--- title: How to build an AI-Native Health Company — Dan Feng,... description: Maven Clinic's engineering leadership describes their two-year transition to becoming an 'AI native' company, covering internal AI tool adoption, building AI... canonical: https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: How to build an AI-Native Health Company — Dan Feng, Maven Clinic | daily.dev og:description: Maven Clinic's engineering leadership describes their two-year transition to becoming an 'AI native' company, covering internal AI tool adoption, building AI... og:url: https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis og:image: https://api.daily.dev/og/posts/hXirr8uIs.png og:image:alt: How to build an AI-Native Health Company — Dan Feng, Maven Clinic og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # How to build an AI-Native Health Company — Dan Feng, Maven Clinic **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 17 min read · 0 upvotes · 0 comments ## Summary Maven Clinic's engineering leadership describes their two-year transition to becoming an 'AI native' company, covering internal AI tool adoption, building AI into products, and reshaping culture and process. Topics include hiring for AI-fluent engineers who resist delegating implementation work, shrinking planning cycles from months to two-to-four-week sprints, shifting code review practices (self-identified review skip, 500-line PR limits, stacked PRs, avoiding rubber-stamp approvals), and handling LLM hallucination risk through tiered failure tolerance, multi-model verification for high-stakes tasks like reimbursement claims, high-repetition integration testing with pass-rate thresholds, and post-launch conversation auditing. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How should code review change when engineers use AI coding tools to write much more code than before? Allow engineers to self-identify whether a pull request needs review at all, skipping it for simple, low-risk changes while still holding them accountable for what they merge. For PRs that do need review, cap them at 500 lines of code and use stacked PRs to break large features into reviewable chunks, avoiding rubber-stamp approvals that create false confidence. _Teams rethinking review workflows around AI-generated code can follow ongoing coverage of engineering practices on daily.dev._ ### How do you handle LLM hallucination risk in production features with different tolerance levels, like appointment scheduling versus payment processing? Classify failures by acceptable risk: a scheduling feature can tolerate occasional failures since users can simply retry, but a reimbursement claim system cannot tolerate errors because miscalculating an amount causes immediate escalation. High-stakes cases use multiple models to independently verify the same input, such as a receipt, and only proceed when results agree, otherwise routing to a human agent. _Anyone designing safeguards around LLM-driven financial or transactional features can track similar reliability patterns on daily.dev._ ### How can a company test AI-powered software when LLM outputs are non-deterministic? Run each integration test case multiple times rather than once, since large language models can produce different results across runs, and require a consistently high pass rate, such as 90%, across all runs. After launch, an automated evaluation system scores conversations against predefined rubrics, supplemented by a dedicated team that manually reviews a sample, scaling review coverage up to around 20% for newly launched features. _Developers building test strategies for non-deterministic LLM systems can follow related engineering practices on daily.dev._ ## Similar posts on daily.dev - [From one AI product to an AI factory](https://daily.dev/posts/from-one-ai-product-to-an-ai-factory-yrpjourpj) · Medium · 1 upvotes · 0 comments - [AI-Native Leaders: The Organizational Playbook for Engineering Transformation at Scale](https://daily.dev/posts/ai-native-leaders-the-organizational-playbook-for-engineering-transformation-at-scale-rhpzjmdpd) · ByteByteGo · 10 upvotes · 0 comments - [How To Build an AI-First Organization, One Engineer at a Time](https://daily.dev/posts/how-to-build-an-ai-first-organization-one-engineer-at-a-time-qmk9dttgy) · The New Stack · 1 upvotes · 0 comments --- Tags: [#ai-coding](https://daily.dev/tags/ai-coding), [#code-review](https://daily.dev/tags/code-review), [#company-culture](https://daily.dev/tags/company-culture) [View this post on daily.dev](https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"How to build an AI-Native Health Company — Dan Feng, Maven Clinic","url":"https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis"},"datePublished":"2026-08-19T17:54:34.313Z","dateModified":"2026-08-19T17:54:59.598Z","description":"Maven Clinic's engineering leadership describes their two-year transition to becoming an 'AI native' company, covering internal AI tool adoption, building AI...","image":"https://i.ytimg.com/vi/WJRdLNhrsLQ/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/WJRdLNhrsLQ/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-coding,code-review,company-culture","timeRequired":"PT17M","video":{"@type":"VideoObject","name":"How to build an AI-Native Health Company — Dan Feng, Maven Clinic","description":"Maven Clinic's engineering leadership describes their two-year transition to becoming an 'AI native' company, covering internal AI tool adoption, building AI...","thumbnailUrl":"https://i.ytimg.com/vi/WJRdLNhrsLQ/sddefault.jpg","uploadDate":"2026-08-19T17:54:34.313Z","duration":"PT17M","url":"https://api.daily.dev/r/hXirr8uIs","embedUrl":"https://www.youtube.com/embed/WJRdLNhrsLQ"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"How to build an AI-Native Health Company — Dan Feng, Maven Clinic"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/how-to-build-an-ai-native-health-company-dan-feng-maven-clinic-hxirr8uis#faq","mainEntity":[{"@type":"Question","name":"How should code review change when engineers use AI coding tools to write much more code than before?","acceptedAnswer":{"@type":"Answer","text":"Allow engineers to self-identify whether a pull request needs review at all, skipping it for simple, low-risk changes while still holding them accountable for what they merge. For PRs that do need review, cap them at 500 lines of code and use stacked PRs to break large features into reviewable chunks, avoiding rubber-stamp approvals that create false confidence. Teams rethinking review workflows around AI-generated code can follow ongoing coverage of engineering practices on daily.dev."}},{"@type":"Question","name":"How do you handle LLM hallucination risk in production features with different tolerance levels, like appointment scheduling versus payment processing?","acceptedAnswer":{"@type":"Answer","text":"Classify failures by acceptable risk: a scheduling feature can tolerate occasional failures since users can simply retry, but a reimbursement claim system cannot tolerate errors because miscalculating an amount causes immediate escalation. High-stakes cases use multiple models to independently verify the same input, such as a receipt, and only proceed when results agree, otherwise routing to a human agent. Anyone designing safeguards around LLM-driven financial or transactional features can track similar reliability patterns on daily.dev."}},{"@type":"Question","name":"How can a company test AI-powered software when LLM outputs are non-deterministic?","acceptedAnswer":{"@type":"Answer","text":"Run each integration test case multiple times rather than once, since large language models can produce different results across runs, and require a consistently high pass rate, such as 90%, across all runs. After launch, an automated evaluation system scores conversations against predefined rubrics, supplemented by a dedicated team that manually reviews a sample, scaling review coverage up to around 20% for newly launched features. Developers building test strategies for non-deterministic LLM systems can follow related engineering practices on daily.dev."}}]} ```