1(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[2573],{4184:function(e,t){var s;!function(){"use strict";var a={}.hasOwnProperty;function i(){for(var e=[],t=0;t<arguments.length;t++){var s=arguments[t];if(s){var r=typeof s;if("string"===r||"number"===r)e.push(s);else if(Array.isArray(s)){if(s.length){var n=i.apply(null,s);n&&e.push(n)}}else if("object"===r){if(s.toString!==Object.prototype.toString&&!s.toString.toString().includes("[native code]")){e.push(s.toString());continue}for(var o in s)a.call(s,o)&&s[o]&&e.push(o)}}}return e.join(" ")}e.exports?(i.default=i,e.exports=i):void 0===(s=function(){return i}.apply(t,[]))||(e.exports=s)}()},4166:function(e,t,s){(window.__NEXT_P=window.__NEXT_P||[]).push(["/ai-customer-service-automation",function(){return s(9480)}])},1210:function(e,t){"use strict";Object.defineProperty(t,"__esModule",{value:!0}),t.getDomainLocale=function(e,t,s,a){return!1};("function"===typeof t.default||"object"===typeof t.default&&null!==t.default)&&"undefined"===typeof t.default.__esModule&&(Object.defineProperty(t.default,"__esModule",{value:!0}),Object.assign(t.default,t),e.exports=t.default)},8418:function(e,t,s){"use strict";Object.defineProperty(t,"__esModule",{value:!0});var a=s(4941).Z;s(5753).default;Object.defineProperty(t,"__esModule",{value:!0}),t.default=void 0;var i=s(2648).Z,r=s(7273).Z,n=i(s(7294)),o=s(6273),l=s(2725),c=s(3462),d=s(1018),u=s(7190),h=s(1210),p=s(8684),m={};function _(e,t,s,a){if(e&&o.isLocalURL(t)){Promise.resolve(e.prefetch(t,s,a)).catch((function(e){0}));var i=a&&"undefined"!==typeof a.locale?a.locale:e&&e.locale;m[t+"%"+s+(i?"%"+i:"")]=!0}}var f=n.default.forwardRef((function(e,t){var s,i=e.href,f=e.as,g=e.children,x=e.prefetch,y=e.passHref,j=e.replace,v=e.shallow,w=e.scroll,b=e.locale,A=e.onClick,k=e.onMouseEnter,I=e.onTouchStart,C=e.legacyBehavior,N=void 0===C?!0!==Boolean(!1):C,P=r(e,["href","as","children","prefetch","passHref","replace","shallow","scroll","locale","onClick","onMouseEnter","onTouchStart","legacyBehavior"]);s=g,!N||"string"!==typeof s&&"number"!==typeof s||(s=n.default.createElement("a",null,s));var T=!1!==x,S=n.default.useContext(c.RouterContext),R=n.default.useContext(d.AppRouterContext);R&&(S=R);var M,B=n.default.useMemo((function(){var e=a(o.resolveHref(S,i,!0),2),t=e[0],s=e[1];return{href:t,as:f?o.resolveHref(S,f):s||t}}),[S,i,f]),q=B.href,L=B.as,F=n.default.useRef(q),O=n.default.useRef(L);
1N&&(M=n.default.Children.only(s));var W=N?M&&"object"===typeof M&&M.ref:t,E=a(u.useIntersection({rootMargin:"200px"}),3),G=E[0],Z=E[1],H=E[2],D=n.default.useCallback((function(e){O.current===L&&F.current===q||(H(),O.current=L,F.current=q),G(e),W&&("function"===typeof W?W(e):"object"===typeof W&&(W.current=e))}),[L,W,q,H,G]);n.default.useEffect((function(){var e=Z&&T&&o.isLocalURL(q),t="undefined"!==typeof b?b:S&&S.locale,s=m[q+"%"+L+(t?"%"+t:"")];e&&!s&&_(S,q,L,{locale:t})}),[L,q,Z,b,T,S]);var $={ref:D,onClick:function(e){N||"function"!==typeof A||A(e),N&&M.props&&"function"===typeof M.props.onClick&&M.props.onClick(e),e.defaultPrevented||function(e,t,s,a,i,r,l,c,d,u){if("A"!==e.currentTarget.nodeName.toUpperCase()||!function(e){var t=e.currentTarget.target;return t&&"_self"!==t||e.metaKey||e.ctrlKey||e.shiftKey||e.altKey||e.nativeEvent&&2===e.nativeEvent.which}(e)&&o.isLocalURL(s)){e.preventDefault();var h=function(){"beforePopState"in t?t[i?"replace":"push"](s,a,{shallow:r,locale:c,scroll:l}):t[i?"replace":"push"](s,{forceOptimisticNavigation:!u})};d?n.default.startTransition(h):h()}}(e,S,q,L,j,v,w,b,Boolean(R),T)},onMouseEnter:function(e){N||"function"!==typeof k||k(e),N&&M.props&&"function"===typeof M.props.onMouseEnter&&M.props.onMouseEnter(e),!T&&R||o.isLocalURL(q)&&_(S,q,L,{priority:!0})},onTouchStart:function(e){N||"function"!==typeof I||I(e),N&&M.props&&"function"===typeof M.props.onTouchStart&&M.props.onTouchStart(e),!T&&R||o.isLocalURL(q)&&_(S,q,L,{priority:!0})}};if(!N||y||"a"===M.type&&!("href"in M.props)){var Q="undefined"!==typeof b?b:S&&S.locale,U=S&&S.isLocaleDomain&&h.getDomainLocale(L,Q,S.locales,S.domainLocales);$.href=U||p.addBasePath(l.addLocale(L,Q,S&&S.defaultLocale))}return N?n.default.cloneElement(M,$):n.default.createElement("a",Object.assign({},P,$),s)}));t.default=f,("function"===typeof t.default||"object"===typeof t.default&&null!==t.default)&&"undefined"===typeof t.default.__esModule&&(Object.defineProperty(t.default,"__esModule",{value:!0}),Object.assign(t.default,t),e.exports=t.default)},1018:function(e,t,s){"use strict";Object.defineProperty(t,"__esModule",{value:!0}),t.TemplateContext=t.GlobalLayoutRouterContext=t.LayoutRouterContext=t.AppRouterContext=void 0;var a=(0,s(2648).Z)(s(7294)),i=a.default.createContext(null);t.AppRouterContext=i;var r=a.default.createContext(null);t.LayoutRouterContext=r;var n=a.default.createContext(null);t.GlobalLayoutRouterContext=n;var o=a.default.createContext(null);t.TemplateContext=o},6672:function(e,t,s){"use strict";var a=s(5893),i=(s(7294),s(2229)),r=s.n(i);t.Z=function(e){var t=e.title,s=e.text,i=e.buttonLabel,n=void 0===i?"Contact Us":i,o=e.buttonHref,l=void 0===o?"/ai-workflow-audit":o;return(0,a.jsxs)("div",{className:r().cta,children:[(0,a.jsx)("h3",{className:r().cta__title,children:t}),s&&(0,a.jsx)("p",{className:r().cta__text,children:s}),(0,a.jsx)("a",{href:l,className:r().cta__button,"data-analytics-location":"cta_block",children:n})]})}},337:function(e,t,s){"use strict";var a=s(5893),i=s(7294),r=s(1664),n=s.n(r),o=s(8937),l=s.n(o);t.Z=function(e){var t=e.items,s=e.title,r=void 0===s?"Related Resources":s,o=(0,i.useState)(!1),c=o[0],d=o[1];if(!t||0===t.length)return null;var u=t.filter((function(e){return Boolean((null===e||void 0===e?void 0:e.href)&&e.title&&e.description)}));if(0===u.length)return null;var h=u.length>3;return(0,a.jsxs)("div",{className:l().related,children:[(0,a.jsx)("h3",{className:l().related__title,children:r}),(0,a.jsx)("div",{className:l().related__grid,children:u.map((function(e,t){return(0,a.jsx)(n(),{href:e.href,children:(0,a.jsxs)("a",{className:[l().related__card,!c&&t>=3?l()["related__card--hidden"]:""].filter(Boolean).join(" "),children:[e.category&&(0,a.jsx)("span",{className:l()["related__card-category"],children:e.category}),(0,a.jsx)("h4",{className:l()["related__card-title"],children:e.title}),(0,a.jsx)("p",{className:l()["related__card-desc"],children:e.description})]})},t)}))}),h&&!c&&(0,a.jsxs)("button",{className:l()["related__more-btn"],onClick:function(){return d(!0)},type:"button",children:["See ",u.length-3," more"]})]})}},1856:function(e,t,s){"use strict";var a=s(5893),i=(s(7294),s(8214)),r=s.n(i);t.Z=function(e){var t=e.items,s=e.title,i=void 0===s?"Contents":s;return t&&0!==t.length?(0,a.jsxs)("nav",{className:r().toc,children:[(0,a.jsx)("div",{className:r().toc__title,children:i}),(0,a.jsx)("ol",{className:r().toc__list,children:t.map((function(e){return(0,a.jsx)("li",{className:r().toc__item,children:(0,a.jsx)("a",{href:"#".concat(e.id),children:e.label})},e.id)}))})]}):null}},9480:function(e,t,s){"use strict";s.r(t);var a=s(5893),i=s(1664),r=s.n(i),n=(s(7294),s(2852)),o=s(6672),l=s(3837),c=s(6392),d=s(337),u=s(2534),h=s(1856),p=s(4526),m=s(711),_=s.n(m),f="https://www.layer3labs.io",g=[{id:"what-is-customer-service-automation",label:"What Is Customer Service Automation?"},{id:"what-you-can-automate",label:"What You Can (and Cannot) Automate"},{id:"chatbots",label:"AI Customer Service Chatbots"}
1,{id:"architecture",label:"Architecture of an AI Support System"},{id:"workflows",label:"Key Workflows to Automate"},{id:"stack",label:"Tools and Platforms"},{id:"implementation",label:"Implementation Playbook"},{id:"metrics",label:"Metrics That Matter"},{id:"deflection-benchmarks",label:"Deflection Benchmarks (by Ticket Type)"},{id:"ai-vs-outsourcing",label:"AI Automation vs Outsourcing"},{id:"autonomous-ai",label:"Autonomous / Agentic AI Customer Service"},{id:"generative-ai-for-cx",label:"Generative AI for Customer Experience"},{id:"risks",label:"Risks and How to Manage Them"},{id:"diy-vs-hire",label:"DIY vs. Implementation Partner"},{id:"faq",label:"FAQ"}],x=[{question:"Will AI replace my customer support team?",answer:"No. AI handles the repetitive, high-volume tasks (ticket classification, FAQ responses, data lookup) so your team can focus on complex issues, escalations, and relationship-building. Most businesses redeploy support staff to higher-value work rather than reducing headcount."},{question:"How accurate is AI at answering customer questions?",answer:"For well-scoped FAQ-type questions with a good knowledge base, AI accuracy is typically 85\u201395%. For ambiguous or multi-step questions, accuracy drops to 60\u201375%, which is why human escalation paths are essential. The key is knowing where the boundary is for your business."},{question:"What if the AI gives a customer wrong information?",answer:"Every AI support system should include confidence scoring and escalation rules. Low-confidence responses get routed to a human instead of sent to the customer. During the first 2\u20134 weeks, most teams review 100% of AI outputs before they reach customers."},{question:"How long does it take to set up AI customer support?",answer:"A basic FAQ chatbot takes 1\u20133 weeks. A full triage-and-response system with CRM integration takes 4\u20138 weeks. The timeline depends mainly on how organized your existing knowledge base and ticket data are."},{question:"What is the ROI of AI customer support automation?",answer:"Typical results for SMBs: 30\u201350% reduction in first-response time, 20\u201340% of tickets fully resolved without human intervention, and 15\u201325 hours/week saved for the support team. Payback period is usually 2\u20134 months."},{question:"Is a customer service chatbot the same as AI customer service automation?",answer:"No. A customer service chatbot is one interface. AI customer service automation includes triage, routing, draft replies, CRM updates, escalation rules, knowledge base search, and reporting. The chatbot may be part of the system, but the workflow around it matters more."},{question:"What types of businesses benefit most from AI customer service automation?",answer:"Businesses with high ticket volume and repetitive questions \u2014 e-commerce, SaaS, home services, healthcare scheduling, and professional services \u2014 see the fastest ROI. The threshold for viability is roughly 20\u201330 tickets per day. Below that, the tooling cost may not justify the savings."},{question:"How do I build a knowledge base that AI can actually use?",answer:'Start with your 20 most common ticket topics. Write clear, concise answers (2\u20134 sentences each) in plain language. Organize by category and tag synonyms \u2014 customers ask for "refund" and "return" and "money back." A clean, well-structured knowledge base directly determines AI answer quality.'},{question:"Should I disclose that customers are talking to AI?",answer:'Yes \u2014 always. Transparency is both a best practice and increasingly a legal requirement in some jurisdictions. Customers who discover they were talking to AI without disclosure trust the company less. A simple "Hi, I am an AI assistant" greeting sets accurate expectations and filters customers who want a human immediately.'},{question:"Can AI build a knowledge base for my small business support team?",answer:"Yes. AI can turn your existing tickets, docs, and call transcripts into an organized internal knowledge base \u2014 drafting first-pass articles, tagging by category, and flagging gaps where no answer exists yet. A human should still review and approve each article, since an internal knowledge base built for agents feeds directly into the customer-facing AI answers above, so mistakes compound. Most small teams get a usable first draft in 1\u20132 weeks."},{question:"Will AI replace call center agents?",answer:"No, not entirely. AI takes over the high-volume, repetitive share of call center work \u2014 FAQ answers, order status, basic troubleshooting \u2014 while agents handle escalations, complex cases, and anything emotional. Most SMBs land on a tiered model: AI resolves tier-1 calls outright, co-pilots tier-2 with a human reviewing the draft reply, and agents run tier-3 and overflow. See our AI vs traditional customer-service outsourcing guide for how this compares to a full BPO call center on cost."},{question:"What is customer support automation?",answer:"Customer support automation is another name for customer service automation. Both terms describe software that classifies tickets, answers common questions, and routes harder issues to a person. The two are used interchangeably across the industry."},{question:"Can I get an AI-answered toll-free number for customer service?",answer:"Yes, you can route a toll-free number through an AI voice agent. The agent answers calls, checks order or account status, and resolves the call or transfers it to a person. Setup usually takes one to three weeks, using the same escalation rules as your chat and email support."}],y=[{title:"AI Customer Service Chatbot for Small Business",description:"Build a customer service chatbot with knowledge base content, handoffs, and metrics.",href:"/guides/ai-customer-service-chatbot-for-small-business",category:"Guide"},{title:"AI Customer Service Escalation Workflow",description:"Use a support triage and escalation template before automating customer replies.",href:"/templates/ai-customer-service-escalation-workflow",category:"Template"},{title:"ChatGPT Customer Service",description:"Use ChatGPT for support drafts, summaries, knowledge articles, and escalation rules.",href:"/guides/chatgpt-customer-service",category:"Guide"},{title:"AI Voice Agents for Small Business",description:"Use voice agents for call intake, booking, routing, and after-hours coverage.",href:"/guides/ai-voice-agents-for-small-business",category:"Voice"},{title:"AI Contact Center",description:"Scaling past chat and email? What a full AI contact center adds, real platform costs, and whether you need one.",href:"/guides/ai-contact-center-for-small-business",category:"Guide"},{title:"AI Virtual Receptionist for Small Business",description:"24/7 AI virtual receptionist \u2014 costs, providers, and setup for SMBs.",href:"/guides/ai-virtual-receptionist-for-small-business",category:"Guide"},{title:"AI Workflow Automation",description:"Step-by-step guide to mapping, automating, and monitoring business workflows with AI.",href:"/ai-workflow-automation",category:"Guide"},{title:"AI CRM Integration",description:"How to connect AI to your CRM for lead scoring, data enrichment, and pipeline automation.",href:"/ai-crm-integration",category:"Integration"},{title:"AI for Small Business",description:"Complete guide to where AI fits in a small business \u2014 workflows, costs, tools, and risks.",href:"/ai-for-small-business",category:"Guide"},{title:"Support Automation Savings Calculator",description:"Estimate how much your team could save by automating customer support with AI.",href:"/tools/support-savings-calculator",category:"Calculator"},{title:"AI Receptionist for Small Business",description:"AI receptionists that answer calls 24/7, book appointments, and capture leads.",href:"/guides/ai-receptionist-for-small-business",category:"Guide"},{title:"Custom AI Agent vs AI Chatbot",description:"Compare custom AI agents and off-the-shelf chatbots for customer service.",href:"/comparisons/custom-ai-agent-vs-ai-chatbot",category:"Comparison"},{title:"AI for Retail Businesses",description:"Automate inventory forecasting, POS sync, and 24/7 order-status chat for retail.",href:"/ai-for/retail",category:"Industry"},{title:"Missed Call Revenue Calculator",description:"Calculate how much revenue your business loses to unanswered calls.",href:"/tools/missed-call-calculator",category:"Calculator"}];t.default=function(){var e="".concat(f,"/ai-customer-service-automation"),t="AI Customer Service Automation & Conversational AI for SMBs | Layer3Labs",s="Conversational AI for customer service: automate 40\u201360% of support volume without losing trust. Which tickets to hand to AI vs. keep human, and the cost.";return(0,a.jsxs)(p.Z,{title:t,description:s,children:[(0,a.jsx)(c.ZP,{data:(0,c.Pe)({title:t,description:s,url:e,datePublished:"2026-05-14",dateModified:"2026-09-14"})}),(0,a.jsx)(c.ZP,{data:(0,c.pI)({name:"AI Customer Service Automation",description:s,url:e,serviceType:"AI Automation"})}),(0,a.jsx)(c.ZP,{data:(0,c.dl)([{name:"Home",url:f},{name:"Resources",url:"".concat(f,"/resources")},{name:"AI Customer Service Automation",url:e}])}),(0,a.jsx)(c.ZP,{data:(0,c.Cm)(x)}),(0,a.jsx)(c.ZP,{data:(0,c.OX)()}),(0,a.jsxs)("div",{className:_().pillar,children:[(0,a.jsx)(n.Z,{}),(0,a.jsx)("div",{className:_().pillar__hero,children:(0,a.jsx)(u.Z,{title:"AI Customer Service Automation: What to Automate, What to Keep Human",subtitle:"A practical guide to automating customer support \u2014 triage, drafting, escalation, and self-service, without losing the human touch.",breadcrumbs:[{label:"Home",href:"/"},{label:"Resources",href:"/resources"},{label:"AI Customer Service Automation"}]})}),(0,a.jsxs)("div",{className:_().pillar__body,children:[(0,a.jsx)("div",{className:_().pillar__toc,children:(0,a.jsx)(h.Z,{items:g})}),(0,a.jsxs)("section",{id:"what-is-customer-service-automation",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"What Is Customer Service Automation?"}),(0,a.jsx)("p",{className:_().pillar__p,children:"Customer service automation uses software, often AI-powered, to handle support tasks without a person doing each one by hand. It reads
1incoming messages, classifies the issue, pulls the right information, and answers or routes the ticket. Customer support automation describes the same shift under a different name."}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Examples of AI in customer service"}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Ticket triage."})," AI reads a new ticket, tags its category and urgency, and sends it to the right queue."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"FAQ chatbot."})," AI answers common questions, such as store hours or pricing, from your knowledge base."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Order status lookup."})," AI checks your order system and tells the customer where their shipment stands."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Draft-and-review replies."})," AI writes a response for harder tickets, and a human agent edits it before sending."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Proactive follow-up."})," After a ticket closes, AI sends a satisfaction check and flags any sign of lingering frustration."]})]}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Benefits of AI in customer service"}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Faster response times."})," AI answers common questions in seconds instead of hours."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Round-the-clock coverage."})," AI keeps answering after your team goes home, on weekends, and on holidays."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Lower cost per ticket."})," Automated tickets cost cents to resolve instead of dollars in agent time."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Consistent answers."})," AI pulls from the same knowledge base every time, so customers get the same answer no matter who responds."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"More time for hard problems."})," Freeing agents from repetitive tickets lets them focus on complaints, escalations, and relationships."]})]})]}),(0,a.jsxs)("section",{id:"what-you-can-automate",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"What You Can (and Cannot) Automate"}),(0,a.jsx)("p",{className:_().pillar__p,children:"AI excels at the first three steps of customer support: reading the incoming message, classifying the issue, and pulling relevant information. It struggles with empathy-heavy interactions, novel situations, and multi-step troubleshooting that requires back-and-forth."}),(0,a.jsx)("div",{className:_()["pillar__table-wrap"],children:(0,a.jsxs)("table",{className:_().pillar__table,children:[(0,a.jsx)("thead",{children:(0,a.jsxs)("tr",{children:[(0,a.jsx)("th",{children:"Automate Well"}),(0,a.jsx)("th",{children:"Keep Human"})]})}),(0,a.jsxs)("tbody",{children:[(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Ticket classification and routing"}),(0,a.jsx)("td",{children:"Angry customer de-escalation"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"FAQ responses (password resets, hours, pricing)"}),(0,a.jsx)("td",{children:"Complex billing disputes"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Order status lookups"}),(0,a.jsx)("td",{children:"Product returns requiring judgment"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Response drafting for common issues"}),(0,a.jsx)("td",{children:"VIP or high-value account issues"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Knowledge base search and retrieval"}),(0,a.jsx)("td",{children:"Legal or compliance-sensitive requests"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Follow-up scheduling"}),(0,a.jsx)("td",{children:"Multi-party coordination"})]})]})]})})]}),(0,a.jsxs)("section",{id:"chatbots",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"AI Customer Service Chatbots vs. Full Support Automation"}),(0,a.jsx)("p",{className:_().pillar__p,children:"An AI customer service chatbot answers questions in a chat window. A full AI customer service automation system handles the work around the chat: classifying the issue, checking customer context, drafting a response, updating the ticket, and escalating when confidence is low."}),(0,a.jsx)("p",{className:_().pillar__p,children:"Conversational AI for customer service takes this further \u2014 using large language models to hold natural, multi-turn dialogues across chat, email, and voice channels. Rather than selecting from pre-defined options, customers describe their issue in their own words and the AI responds contextually. This is what distinguishes conversational AI customer service from older rule-based chatbots."}),(0,a.jsx)("p",{className:_().pillar__p,children:"Small businesses often start with a chatbot because it is visible. But the biggest time savings usually come from the back-end workflow: routing, tagging, summary writing, CRM updates, and follow-up reminders."}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Use a chatbot"})," for high-volume, low-risk questions with clear answers."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Use draft-and-review automation"})," for questions that need context or judgment."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Use escalation rules"})," for billing disputes, angry customers, legal issues, refunds, and VIP accounts."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Use reporting"})," to find gaps in your knowledge base and improve automation coverage over time."]})]})]}),(0,a.jsxs)("section",{id:"architecture",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Architecture of an AI Support System"}),(0,a.jsx)("p",{className:_().pillar__p,children:"A well-designed AI support system has four layers that work together:"}),(0,a.jsxs)("ol",{className:_().pillar__ol,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Intake layer"})," \u2014 Receives messages from all channels (email, chat, social, phone transcripts) and normalizes them into a standard format."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Intelligence layer"})," \u2014 Classifies the issue, determines urgency, checks the knowledge base, and generates a draft response. This is where the AI model lives."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Routing layer"})," \u2014 Based on confidence scores and business rules, either sends the AI response directly, queues it for human review, or routes it to a specialist."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Action layer"})," \u2014 Executes the resolution: sends the response, updates the ticket, logs to CRM, triggers follow-up sequences."]})]}),(0,a.jsxs)("div",{className:_().pillar__callout,children:[(0,a.jsx)("strong",{children:"Critical design principle:"})," The routing layer is where quality control lives. Set confidence thresholds conservatively at launch (review everything), then relax them as the system proves itself."]})]}),(0,a.jsxs)("section",{id:"workflows",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Key Workflows to Automate"}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Tier 1: Ticket triage and classification"}),(0,a.jsx)("p",{className:_().pillar__p,children:"Automatically tag incoming tickets by category, urgency, and required skill set. Routes to the right queue without human sorting. Impact: eliminates 20\u201330 minutes of daily triage work per support agent."}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Tier 2: FAQ auto-response"}),(0,a.jsx)("p",{className:_().pillar__p,children:"For questions that map to existing documentation (hours, pricing, how-to guides), generate and send responses automatically. Requires a well-maintained knowledge base. Typically handles 20\u201340% of total ticket volume."}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Tier 3: Response drafting with human review"}),(0,a.jsx)("p",{className:_().pillar__p,children:"For moderately complex issues, AI drafts a response and presents it to a human agent for review and editing. Reduces response crafting time from 5\u201310 minutes to 1\u20132 minutes per ticket."}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Tier 4: Proactive follow-up"}),(0,a.jsx)("p",{className:_().pillar__p,children:"After ticket resolution, automatically send satisfaction checks, resurface unresolved
1issues, and trigger escalation if a customer responds with continued frustration."})]}),(0,a.jsxs)("section",{id:"stack",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Tools and Platforms"}),(0,a.jsx)("div",{className:_()["pillar__table-wrap"],children:(0,a.jsxs)("table",{className:_().pillar__table,children:[(0,a.jsx)("thead",{children:(0,a.jsxs)("tr",{children:[(0,a.jsx)("th",{children:"Category"}),(0,a.jsx)("th",{children:"Options"}),(0,a.jsx)("th",{children:"Best For"})]})}),(0,a.jsxs)("tbody",{children:[(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Help Desk + AI"}),(0,a.jsx)("td",{children:"Intercom Fin, Zendesk AI, Freshdesk Freddy"}),(0,a.jsx)("td",{children:"Teams already on these platforms"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Custom AI Layer"}),(0,a.jsx)("td",{children:"OpenAI API + n8n/Make + your help desk"}),(0,a.jsx)("td",{children:"Custom workflows, multi-system integration"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Knowledge Base"}),(0,a.jsx)("td",{children:"Notion + AI search, Guru, custom RAG pipeline"}),(0,a.jsx)("td",{children:"Internal and external documentation"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Voice/Phone"}),(0,a.jsx)("td",{children:"Bland.ai, Synthflow, Retell"}),(0,a.jsx)("td",{children:"Phone-first support operations"})]})]})]})}),(0,a.jsx)("p",{className:_().pillar__p,children:"For most SMBs handling 50\u2013500 tickets/day, a custom AI layer connected to your existing help desk provides the best balance of control and cost. Pre-built AI features in help desk platforms work well for simpler setups."}),(0,a.jsxs)("div",{className:_().pillar__callout,children:[(0,a.jsx)("strong",{children:"Considering an enterprise contact-center platform?"})," Vendors like NICE CXone sell real AI capability, but the pricing and implementation cost are built for 100+-agent operations. See our ",(0,a.jsx)(r(),{href:"/comparisons/nice-cxone-alternatives",children:(0,a.jsx)("a",{children:"NICE CXone alternatives comparison"})})," for SMB-sized options if you have roughly 5\u201350 agents."]})]}),(0,a.jsxs)("section",{id:"implementation",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Implementation Playbook"}),(0,a.jsxs)("ol",{className:_().pillar__ol,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Week 1 \u2014 Audit your tickets:"})," Export 30 days of tickets. Categorize them by type, complexity, and resolution path. Identify the top 5 categories by volume."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Week 2 \u2014 Build your knowledge base:"})," Compile and clean the documentation needed to answer the top 5 ticket categories. This is the foundation the AI will draw from."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Week 3\u20134 \u2014 Build the triage pipeline:"})," Set up classification, knowledge base search, and response drafting. Test against historical tickets to measure accuracy."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Week 5 \u2014 Supervised launch:"})," Go live with 100% human review. Every AI-drafted response is reviewed by a support agent before sending. Track accuracy and customer satisfaction."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Week 6\u20138 \u2014 Scale automation:"})," For ticket types where AI accuracy exceeds 90%, begin sending responses with reduced human review. Keep full review for edge cases and low-confidence outputs."]})]})]}),(0,a.jsxs)("section",{id:"metrics",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Metrics That Matter"}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Deflection rate"})," \u2014 Percentage of tickets resolved without human intervention. Target: 20\u201340% within 60 days."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"First-response time"})," \u2014 Time from ticket creation to first meaningful response. AI typically reduces this by 50\u201380%."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Agent handling time"})," \u2014 Time each agent spends per ticket. AI drafting reduces this by 40\u201360%."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"CSAT on AI-handled tickets"})," \u2014 Compare satisfaction scores for AI-resolved vs. human-resolved tickets. If AI tickets score 10%+ lower, tighten the automation boundaries."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Escalation rate"})," \u2014 Percentage of AI-handled tickets that escalate to a human. Should decrease over time as the system improves."]})]})]}),(0,a.jsxs)("section",{id:"deflection-benchmarks",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Deflection Benchmarks: What 60% Actually Looks Like"}),(0,a.jsx)("p",{className:_().pillar__p,children:'"Resolve 60% of tickets without an agent" is the headline most AI customer service vendors lead with. The number is real for some businesses, misleading for others. The benchmark below shows what deflection actually looks like by ticket type \u2014 so you can predict yours before signing.'}),(0,a.jsx)("div",{className:_()["pillar__table-wrap"],children:(0,a.jsxs)("table",{className:_().pillar__table,children:[(0,a.jsx)("thead",{children:(0,a.jsxs)("tr",{children:[(0,a.jsx)("th",{children:"Ticket Type"}),(0,a.jsx)("th",{children:"Realistic AI Deflection Rate"}),(0,a.jsx)("th",{children:"Why"})]})}),(0,a.jsxs)("tbody",{children:[(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Account / password resets"}),(0,a.jsx)("td",{children:"80\u201395%"}),(0,a.jsx)("td",{children:"Structured workflow, clear escalation triggers"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Order status / shipping"}),(0,a.jsx)("td",{children:"70\u201390%"}),(0,a.jsx)("td",{children:"Data is in a system AI can read"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"FAQ / policy questions"}),(0,a.jsx)("td",{children:"60\u201380%"}),(0,a.jsx)("td",{children:"Good knowledge base = high deflection"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Product how-to / setup"}),(0,a.jsx)("td",{children:"40\u201365%"}),(0,a.jsx)("td",{children:"Limited by docs quality + edge cases"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Billing disputes"}),(0,a.jsx)("td",{children:"20\u201335%"}),(0,a.jsx)("td",{children:"Most need human judgment; AI gathers info"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Complaints / refunds"}),(0,a.jsx)("td",{children:"5\u201315%"}),(0,a.jsx)("td",{children:"Almost always need a person"})]}),(0,a.jsxs)("tr",{children:[(0,a.jsx)("td",{children:"Technical bugs / outages"}),(0,a.jsx)("td",{children:"10\u201325%"}),(0,a.jsx)("td",{children:"AI tags + routes; humans resolve"})]})]})]})}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"Cost Per Ticket \u2014 Before vs. After"}),(0,a.jsx)("p",{className:_().pillar__p,children:"Useful way to model the savings: cost per ticket, split into agent labor and AI cost. Typical 2026 numbers for an SMB doing 4,000 tickets/month:"}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Before AI:"})," ~$5.20 per ticket fully loaded (agent time + tooling + overhead)."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"After AI at 35% blended deflection:"})," ~$3.55 per ticket. Monthly savings: ~$6,600."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"After AI at 55% blended deflection:"})," ~$2.70 per ticket. Monthly savings: ~$10,000."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Common build cost:"})," $18,000\u2013$45,000 for a single-channel deployment with knowledge-base grounding, escalation logic, and CSAT tracking."]})]}),(0,a.jsxs)("div",{className:_().pillar__callout,children:[(0,a.jsx)("strong",{children:"The trap to avoid:"}),' chasing a 60%+ headline rate by tightening the AI\'s allowed answers too narrowly. You end up with a bot that defers to a human on anything interesting \u2014 which hurts CSAT and trains customers to
1start with "talk to a human." Aim for the right rate per ticket type, not a single headline number.']})]}),(0,a.jsxs)("section",{id:"ai-vs-outsourcing",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"AI Automation vs Outsourcing Your Customer Service"}),(0,a.jsx)("p",{className:_().pillar__p,children:'For two decades, the default answer to "support volume is breaking us" was customer support outsourcing \u2014 a BPO in the Philippines, India, or Latin America at $8\u2013$18 per agent-hour. In 2026, AI customer service automation has become the cheaper option on most metrics, but the choice is not binary. The smartest support teams blend both, and the boundary between them is where most of the savings live.'}),(0,a.jsx)("p",{className:_().pillar__p,children:"Outsourcing customer service buys you human warmth, judgment on edge cases, and instant capacity ramp. The cost is per-agent-hour, language coverage is uneven, and quality drifts when attrition is high. AI automation buys you sub-second response, 24/7 availability, predictable per-ticket cost, and infinite concurrency \u2014 at the cost of empathy on emotionally charged tickets and the ongoing maintenance burden of knowledge-base hygiene."}),(0,a.jsx)("h3",{className:_().pillar__h3,children:"The blended model that wins in 2026"}),(0,a.jsx)("p",{className:_().pillar__p,children:"The pattern most SMBs are landing on: AI handles tier-1 (FAQs, order status, basic troubleshooting, account questions) and tier-2 with co-pilot assist (AI drafts the reply, the human reviews and sends). Outsourced human agents handle tier-3 (escalations, complaints, retention, anything emotional) and overflow during seasonal spikes."}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Cost per ticket \u2014 AI:"})," $0.10\u2013$0.50 fully loaded (LLM + platform + KB hygiene)"]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Cost per ticket \u2014 outsourcing:"})," $1.50\u2013$7.00 depending on geography and complexity"]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Best for AI:"})," high-volume, FAQ-heavy tickets, predictable patterns, 24/7 coverage"]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Best for outsourcing:"})," emotional tickets, complex policy decisions, surge capacity, languages your AI cannot cover"]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Best for blended:"})," any team with both transactional and judgment-heavy tickets \u2014 which is most SMBs"]})]}),(0,a.jsx)("p",{className:_().pillar__p,children:"The honest tradeoff outsourcing vendors will not raise: AI scales without quality drift, but a knowledge base that goes 6 months without grooming starts producing wrong answers \u2014 and unlike a human, the AI never tells you it is uncertain. The maintenance burden is real and needs an owner."})]}),(0,a.jsxs)("section",{id:"autonomous-ai",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Autonomous / Agentic AI Customer Service"}),(0,a.jsx)("p",{className:_().pillar__p,children:"Autonomous AI customer service means the AI resolves a ticket end-to-end \u2014 looking up an order, issuing a refund, updating an account, or rebooking an appointment \u2014 without a human approving each step. This goes further than the tiered automation above, where a human still reviews AI-drafted replies at tier 2."}),(0,a.jsx)("p",{className:_().pillar__p,children:"The distinction that matters: tiered automation drafts, a human sends. Agentic automation acts, then reports what it did. That shift only makes sense once the AI has reliable tool access (order systems, billing, scheduling) and clear guardrails on what it is allowed to do without asking first."}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Good fits for autonomous resolution:"})," order status lookups, standard refunds under a fixed dollar threshold, appointment rescheduling, password/account resets, subscription pauses \u2014 actions that are reversible and bounded."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Keep a human in the loop for:"})," refunds above a set threshold, account cancellations, anything involving a complaint about being harmed or misled, and any action that is hard to reverse."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Guardrails that make autonomy safe:"})," a hard allow-list of actions the AI can take unsupervised, dollar/frequency caps per action, full audit logging of every autonomous action taken, and a standing rule that any action outside the allow-list routes to tier-2 review instead of failing silently."]})]}),(0,a.jsxs)("div",{className:_().pillar__callout,children:[(0,a.jsx)("strong",{children:"Where this fits your rollout:"})," most SMBs should not start here. Get tier-1 FAQ deflection and tier-2 co-pilot assist solid first (see the implementation playbook above), then graduate the highest-volume, lowest-risk actions to autonomous resolution once you trust the AI's accuracy on that specific workflow."]})]}),(0,a.jsxs)("section",{id:"generative-ai-for-cx",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Generative AI for Customer Experience"}),(0,a.jsx)("p",{className:_().pillar__p,children:"Generative artificial intelligence (AI) improves customer experience (CX) by drafting complete replies, summarizing long message threads, and personalizing responses using live account records. At Layer3 Labs, we see small and midsize business (SMB) teams use generative drafting to speed up routine replies while keeping human agents in control of approvals. This focus on reply quality and context separates customer experience improvements from basic ticket routing."}),(0,a.jsx)("p",{className:_().pillar__p,children:"Generative systems improve response quality and accuracy across four practical areas:"}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Drafting complete responses:"})," the model uses a short list of facts to create a full, on-brand reply, which frees agents from typing standard messages from scratch."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Summarizing interaction history:"})," the system condenses long email threads or chat transcripts into a brief summary, which saves staff from re-reading past messages."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Personalizing customer context:"})," replies reference the buyer's actual subscription tier, order history, or product details instead of relying on generic email templates."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Grounding data with retrieval-augmented generation (RAG):"})," connecting the drafting engine to verified account records stops the system from inventing policies or confusing account terms."]})]}),(0,a.jsx)("p",{className:_().pillar__p,children:"The most common failure mode occurs when a business deploys generative drafting without connecting the tool to verified account data. The model then writes fluent, confident replies that reference the wrong service plan or quote an invali
1d return policy. This is a data-grounding problem. Better prompts cannot solve missing customer context."}),(0,a.jsx)("p",{className:_().pillar__p,children:"Teams solve this error by using retrieval-augmented generation against the customer database. That setup injects current order records and account status directly into the drafting step. Grounding the output in verified account facts protects customer trust while cutting the time agents spend typing routine messages."})]}),(0,a.jsxs)("section",{id:"risks",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"Risks and How to Manage Them"}),(0,a.jsxs)("ul",{className:_().pillar__ul,children:[(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Hallucinated answers"})," \u2014 AI invents product features or policies that do not exist. Mitigation: constrain the AI to your knowledge base (RAG architecture) and set strict citation requirements."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Tone mismatches"})," \u2014 AI responds too formally to a casual customer or too casually to a frustrated one. Mitigation: include sentiment detection in the pipeline and adjust tone instructions dynamically."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Feedback loops"})," \u2014 AI learns from its own mistakes if incorrect responses get marked as resolved. Mitigation: separate AI accuracy tracking from ticket resolution metrics."]}),(0,a.jsxs)("li",{children:[(0,a.jsx)("strong",{children:"Customer frustration with bots"})," \u2014 Some customers want a human immediately. Mitigation: always provide a clear, one-click path to a human agent. Never trap customers in an AI loop."]})]})]}),(0,a.jsxs)("section",{id:"diy-vs-hire",className:_().pillar__section,children:[(0,a.jsx)("h2",{className:_().pillar__h2,children:"DIY vs. Implementation Partner"}),(0,a.jsxs)("p",{className:_().pillar__p,children:[(0,a.jsx)("strong",{children:"DIY works if"})," you are already on a platform with built-in AI (Intercom, Zendesk) and want basic FAQ automation with minimal customization. Budget: $0\u2013$500/month in additional platform costs."]}),(0,a.jsxs)("p",{className:_().pillar__p,children:[(0,a.jsx)("strong",{children:"An implementation partner is worth it if"})," you need custom triage logic, multi-channel integration, CRM sync, or voice/phone support automation. Budget: $15,000\u2013$50,000 for the initial build, $1,000\u2013$3,000/month for ongoing optimization."]})]})]}),(0,a.jsx)("hr",{className:_().pillar__divider}),(0,a.jsx)("div",{className:_()["pillar__faq-wrap"],children:(0,a.jsx)(l.Z,{items:x})}),(0,a.jsx)("div",{className:_()["pillar__cta-wrap"],children:(0,a.jsx)(o.Z,{title:"Ready to Automate Your Customer Support?",text:"We build AI-powered support systems that handle triage, drafting, and escalation \u2014 without replacing your team. Start with a free support workflow audit.",buttonLabel:"Get a Support Automation Audit",buttonHref:"/ai-workflow-audit"})}),(0,a.jsx)("div",{className:_()["pillar__related-wrap"],children:(0,a.jsx)(d.Z,{items:y})})]})]})}},711:function(e){e.exports={pillar:"Pillar_pillar__Cf6Gt",pillar__hero:"Pillar_pillar__hero__c4vPo",pillar__body:"Pillar_pillar__body__XeMxB",pillar__toc:"Pillar_pillar__toc__R0j4S",pillar__section:"Pillar_pillar__section__monQV",pillar__h2:"Pillar_pillar__h2__Lar__",pillar__h3:"Pillar_pillar__h3__voFAk",pillar__p:"Pillar_pillar__p__QOEwr",pillar__ul:"Pillar_pillar__ul__UQRNa",pillar__ol:"Pillar_pillar__ol__0nsX4","pillar__table-wrap":"Pillar_pillar__table-wrap__ig0fi",pillar__table:"Pillar_pillar__table___lae_",pillar__callout:"Pillar_pillar__callout__Jj8Ex",pillar__sources:"Pillar_pillar__sources__roz4C","pillar__sources-label":"Pillar_pillar__sources-label__u56te","pillar__sources-list":"Pillar_pillar__sources-list__RIvYT",pillar__source:"Pillar_pillar__source__mPXkl","pillar__source-icon":"Pillar_pillar__source-icon__W8mN1",pillar__cards:"Pillar_pillar__cards__gIPA8",pillar__card:"Pillar_pillar__card__aiADj","pillar__card-num":"Pillar_pillar__card-num__qlNys","pillar__card-title":"Pillar_pillar__card-title__DrTIf","pillar__card-text":"Pillar_pillar__card-text__KxmHd",pillar__crosslink:"Pillar_pillar__crosslink__9aE1R",pillar__disclaimer:"Pillar_pillar__disclaimer__5WFc0","pillar__inline-cta":"Pillar_pillar__inline-cta__VyN9O","pillar__inline-cta-text":"Pillar_pillar__inline-cta-text__da2fH","pillar__inline-cta-button":"Pillar_pillar__inline-cta-button__M7f68",pillar__form:"Pillar_pillar__form__7B00C","pillar__field-grid":"Pillar_pillar__field-grid__V02T_",pillar__field:"Pillar_pillar__field__8FUMT","pillar__form-button":"Pillar_pillar__form-button__RCRCw",pillar__divider:"Pillar_pillar__divider__kiDmZ","pillar__cta-wrap":"Pillar_pillar__cta-wrap___9gQg","pillar__related-wrap":"Pillar_pillar__related-wrap__qbcVo","pillar__faq-wrap":"Pillar_pillar__faq-wrap__G3ga8"}},2229:function(e){e.exports={cta:"CtaBlock_cta__GAcFA",cta__title:"CtaBlock_cta__title__r8_dK",cta__text:"CtaBlock_cta__text__nubF8",cta__button:"CtaBlock_cta__button__ury_A"}},8937:function(e){e.exports={related:"RelatedResources_related__t0w8B",related__title:"RelatedResources_related__title__eq30W",related__grid:"RelatedResources_related__grid__Pt53_",related__card:"RelatedResources_related__card__uJcZ0","related__card-category":"RelatedResources_related__card-category__WY8_n","related__card-title":"RelatedResources_related__card-title__YuLza","related__card-desc":"RelatedResources_related__card-desc__SsbMI","related__card--hidden":"RelatedResources_related__card--hidden__uyF0s","related__more-btn":"RelatedResources_related__more-btn__JnqeM"}},8214:function(e){e.exports={toc:"TableOfContents_toc__Jcvth",toc__title:"TableOfContents_toc__title__h5v_t",toc__list:"TableOfContents_toc__list__CRtXV",toc__item:"TableOfContents_toc__item__1HkyI"}},9008:function(e,t,s){e.exports=s(5443)},1664:function(e,t,s){e.exports=s(8418)},1351:function(e,t,s){"use strict";function a(e){throw e}s.d(t,{Z:function(){return a}})}},function(e){e.O(0,[4526,2852,1973,5299,7827,9774,2888,179],(function(){return t=4166,e(e.s=t);var t}));var t=e.O();_N_E=t}]);
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.