--- title: Healthcare’s Agent Bytecode: X12 as the Harness for AI... description: A conference talk explores how to build reliable AI agent systems for healthcare insurance claims processing, using the X12 EDI standard as a structured... canonical: https://daily.dev/posts/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay | daily.dev og:description: A conference talk explores how to build reliable AI agent systems for healthcare insurance claims processing, using the X12 EDI standard as a structured... og:url: https://daily.dev/posts/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc og:image: https://api.daily.dev/og/posts/dVGIic0yC.png og:image:alt: Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay 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. # Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 20 min read · 0 upvotes · 0 comments ## Summary A conference talk explores how to build reliable AI agent systems for healthcare insurance claims processing, using the X12 EDI standard as a structured 'harness' to constrain agentic reasoning. The speaker argues that pure open-ended agentic execution is too costly, slow, and error-prone across the many steps of a claim lifecycle (eligibility, treatment, submission, payment), while fully hardcoded systems become unmanageable. X12 transactions (270 eligibility, 275 imaging, 999 acknowledgement, 835 payment) provide a public, standardized schema that grounds agent actions and enables rejection of incorrect outputs. The talk also covers multimodal context loss when extracting findings from images before combining with other data, the use of database-backed memory instead of local file memory for enterprise compliance, the risk of memory introducing bias, and the point that insurance-provided data (via X12, browser, or phone) is never guaranteed to be ground truth and must be treated as 'semi-correct until proven otherwise.' ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why use the X12 standard instead of a custom schema for AI agents handling insurance claims? X12 transactions such as 270 for eligibility requests, 275 for imaging, 999 for acknowledgements, and 835 for payment receipts are public and standardized, so any engineer or agent can look up the exact schema instead of inventing one. This gives AI agents a strict, predictable structure to reason within across the many steps of a claim lifecycle, reducing errors compared to an unconstrained custom format an agent might otherwise generate inconsistently. _daily.dev surfaces practical patterns for grounding AI agents in domain standards like X12._ ### Why shouldn't AI agent memory be stored locally like Claude Code or Codex do it? In enterprise healthcare settings, local desktop memory used by tools like Claude Code or Codex isn't viable, so memory is instead stored in a database to maintain logical separation and auditability. This approach still requires balancing helpfulness against bias, since persistent memory across chats and days can wrongly steer a user toward repeating yesterday's task instead of their actual current need. _Developers designing agent memory systems can track these architecture tradeoffs on daily.dev._ ### Can insurance company data always be trusted as accurate when building automated claims systems? No, insurance-provided data delivered through X12, phone calls, or web portals is never guaranteed to be ground truth, even when all channels agree. A patient can be confirmed as covered by phone, browser, and X12 alike, yet the claim can still later be denied for lack of coverage during that period, so internal records should be treated as semi-correct until downstream evidence confirms or contradicts them. _daily.dev helps engineers building claims automation stay grounded in real-world reliability tradeoffs._ ## Similar posts on daily.dev - [Beyond parsing X12: Closing the gap for revenue cycle workflows in healthcare](https://daily.dev/posts/beyond-parsing-x12-closing-the-gap-for-revenue-cycle-workflows-in-healthcare-zucs8ei4r) · databricks · 0 upvotes · 0 comments - [Stedi and EDI as Modern API Workflows](https://daily.dev/posts/stedi-and-edi-as-modern-api-workflows-d3q5sumw9) · API Evangelist · 0 upvotes · 0 comments - [Rehumanizing global health care with agentic AI](https://daily.dev/posts/rehumanizing-global-health-care-with-agentic-ai-xzrofzo6q) · MIT Technology Review · 0 upvotes · 0 comments - [Insurers claim AI is already increasing healthcare costs](https://daily.dev/posts/insurers-claim-ai-is-already-increasing-healthcare-costs-wjxxuuylt) · TechCrunch · 0 upvotes · 0 comments - [Agentic Development: We Handed Off the Code, Not the Job](https://daily.dev/posts/agentic-development-we-handed-off-the-code-not-the-job-2ou02tzdx) · Atomic Spin · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#healthcare](https://daily.dev/tags/healthcare) [View this post on daily.dev](https://daily.dev/posts/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc) ```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":"Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay","url":"https://daily.dev/posts/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc"},"datePublished":"2026-08-19T16:54:27.140Z","dateModified":"2026-08-19T16:54:54.635Z","description":"A conference talk explores how to build reliable AI agent systems for healthcare insurance claims processing, using the X12 EDI standard as a structured...","image":"https://i.ytimg.com/vi/UyyOoJmuATU/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/UyyOoJmuATU/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/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,healthcare","timeRequired":"PT20M","video":{"@type":"VideoObject","name":"Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay","description":"A conference talk explores how to build reliable AI agent systems for healthcare insurance claims processing, using the X12 EDI standard as a structured...","thumbnailUrl":"https://i.ytimg.com/vi/UyyOoJmuATU/sddefault.jpg","uploadDate":"2026-08-19T16:54:27.140Z","duration":"PT20M","url":"https://api.daily.dev/r/dVGIic0yC","embedUrl":"https://www.youtube.com/embed/UyyOoJmuATU"}} {"@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":"Healthcare’s Agent Bytecode: X12 as the Harness for AI Agents — Vasant Kearney, Onlay"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents-vasant-kearney-onlay-dvgiic0yc#faq","mainEntity":[{"@type":"Question","name":"Why use the X12 standard instead of a custom schema for AI agents handling insurance claims?","acceptedAnswer":{"@type":"Answer","text":"X12 transactions such as 270 for eligibility requests, 275 for imaging, 999 for acknowledgements, and 835 for payment receipts are public and standardized, so any engineer or agent can look up the exact schema instead of inventing one. This gives AI agents a strict, predictable structure to reason within across the many steps of a claim lifecycle, reducing errors compared to an unconstrained custom format an agent might otherwise generate inconsistently. daily.dev surfaces practical patterns for grounding AI agents in domain standards like X12."}},{"@type":"Question","name":"Why shouldn't AI agent memory be stored locally like Claude Code or Codex do it?","acceptedAnswer":{"@type":"Answer","text":"In enterprise healthcare settings, local desktop memory used by tools like Claude Code or Codex isn't viable, so memory is instead stored in a database to maintain logical separation and auditability. This approach still requires balancing helpfulness against bias, since persistent memory across chats and days can wrongly steer a user toward repeating yesterday's task instead of their actual current need. Developers designing agent memory systems can track these architecture tradeoffs on daily.dev."}},{"@type":"Question","name":"Can insurance company data always be trusted as accurate when building automated claims systems?","acceptedAnswer":{"@type":"Answer","text":"No, insurance-provided data delivered through X12, phone calls, or web portals is never guaranteed to be ground truth, even when all channels agree. A patient can be confirmed as covered by phone, browser, and X12 alike, yet the claim can still later be denied for lack of coverage during that period, so internal records should be treated as semi-correct until downstream evidence confirms or contradicts them. daily.dev helps engineers building claims automation stay grounded in real-world reliability tradeoffs."}}]} ```