{"id":35057,"date":"2025-08-17T13:27:03","date_gmt":"2025-08-17T11:27:03","guid":{"rendered":"https:\/\/www.graviton.at\/letterswaplibrary\/how-do-you-collect-and-structure-data-for-an-ai-after-sales-sav-agent-in-banking-insurance\/"},"modified":"2025-08-17T13:27:03","modified_gmt":"2025-08-17T11:27:03","slug":"how-do-you-collect-and-structure-data-for-an-ai-after-sales-sav-agent-in-banking-insurance","status":"publish","type":"post","link":"https:\/\/www.graviton.at\/letterswaplibrary\/how-do-you-collect-and-structure-data-for-an-ai-after-sales-sav-agent-in-banking-insurance\/","title":{"rendered":"How Do You Collect And Structure Data For An AI After-sales (SAV) Agent In Banking\/insurance?"},"content":{"rendered":"<p><!-- SC_OFF --><\/p>\n<div class=\"md\">\n<p>Hey everyone,<\/p>\n<p>I\u2019m an intern at a new AI startup, and my current task is to <strong>collect, store, and organize data<\/strong> for a project where the end goal is to build an <em>archetype after-sales (SAV) agent<\/em> for financial institutions.<\/p>\n<p>I\u2019m focusing on <strong>3 banks<\/strong> and an <strong>insurance company<\/strong> . My first step was scraping their websites, mainly <strong>FAQ pages<\/strong> and <strong>product descriptions<\/strong> (loans, cards, accounts, insurance policies). The problem is:<\/p>\n<ul>\n<li>Their websites are often outdated, with little useful product\/service info.<\/li>\n<li>Most of the content is just <strong>news, press releases, and conferences<\/strong> (which seems irrelevant for an after-sales agent).<\/li>\n<li>Their social media is also mostly marketing and event announcements.<\/li>\n<\/ul>\n<p>This left me with a <strong>small and incomplete dataset<\/strong> that doesn\u2019t look sufficient for training a useful customer support AI. When I raised this, my supervisor suggested scraping <em>everything<\/em> (history, news, events, conferences), but I\u2019m not convinced that this is valuable for a <strong>customer-facing SAV agent<\/strong>.<\/p>\n<p>So my questions are:<\/p>\n<ul>\n<li><strong>What kinds of data do people usually collect to build an AI agent for after-sales service (in banking\/insurance)?<\/strong><\/li>\n<li>How is this data typically <strong>organized\/divided<\/strong> (e.g., FAQs, workflows, escalation cases)?<\/li>\n<li>Where else (beyond the official sites) should I look for <em>useful, domain-specific data<\/em> that actually helps the AI answer real customer questions?<\/li>\n<\/ul>\n<p>Any advice, examples, or references would be hugely appreciated .<\/p>\n<\/div>\n<p><!-- SC_ON -->   submitted by   <a href=\"https:\/\/www.reddit.com\/user\/seriousdeadmen47\"> \/u\/seriousdeadmen47 <\/a> <br \/> <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1msobmk\/how_do_you_collect_and_structure_data_for_an_ai\/\">[link]<\/a><\/span>   <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1msobmk\/how_do_you_collect_and_structure_data_for_an_ai\/\">[comments]<\/a><\/span><\/p><div class='watch-action'><div class='watch-position align-right'><div class='action-like'><a class='lbg-style1 like-35057 jlk' href='javascript:void(0)' data-task='like' data-post_id='35057' data-nonce='65e0e39b87' rel='nofollow'><img class='wti-pixel' src='https:\/\/www.graviton.at\/letterswaplibrary\/wp-content\/plugins\/wti-like-post\/images\/pixel.gif' title='Like' \/><span class='lc-35057 lc'>0<\/span><\/a><\/div><\/div> <div class='status-35057 status align-right'><\/div><\/div><div class='wti-clear'><\/div>","protected":false},"excerpt":{"rendered":"<p>Hey everyone, I\u2019m an intern at a new AI startup, and my current task is to collect,&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[85],"tags":[],"class_list":["post-35057","post","type-post","status-publish","format-standard","hentry","category-datatards","wpcat-85-id"],"_links":{"self":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/35057","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/comments?post=35057"}],"version-history":[{"count":0,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/35057\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/media?parent=35057"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/categories?post=35057"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/tags?post=35057"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}