Back office ops · Production

Intuit's platform-centric AI code generation achieves 56% faster PR merges and 3x faster backend generation

The problem

Off-the-shelf AI coding assistants lacked awareness of Intuit-specific APIs, architectural conventions, code quality standards, and compliance requirements, causing suggestions to require heavy manual editing and usage to decline especially among senior developers.

First attempt

Generic IDE-based AI coding assistants generated code that did not conform to Intuit's standards, requiring developers to examine output, check against existing code, and manually adjust each component — and sometimes the generated code was outright incorrect.

Workflow diagram · grounded in source
1
Developer submits context-aware prompt
trigger
“Generate a loan eligibility widget by following Intuit Best Practices and coding patterns.”
2
Intuit Context injection from golden repos
integration
“injecting internal knowledge (dubbed the Intuit Context) into the development experience, served in real time by our scalable backend with context-enriched query pipelines”
3
Context-aware code generation
ai_action
“the coding assistant has the foresight of pulling in the relevant Intuit Design System and AppFabric libraries and building the form accordingly”
4
Developer reviews AI-generated tests
human_review
“58% of AI-generated tests are used without modification, after review”
5
PR creation and merge
output
“Engineers using AI-assisted workflows merge PRs 56% faster”
Reported outcome

With context-aware AI grounded in golden repositories, 58% of AI-generated tests are accepted without modification, PR merges are 56% faster, backend code generation time dropped by 3x, and frontend generation tasks improved by over 10x.

Reported metrics
AI-generated tests accepted without modification58%
PR merge speed improvement56% faster
Backend code generation timedropped by 3×
Frontend generation task improvementover 10×
Show all 5 reported metrics
AI-generated tests accepted without modification58%
PR merge speed improvement56% faster
Backend code generation timedropped by 3×
Frontend generation task improvementover 10×
Developer velocity8x
Reported stack
LLMsIntuit Design SystemAppFabric
Source
https://medium.com/intuit-engineering/a-platform-centric-approach-to-ai-assisted-code-generation-at-intuit-03984a85558e
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

With context-aware AI grounded in golden repositories, 58% of AI-generated tests are accepted without modification, PR merges are 56% faster, backend code generation time dropped by 3x, and frontend generation tasks i…

What tools did this team use?

LLMs, Intuit Design System, AppFabric.

What results were reported?

AI-generated tests accepted without modification: 58%; PR merge speed improvement: 56% faster; Backend code generation time: dropped by 3×; Frontend generation task improvement: over 10× (source-reported, not independently verified).

What failed first in this deployment?

Generic IDE-based AI coding assistants generated code that did not conform to Intuit's standards, requiring developers to examine output, check against existing code, and manually adjust each component — and sometimes…

How is this back office ops AI workflow structured?

Developer submits context-aware prompt → Intuit Context injection from golden repos → Context-aware code generation → Developer reviews AI-generated tests → PR creation and merge.