Merge's Shensi Ding: AI cuts dinner prep to 30 minutes

Merge's story is less about one tool than about how a small software company spreads AI across roles. Ding uses Codex to review code for a new provisioning tab in Merge for Workforce, while her colleagues lean on GPT‑6 Astra for product testing and marketing. The dinner bot shows the pattern in miniature: Merge's connectors pull guests from the CRM and Apollo, the model drafts outreach, and Ding edits the output and curates the guest list using relationship context the bot lacks.

What stands out is who is doing the building. Ding says the shift has been especially valuable for team members outside engineering, and Charlie Lynch on the sales side built reusable workflows for deal approvals, forecast updates and post-renewal next steps. That fits her leadership advice: executives set the ceiling for AI fluency, teams should experiment somewhere less risky first, and people need room to tinker and make mistakes.

The reported reliability gap gives the rollout a concrete edge. Ding says tool calls through Merge succeed 99.9% of the time, compared with around 60% for public MCP servers, a difference that may matter as companies let AI agents act across business systems rather than just draft text.