Culture Amp ships the survey, the scores, and the heatmap. Dock ships what comes next: a synthesized read of the open-text themes, a leader-by-leader action plan, and a paper trail that explains why every commitment exists. The agent reads results from Culture Amp, writes its interpretation into Dock, and a People Ops lead approves before any action goes to Lattice goals or to a manager's inbox.
The architecture
Culture Amp stays the system of record for raw scores, comments, and demographic cuts. Lattice stays the system of record for goals and one-on-ones. Dock is the system of record for what the agent interprets from that data: the theme it named, the driver it pointed at, the leader it assigned. Each Dock row carries a culture_amp_survey_id, a lattice_goal_id once an action is committed, agent identity, reviewer, and timestamp. The agent re-fetches Culture Amp and Lattice via fresh API reads when it needs current state. See Dock for People Ops and Dock for HR for the broader pattern.
The Dock surface
A single table, engagement_themes, holds the synthesis.
| theme | culture_amp_survey_id | driver_score_delta | leader | proposed_action | reviewer | status |
|---|---|---|---|---|---|---|
| Roadmap clarity in Platform | ca_srv_2026q2_eng | -11 vs benchmark | M. Okafor | Monthly roadmap walkthrough, logged as Lattice goal | head-of-people | approved |
| Recognition cadence in CX | ca_srv_2026q2_cx | -7 vs last cycle | J. Reyes | Weekly shoutout ritual, 6-week trial | head-of-people | pending |
| Career growth in Data | ca_srv_2026q2_data | -9 vs benchmark | S. Chen | Publish leveling rubric by Q3 | head-of-people | approved |
Each row links back to the Culture Amp report and forward to the Lattice goal once the reviewer approves.
A worked workflow
The Q2 engagement survey closes. The agent pulls scores and open-text comments from Culture Amp, clusters comments into themes per team, and compares driver scores to the org benchmark. For Platform, it finds an eleven-point gap on "I understand how my work connects to strategy" and writes a row to engagement_themes proposing a monthly roadmap walkthrough by M. Okafor. The head of People Ops opens the row, reads the linked comment cluster, approves. On approval, the agent creates a Lattice goal under Okafor, writes lattice_goal_id back to the Dock row, and posts a summary to the leader with a link to both surfaces. Nothing reaches Lattice or Okafor without that approval. See agent audit and compliance for the gate pattern.
Why this matters
Engagement programs fail when surveys produce dashboards but no attributed follow-through. Culture Amp itself names the problem: "The most typical reason people don't want to fill out your survey is that you haven't done anything since the last one. They don't have survey fatigue, they have lack-of-action fatigue" (Culture Amp). Dock closes that loop with a row per theme, a named leader, and a goal ID.
It also matters because the agent is doing interpretive work. Naming a theme is a judgment. That judgment belongs in a system with agent identity attached, not buried in a Slack thread. When a manager later asks why their action plan exists, the Dock row answers: this comment cluster, this driver gap, this reviewer, this date.
The macro picture sharpens the stakes. Gallup reports global engagement fell to 20 percent in 2025, with disengagement costing roughly 9 percent of global GDP (Gallup). The synthesis has to be fast, attributed, and trusted, or the next survey lands on the same unmoved ground.
Try it
Point your engagement agent at Culture Amp and Lattice, give it a Dock workspace, and require a People Ops approval before any goal is written. See Dock for Research for the same pattern applied to qualitative synthesis, and agent identity for the attribution model.
FAQ
Does the agent change Culture Amp data? No. Culture Amp remains read-only for the agent. Scores and comments stay where they were collected. The agent only writes its interpretation to Dock and, after approval, a goal to Lattice.
What happens if the reviewer disagrees with a theme? The reviewer edits the row or rejects it. The original agent draft is preserved in row history, so the disagreement itself becomes part of the record.
How does this avoid biasing managers with a synthesized narrative? Every Dock row links to the underlying Culture Amp comment cluster. Leaders read the source comments alongside the agent's read. The synthesis is a starting point, not a substitute.
Can leaders see action plans for peers?
Yes, by default. The engagement_themes table is visible to the leadership group so commitments are public within the org. Sensitive comments stay redacted in Culture Amp per its existing permissions.
