Back office ops · Production

Spotify's Honk background coding agent automates 240 dataset migration PRs, saving an estimated 10 engineering weeks

The problem

Spotify needed to migrate approximately 1,800 direct downstream data pipelines off deprecated user datasets in only six months across three pipeline frameworks — a task estimated at 10 engineering weeks of manual effort.

First attempt

All-in-one prompts for Scio pipelines were abandoned because Scio's lack of standardisation made comprehensive prompts unworkable; initial context files generated from human migration guides were too vague, causing Honk to make incorrect field mapping assumptions.

Workflow diagram · grounded in source
1
Dataset deprecation triggers migration
trigger
“we needed to deprecate two of the most heavily-used user datasets in order to release new versions with additional dimensions that would unlock new features. These deprecated datasets had ~1,800 direct downstream data pipelines between them”
2
Backstage identifies in-scope repositories
integration
“With Codesearch, we wrote queries that would find target repositories across the Spotify GitHub Enterprise landscape, and mark them as in-scope for our migrations”
3
Honk performs code migration
ai_action
“we made all mappings clear using tables in the context file — keeping in mind that Honk could only access the context we had written for it and little else — we began to see solid performance across the majority of target repositories”
4
Judgment-call fields flagged for humans
validation
“in cases where a use case–specific judgement call was required. In these cases, we asked Honk to leave the fields unchanged, but to add comments above them with links to human engineer migration guides to make the task as easy as possibl…”
5
Automated PRs created via Fleetshift
output
“we successfully rolled out 240 automated migration PRs using Fleetshift”
6
Owning teams review and merge PRs
human_review
“we had to rely on the downstream owning teams to perform their own manual testing before merging the automated PRs”
Reported outcome

Spotify successfully rolled out 240 automated migration PRs using Fleetshift, saving an estimated 10 engineering weeks of manual effort.

Reported metrics
Engineering effort saved10 engineering weeks
automated migration PRs rolled out240
Direct downstream data pipelines in scope~1,800
Reported stack
HonkBackstageCodesearchFleet ManagementClaude CodeBigQuery RunnerdbtScioGitHub Enterprise
Source
https://engineering.atspotify.com/2026/4/background-coding-agents-dataset-migrations-honk-part-4
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Spotify successfully rolled out 240 automated migration PRs using Fleetshift, saving an estimated 10 engineering weeks of manual effort.

What tools did this team use?

Honk, Backstage, Codesearch, Fleet Management, Claude Code, BigQuery Runner, dbt, Scio, GitHub Enterprise.

What results were reported?

Engineering effort saved: 10 engineering weeks; automated migration PRs rolled out: 240; Direct downstream data pipelines in scope: ~1,800 (source-reported, not independently verified).

What failed first in this deployment?

All-in-one prompts for Scio pipelines were abandoned because Scio's lack of standardisation made comprehensive prompts unworkable; initial context files generated from human migration guides were too vague, causing Ho…

How is this back office ops AI workflow structured?

Dataset deprecation triggers migration → Backstage identifies in-scope repositories → Honk performs code migration → Judgment-call fields flagged for humans → Automated PRs created via Fleetshift → Owning teams review and merge PRs.

Related back office ops cases