
A travel and convenience retail chain cuts dbt Platform migration effort by 70% with AI
A travel and convenience retail chain operating 650+ locations across 42 states migrated roughly 85 dbt Core projects onto dbt Platform, meeting the engagement’s hard deadline. An AI-assisted migration script built by phData cut the manual migration effort per project by roughly 70%.
At a glance
| Industry | Travel and convenience retail |
| Scale | 650+ locations across 42 states |
| Challenge | Migrate a dbt Core estate to dbt Platform by a hard deadline without manual, one-by-one conversion work consuming the data team’s year, while also modernizing a sprawling, duplicated Terraform footprint. |
| Technology | dbt Core, dbt Platform, Snowflake, Snowflake Cortex, Terraform, Atmos, GitHub Action, Anthropic Claude |
| Timeline | 6 months |
| Results | Roughly 70% reduction in per-project migration effort of 85 projects without missing a hard deadline |
| Phdata Service | Data Engineering, dbt, Anthropic |
How to migrate a growing dbt Core estate to dbt Platform without missing the deadline
Many enterprises got their start with dbt Core, but when they try to migrate to dbt Platform, they discover the conversion process isn’t straightforward, especially for early adopters who have accumulated large numbers of models and projects. A travel and convenience retail chain operating 650+ locations across 42 states hit that wall with a dbt Core estate of roughly 85 legacy projects, with each project requiring about 4 hours of conversion work, 2 hours of testing and fine-tuning, and 1 hour of deployment. A fully manual approach across an estate that size risked missing the project’s fixed deadline, while the team’s existing Terraform footprint, spread across 70+ separate repositories, was becoming just as hard to maintain.
dbt Core is the free, open-source, command-line version of dbt that a team installs and runs on its own infrastructure.
dbt Platform is dbt Labs’ managed platform that adds hosted orchestration, scheduling, environment management, and built-in CI/CD on top of dbt project code.
Snowflake Cortex is Snowflake’s built-in suite of AI functions that can generate text, such as draft documentation, directly against data stored in Snowflake.
Atmos is an open-source framework that organizes Terraform configuration into reusable, YAML-defined components shared across a single infrastructure repository.
What phData did
Four decisions shaped the delivery.
Built a Python migration app with AI-assisted development instead of converting each project by hand
phData used AI-assisted development (AIAD), including Anthropic’s Claude, to build the Meshify Migration Assistant, a Python application, rather than converting dozens of dbt projects to dbt Platform manually, one at a time. The client needed a process that was auditable, debuggable, repeatable, and lossless, so the app’s core conversion logic stayed deterministic Python with testable output, and AIAD cut the time it took to build it.
Routed rare edge cases to dbt Wizard instead of expanding the migration script
Rather than growing the core script to handle every unusual, one-off project quirk, phData used dbt Wizard to resolve the long tail of edge cases the script did not cover. Expanding the script for rare exceptions would have made it harder to test and trust, so the team kept the deterministic path narrow and routed exceptions to a separate tool built for that purpose.
Integrated Snowflake Cortex into the migration tool instead of building a separate documentation tool
phData integrated Snowflake Cortex directly into the Meshify Migration Assistant to draft descriptions for thousands of dbt models and columns, using each model’s table structure, column types, and sample values as prompt context. Because Cortex runs natively inside Snowflake, that sample data never left Snowflake’s secure perimeter. Every AI-generated draft was explicitly labeled as AI-generated, so the client gained documentation coverage it did not have time to write manually, without exposing data outside Snowflake.
Rebuilt Terraform infrastructure alongside the dbt migration instead of treating it as a separate project
phData consolidated the client’s 70+ Terraform repositories into a single Atmos mono-repo in parallel with the dbt migration, standardizing resource tagging and replacing statically provisioned resources with reusable components. Running both efforts together let the team retire duplicated code and inconsistent configuration on the same timeline instead of leaving infrastructure modernization for a future engagement.
Migrating an 85-project dbt estate to dbt Platform without missing the deadline
phData’s Meshify Migration Assistant turned a manual, project-by-project slog into a repeatable, automated process, cutting per-project migration effort by 70% across 85 dbt projects. phData still delivered against a hard deadline. The migration script’s output is deterministic and testable, and every AI-generated documentation draft is clearly labeled as such, so the client’s data team can trust what the automation produced without re-checking it line by line.
Manage models through dbt Platform’s hosted orchestration and CI/CD instead of maintaining a separate process for every legacy dbt Core repo.
Rely on documented model and column descriptions for thousands of dbt objects instead of tribal knowledge.
Deploy infrastructure changes through a single Atmos mono-repo with automated CI checks instead of tracking 70+ separate Terraform repos.
Treat the migration script’s output as production-ready, since every conversion is deterministic and testable rather than manually re-verified.
reduction in per-project dbt Core to dbt Platform migration effort.
dbt projects migrated to dbt Platform.
Ready to see what a similar dbt Platform migration could look like for your data estate?
Why this matters beyond this project
Any enterprise sitting on a large, fragmented dbt Core estate faces the same math: manual, project-by-project migration does not scale, and every hour spent converting boilerplate code is an hour not spent on higher-value engineering work. The fix here, deterministic code for the repeatable majority and AI reserved for building tooling faster and drafting documentation at scale, applies to any team running a large-scale dbt, ETL, or infrastructure migration, and it proves AI can meaningfully cut migration effort without eliminating the testing and edge-case work that still needs a person’s judgment.
Related customer stories

B2B Software Company Modernizes Data Stack by Implementing Snowflake, Airflow, and dbt
B2B SaaS company modernizes its data stack by working with phData to implement Snowflake, Airflow, and dbt.
Frequently asked questions
Can AI automate a dbt Core migration to dbt Platform?
AI can meaningfully cut the manual effort in a dbt Core to dbt Platform migration when it is used to speed up a deterministic application rather than to make conversion decisions itself. In this engagement, the Python application phData built with AI-assisted development (AIAD) cut per-project migration effort by roughly 70%, while testing and a long tail of edge cases still required multiple hours of human effort per project.
Is it safe to use AI-generated documentation for production dbt models?
It is safe when every AI-generated draft is clearly labeled as AI-generated so subject-matter experts can flag or correct anything inaccurate. phData integrated Snowflake Cortex directly into the migration tool to draft descriptions for thousands of dbt models and columns, keeping that sample data inside Snowflake's secure perimeter rather than sending it to an external service.
What technologies are involved in a dbt Core to dbt Platform migration?
A typical migration involves dbt Core, dbt Platform, the underlying data warehouse, and an infrastructure-as-code tool for access and environment management. This engagement used dbt Core, dbt Platform, Snowflake, Snowflake Cortex, Terraform, Atmos, GitHub CI/CD, and Anthropic Claude for migration automation.
How does phData make sure AI-assisted migration output can be trusted?
phData keeps the core conversion logic in deterministic, testable code and limits AI to two roles: building the migration tool faster and drafting documentation content that is clearly labeled as AI-generated. Rare edge cases and long-tail testing are routed to dbt Wizard and human judgment instead of expanding the automated application.
Ready to migrate your dbt Core estate to dbt Platform?
Share your goals with us. A phData expert will follow up to discuss next steps.
- Free scoping conversation, no commitment
- Certified data and AI engineers
- Proven delivery

