Description
"These LLMs were optimized for plausibility and not accuracy," says Wangari founder Ari Joury, who learned this the hard way when his company's AI-generated financial commentary delivered a sentence that sounded like consulting speak but meant absolutely nothing, in front of a tier-one insurance client.
Ari brought this lesson to the O'Reilly AI Superstream, tracing the failure to a common mindset problem around treating AI as a magic box rather than an engineering component that needs the same scrutiny as anything else in your system. He covers architectural decay, the pattern where AI-assisted codebases become unreadable within weeks; why the fix requires shifting focus from coding to design; and how tools like LangGraph enable deterministic orchestration while Claude Code handles design and refactoring. He also walks through the multi-agent critique system Wangi built to catch errors before they reach clients, how full data traceability makes every output auditable back to its original source, and how client rejections feed back into the system to improve future output.
Follow O'Reilly on:
LinkedIn: https://www.linkedin.com/company/oreilly/
Facebook: http://facebook.com/OReilly
Instagram: https://www.instagram.com/oreillymedia
BlueSky: https://bsky.app/profile/oreilly.bsky.social
TikTok: https://www.tiktok.com/@oreillymedia