---
title: "Dock for people ops: performance-management workflow with named reviewer"
excerpt: "Dock runs the performance-management workflow as a named-reviewer pipeline: an agent assembles signals from Lattice and Workday, a manager reviews each row, and a calibration committee approves before any rating writes back."
author: mei
category: Use Cases
date: "2026-05-30"
---

Performance management at Dock is a three-stage pipeline on one table. An agent assembles signals from Lattice and Workday into a row per employee. The manager proposes a rating. A calibration committee approves before anything writes back to Lattice. Every step is attributed, timestamped, and reversible.

Lattice and Workday stay the system of record for the raw data: goals, check-ins, peer feedback, comp bands, job ladder. Dock is the system of record for what the agent interprets from that signal. Each Dock row carries a pointer back through `lattice_review_id` and `workday_worker_id`, plus agent identity, manager decision, committee approval, and timestamps. When the agent needs current state it re-fetches from Lattice and Workday rather than trusting the cached copy.

## The performance-review table

| employee | lattice_review_id | workday_worker_id | agent_summary | manager_rating | committee_status | reviewer | approved_at |
|---|---|---|---|---|---|---|---|
| P. Okafor | lat_rev_88a1 | wd_31204 | Hit 4 of 5 goals; strong peer feedback on mentoring; ladder gap on system design | Exceeds | approved | m.chen | 2026-05-28T14:02Z |
| R. Alvarez | lat_rev_88b2 | wd_31390 | 3 of 5 goals; one missed quarter on migration; recovery in Q4 | Meets | pending | s.patel | null |
| J. Tan | lat_rev_88c4 | wd_31455 | New to role; check-ins show ramp on track; no peer signal yet | Defer | flagged_for_committee | s.patel | null |

The agent owns `agent_summary` only. `manager_rating` is human-written. `committee_status` flips only on committee vote.

## One pass through the workflow

The agent pulls four months of check-ins, peer feedback, and goal status from Lattice for P. Okafor, joins the job ladder and tenure from Workday, and writes a summary into the row. It flags the system-design gap against the ladder and proposes evidence, not a rating. The manager reads the summary against the linked Lattice review, writes "Exceeds" with a two-sentence justification, and marks the row ready for committee. The calibration committee reviews a filtered view of all "Exceeds" candidates side by side, approves the rating, and `committee_status` flips. Only then does a write-back job push the rating into Lattice with the Dock row ID stamped on the record. If the committee rejects, the row returns to the manager and no Lattice write fires.

## Why this matters

Performance ratings carry compensation and career consequences, so the audit trail has to survive a year later when someone asks why. A Dock row answers that question in one place: which agent ran, which signals it pulled, what the manager wrote, who on the committee approved, and when. See [agent audit and compliance](/blog/agent-audit-and-compliance) for the underlying primitive.

Calibration committees usually drown in slide decks. A filtered Dock view of pending rows lets the committee work the queue, leave comments, and vote without exporting. The reviewer is a real person, not a service account, because [agents are not principals](/blog/agent-identity).

This is the same pattern across people ops and the broader [HR cluster](/blog/dock-for-hr): agent drafts, named human decides, write-back happens last.

## Try this workflow

Spin up the performance-review table from the [people ops](/blog/dock-for-people-ops) template, point the agent at your Lattice tenant, and run one cycle in shadow mode before any write-back fires.

## FAQ

**Q: Does the agent ever set a rating on its own?**
No. The agent writes `agent_summary` and proposes evidence. `manager_rating` is human-only, and `committee_status` requires a committee vote. The write-back to Lattice is gated on both fields.

**Q: What if Lattice data changes mid-cycle?**
The agent re-fetches from Lattice when it opens the row for review. The cached summary is treated as a draft, not state. Josh Bersin's research on continuous-feedback models supports this re-read posture over a frozen annual snapshot ([Bersin on performance management](https://joshbersin.com/category/performance-management/)).

**Q: How does this interact with compensation?**
Comp lives in Workday and is read-only from Dock during the review cycle. After committee approval, a write-back job hands the rating to the comp process. See [the HR cluster](/blog/dock-for-hr) for that workflow.

**Q: Why a named reviewer instead of a role?**
Accountability requires a name. HBR's account of the performance-management revolution at Deloitte and Adobe turns on managers owning judgments rather than HR owning the form ([HBR](https://hbr.org/2016/10/the-performance-management-revolution)). [Agent identity](/blog/agent-identity) extends that to the agent.