We moved in 2 weeks from Hubspot to Attio using dltHub's agentic tranformations

A working student migrated a CRM from HubSpot to Attio in 2 weeks using dltHub's agentic transformation toolkit and Claude Code. The post details an 8-step workflow: stakeholder meeting to define scope, building a machine-readable ontology from the transcript, planning stacked PRs, extracting missing data via REST API toolkit, generating transformations from the ontology, creating mock data for testing, writing regression tests for every bug, and a controlled sandbox-to-production rollout. Key insight: AI agents fail on migrations without three inputs — engineering best practices, actual schema data, and business context (e.g., GDPR constraints). A canonical data model (CDM) acts as a system-neutral intermediary, making future migrations cheaper. Guard rails like local DuckDB dev loops, dry-runs, sync limits, and pytest suites kept the process safe. Human code review remained the bottleneck.