guardrails flag risky language.

Edge cases tested: ambiguous AR ownership ("We need to fix the onboarding doc" with no clear owner) → the agent asks rather than assumes; no-AR meetings → the agent surfaces themes instead of inventing tasks; review drafts with vague phrases → the coach flags them with specific rewrites.

Primary model: GPT-4o — selected for the 128K context window, strong unstructured-input handling (messy meeting notes), and reliable JSON output for AR extraction. Deferred to V2: blind spot detection (needs multi-week data), calibration prep (needs full-quarter context), RAG HR Q&A.

Who it's for

Primary end-user: the Internal People Manager — managing 5–15 direct reports in a 5,000+ employee enterprise, time-starved, undertrained on people management, doing 1:1s and reviews on top of their technical or functional job. They don't buy the tool, but they make or break adoption — if managers don't use it, the buyer churns.

Budget-holder buyers: CHRO, VP of HR, HR Technology leaders, sometimes IT/Procurement. They care about ROI, data security, compliance, and adoption rates — the latter especially, because every HR-tech tool that ships goes through an adoption death-march.

Why it matters

HR Tech is undergoing a generational shift from systems of record to systems of intelligence — and the performance management category is where that shift will land hardest. The buyers have committed (manager effectiveness is a board topic), the budget is moving (AI-in-HR at 24–27% CAGR), and the existing tools (Lattice, Culture Amp, 15Five) are documentation-first by design — they help managers record but not think.

ManagerOS's wedge is the 1:1 Copilot data layer: once the AI has a quarter of structured notes per direct report, every other use case (reviews, calibration, blind spot detection, HR Q&A) becomes trivially valuable. The bet is that the next decade of enterprise HR tech is one structured AI conversation at a time, not another form to fill in.

At a glance

Project

ManagerOS

Built by

Anusha Muthiah

One-liner

AI Chief of Staff for enterprise people managers — 1:1 copilot, action item tracking, blind spot alerts, and review drafting with coaching.