AI Agents From a URL in 60 Seconds

The 60-Second Revolution Is Here

What Jannis Moore shows on screen barely looks real. In under 60 seconds, he goes from a random Google search for “plumbers in Seattle” to a fully functional, hyper-personalized voice AI agent that can debate tankless vs. traditional water heaters like a seasoned call center rep.

Traditional chatbot or IVR projects do not work like this. A typical voice IVR rollout can take weeks: scoping flows, writing prompts, provisioning phone numbers, wiring APIs, testing edge cases, then redoing half of it when the client changes their FAQ. Even “modern” chatbots usually need custom code, a hosting stack, and someone who actually enjoys YAML.

Here, the recipe is almost offensively simple: grab a local plumber’s URL, paste it into the ElevenLabs agents platform, set a goal (“create a customer support agent that answers basic FAQ questions”), and click create. ElevenLabs auto-generates a system prompt, scrapes the site, builds a knowledge base, and spins up a voice interface that sounds human enough to pass a quick phone demo.

For small businesses, that shift is bigger than it looks. Voice AI agents used to be a luxury tier: something you bought from a SaaS vendor, a call center outsourcer, or an agency billing four figures a month. Now a solo operator or tiny agency can stand up a working “AI receptionist” for a random plumber in under a minute, then send a preview link to prove it.

Zero-barrier-to-entry is not marketing spin here. You do not: - Write a single line of code - Stand up servers or containers - Design complex dialog flows - Manage your own speech models

You bring a URL and a clear goal; the platform does the rest. That flips the power dynamic for local businesses that never had an IT department, let alone an ML team. When a non-technical owner can talk to a custom, knowledge-aware voice agent built from their own website in one minute, access to advanced AI stops being a multi-year roadmap and starts looking like a browser tab.

From Website URL to Working Voice Agent

Start with a business that actually exists. Jannis Moore opens Google, types “plumber Seattle,” and grabs a random local shop from the results. You copy the homepage URL from Chrome’s address bar—no scraping tools, no CSV, just the same link a customer would click.

Then you jump into the ElevenLabs agents dashboard. A prominent “New agent” button kicks off the flow, where you choose Business agent as the template. Under industry, you drill into “Professional services” and then “Customer support,” signaling that this agent should behave like frontline help, not a sales bot or generic assistant.

From here, the setup hinges on two inputs. First, you paste the plumber’s website URL into the knowledge base field, effectively telling ElevenLabs, “this is the source of truth.” Second, you define the main goal, a single system-level instruction like: “create a customer support agent for a plumber that can answer basic FAQ questions.”

That main goal doubles as your system prompt without you writing prompt-engineering boilerplate. ElevenLabs auto-generates a full instruction block that sets tone, scope, and behavior for the agent. The platform then enriches that prompt with live details pulled from the site—business name, services, locations, and typical jobs.

All of that happens behind one button. You click “create agent,” and that’s the only real moment of generation in the entire workflow. No models to select, no token limits to tweak, no separate vector database to wire up.

Seconds later, the interface shows a prebuilt system prompt populated with references to the actual plumber—Moore’s demo calls out “Beast Plumbing” by name. A knowledge base entry already exists, tied directly to the URL you provided. For extra coverage, you can still head to the Knowledge Base tab, add the same URL again, and let the system crawl deeper pages.

By that point, you already have a working voice agent. You hit “Preview agent,” ask, “hey, is a traditional tank or tankless water heater better?” and a natural-sounding voice answers using real content from the plumber’s own FAQ.

The Magic Under the Hood: Instant Expertise

Magic here starts with retrieval-augmented generation, or RAG. Instead of relying on whatever a base model “remembers,” ElevenLabs agents crawl the plumber’s site, slice the text into chunks, embed those chunks as vectors, and retrieve the most relevant pieces every time someone asks a question. The model then generates an answer tightly anchored to those snippets, not to its own imagination.

When Jannis pastes the URL and types a main goal, the platform auto-builds a system prompt behind the scenes. That prompt doesn’t just say “answer questions”; it encodes role, tone, scope, and boundaries: act as a customer support agent for this specific plumbing business, stick to FAQs, avoid pricing guesses, and redirect emergencies. You never see the prompt engineering, but the agent’s personality and guardrails come from that single line of intent.

RAG and the system prompt work as a kind of double safety net against hallucinations. The system prompt narrows what the model is allowed to talk about, while the retrieval layer feeds it only content scraped from the plumber’s pages: services offered, service areas, hours, brand name, and real FAQs. When the agent explains tankless vs. traditional water heaters, it pulls phrasing and facts from the actual website copy rather than inventing generic plumbing advice.

Contrast that with the old way of building support bots. Teams would manually assemble PDFs, Word docs, and spreadsheets, then upload them into some brittle knowledge base that needed constant reformatting. Every change on the website meant another round of edits, re-uploads, and sync delays.

Here, the website becomes the single source of truth. You add the URL once for the agent definition, again for the knowledge base, and the platform fetches and indexes the content automatically. Updating business info is as simple as editing the site and re-syncing, which collapses hours of content wrangling into a few clicks.

For anyone who wants to go deeper into how this stacks up against traditional bot builders, ElevenLabs has more technical detail at Deploy Conversational AI Agents in Minutes - ElevenLabs.

Why The Voice Is The New User Interface

Voice is quietly becoming the real UI for AI agents, and ElevenLabs is betting everything on making that voice sound human. Their stack focuses on three things: realism, low latency, and emotional expressiveness. That combination turns a scripted chatbot into something that feels like a competent human on the other end of the line.

Traditional voice assistants still sound like they are reading from a teleprompter underwater. Long pauses, chopped-up sentences, and zero sense of when you are finished talking. ElevenLabs’ turn-taking system keeps latency low enough that you can interrupt, clarify, or change direction mid-sentence without the agent freezing or talking over you.

Cadence matters as much as raw audio quality. The demo plumber agent doesn’t just recite a paragraph about tankless water heaters; it paces the answer, adds natural emphasis, and lands on key benefits the way a real salesperson would. That rhythm makes the interaction feel like a conversation, not a support ticket.

Trust is where voice becomes a business weapon. A crisp, confident, emotionally tuned voice makes callers stay on the line longer and ask more questions. When the agent responds instantly, with the right tone and accurate info pulled from the site, customers stop testing it and start relying on it.

Friction drops fast when you remove clunky IVR menus and robotic prompts. Instead of “Press 3 for water heaters,” you just say, “hey, my shower keeps going cold, what should I do?” The agent replies in a reassuring, human-like voice that sounds like it has handled this a thousand times before. That familiarity keeps people from hanging up or bouncing to a competitor.

Global scale is baked in. ElevenLabs supports over 5,000+ voices across 31+ languages, so you can spin up localized agents for different markets without rewriting your whole playbook. One plumber in Seattle today, a chain of service businesses across Europe, Latin America, and Asia tomorrow—same pipeline, different voices.

Beyond the Demo: Enterprise-Ready Power

Forget the plumber demo for a second—ElevenLabs’ agents platform is really a front door to a much bigger Conversational AI 2.0 stack aimed squarely at enterprises. Under the same interface you used for a one-minute mockup, you can wire up production-grade agents that sit inside call centers, healthcare systems, and financial workflows.

Security and compliance move from nice-to-have to table stakes here. ElevenLabs advertises HIPAA support for healthcare data, SOC 2 alignment for broader enterprise security, and a zero-retention mode where audio and transcripts never persist on their servers, which matters if you are handling PHI, legal calls, or high-value financial accounts.

On top of that, role-based access control keeps sprawling teams from stepping on each other. You can segment who can: - Create or edit agents - Change knowledge sources and tools - View logs, analytics, and recordings

That kind of permission model is the difference between a clever demo and something a compliance officer will actually sign off on.

Conversational AI 2.0 also means multimodal by default. Users can speak, pause, and then type follow-up questions in the same session, with the agent tracking context across voice and text and responding in whichever mode makes sense. That flexibility matters when someone starts a call on mobile, moves to desktop, or needs to paste in an order ID or policy number.

Turn-taking is where ElevenLabs leans hardest on its real-time stack. The system uses low-latency streaming so the agent can interrupt itself, respond mid-sentence, or gracefully pause when a human barges in, mimicking how support reps actually talk. No more waiting for a full paragraph of TTS to finish before you can correct a mistake.

Telephony integration turns all of this into a drop-in replacement for parts of your call center. Through SIP, WebRTC, or direct carrier hooks, agents can handle: - Inbound support lines and IVR-style routing - Outbound follow-ups, reminders, and surveys - After-hours triage with full call logging and analytics

Once wired up, that one-minute URL trick stops being a parlor trick and becomes a way to rapidly prototype and then harden real, revenue-facing voice workflows.

Your New Secret Weapon for Closing Deals

Sales agencies and automation consultants just got a brutally effective new wedge: spin up a working, branded voice agent for a prospect in the time it takes them to finish a coffee. Instead of pitching “AI transformation” in the abstract, you send a link where the client literally talks to their own AI agent, already fluent in their services, pricing, and FAQs.

Jannis Moore’s plumber demo is the template. You grab a local business site, drop the URL into ElevenLabs agents platform, define a goal like “handle customer support FAQs,” and in under 60 seconds you have a voice agent that sounds like a real staffer and answers questions pulled straight from the site.

For agencies, that speed turns cold outreach into a hyper-personalized product. You are no longer sending generic Loom videos or PDFs; you are sending a live agent that: - Greets callers in a natural, low-latency voice - Answers niche questions from the business’s own knowledge base - Demonstrates 24/7 support without touching their existing stack

That “shareable link” is the closer. ElevenLabs generates a public preview URL where the lead can open a browser, hit a button, and start talking. No login, no install, no sandbox credentials—just an immediate, visceral “oh, this already works for my business” moment.

For a local plumber, restaurant, clinic, or law office, this turns your offer into a no-brainer. You are not asking them to imagine ROI; they hear it in real time as the agent explains tankless water heaters, booking policies, or insurance details using their own copy. The objection shifts from “Does this tech work?” to “How fast can we roll this out?”

That is where upsell gravity kicks in. Once a prospect experiences a working demo, conversations naturally move to call routing, CRM integration, multi-language support, and analytics. You sell the ongoing design of flows, escalation rules, and guardrails—not the one-minute setup.

Anyone building a serious practice around this will live in the docs. ElevenLabs details agent configuration, tools, and deployment options in the Agents Platform Overview – ElevenLabs Documentation, which effectively becomes your playbook for turning quick demos into long-term retainers.

Rise of the 'Conversational Artist'

No-code platforms always trigger the same anxiety: if anyone can spin up AI agents from a URL in 60 seconds, what happens to the people who actually understand the tech? With ElevenLabs agents, that fear feels sharper, because the scaffolding work—APIs, hosting, transcription, TTS—vanishes behind a single create button.

Jannis Moore’s answer in the demo is blunt: you stop being a “tech expert” and start being a conversational artist. When a plumber’s FAQ bot goes from idea to live voice agent in under a minute, the scarce skill is no longer wiring the stack; it is deciding what the agent should say, when, and why.

Designing that layer is harder than it looks. A good conversational artist defines the agent’s main goal, crafts a system prompt that actually constrains behavior, and curates the knowledge base so the RAG pipeline does not surface outdated prices, wrong service areas, or half-baked blog advice. ElevenLabs can ingest an entire site, but it will happily repeat whatever junk you leave in.

Edge cases are where amateurs crash. A real deployment has to survive: - Customers outside the service area - Emergencies at 3 a.m. - Conflicting information across multiple pages - Accents, background noise, and people who ramble

Handling that means building guardrails: escalation rules to a human, fallback answers when confidence drops, and clear boundaries on what the agent will not guess. Your value shows up in fewer hallucinations, higher task completion, and a measurable lift in bookings or reduced call volume.

Businesses do not care that ElevenLabs handles turn-taking at low latency or supports 31+ languages; they care that call abandonment drops 20% and revenue per lead rises. Aligning an agent’s behavior with those ROI numbers—choosing which questions to prioritize, how aggressively to upsell, when to stop talking and just book the job—is not a checkbox in a dashboard.

That is the new job description. Tools like ElevenLabs agents erase the technical barrier, then brutally expose whether you can design conversations that survive real customers, real stakes, and real money.

Real-World Blueprints That Drive Revenue

Most people will demo this with a plumber, but the real money shows up when you apply the same 60‑second build flow to higher‑value local businesses. Anywhere customers call with the same questions, at all hours, you can drop in an ElevenLabs agents link and quietly replace a chunk of front‑office overhead.

Start with real estate. A real estate agent bot can ingest an entire brokerage site—listings, agent bios, neighborhood guides—and instantly answer questions like “Is 1243 Pine Street still active?” or “What are the HOA fees on the loft near downtown?” Instead of bouncing between Zillow, email, and voicemail, buyers talk to a single voice that knows every property detail the site exposes.

Because the agent runs 24/7, it can qualify leads while humans sleep. You define a goal like “capture contact info and schedule viewings,” and the system prompt steers the bot to: - Confirm budget, timeline, and must‑have features - Offer 2–3 matching listings from the site - Book a viewing directly into a connected calendar

Hospitality gets the same upgrade. A restaurant agent can answer “Do you have a table for four at 7 p.m.?” without putting anyone on hold, then push the booking into OpenTable or an in‑house system. It can read the online menu to handle “gluten‑free,” “nut allergy,” or “vegan” queries with more consistency than a rushed Friday‑night host.

Because it sits on top of the live website, updating the menu or holiday hours updates the agent automatically. That means fewer angry calls about outdated hours on Google Maps and fewer no‑shows from guests who never got a confirmation text.

E‑commerce is where this quietly turns into a revenue engine. An e‑commerce support agent can use order‑lookup tools and your product catalog to handle “Where is my package?”, “Start a return,” and “What charger works with this laptop?” in one continuous conversation. No ticket numbers, no IVR maze.

Once it knows the catalog, the agent can behave like a salesperson, not just a help desk. After processing a return, it can recommend in‑stock alternatives, cross‑sell accessories, and surface bundles that already exist on the site—all driven by the same URL you paste in at setup.

Where ElevenLabs Leads the AI Pack

Most AI agent platforms today feel like stitched-together Frankenstacks. You wire up separate tools for speech-to-text, a large language model, a vector database, and telephony, then pray the latency graph does not spike. ElevenLabs agents flip that script by owning the hardest part of the stack: voice that sounds human and responds almost instantly.

Other players can bolt voice on top of generic chatbots, but ElevenLabs built its reputation on ultra-realistic TTS first. Their models power more than 5,000 voices across 31+ languages, with latency low enough for natural interruptions and back-and-forth. That sub-second response is what makes a plumber FAQ call feel like a conversation, not a voicemail tree.

Most competitors still ask you to wrangle schemas, flows, or complex prompt templates. ElevenLabs agents reduce setup to something that looks closer to “paste URL, define goal, click create”. In Jannis Moore’s demo, a working customer support agent goes from random Google result to live, talking assistant in under 60 seconds.

That URL-to-agent workflow is not just a party trick; it is a major product decision. Instead of forcing you to manually build a knowledge base, the platform crawls the site, auto-generates a system prompt, and wires up retrieval so the agent can quote specific services, hours, and policies. You move from “blank canvas” to “draft that actually knows the business” in one step.

Where many teams juggle 3–5 vendors, ElevenLabs ships an all-in-one conversational stack. Inside a single platform you get: - TTS for lifelike output - ASR for fast, accurate transcription - Agents for orchestration, memory, and tools - Telephony for real calls and routing

That consolidation matters for agencies and enterprises that care about SLAs, debugging, and compliance. Fewer APIs mean fewer failure modes and a single place to monitor quality, latency, and cost. For a deeper breakdown of how this architecture works, ElevenLabs details their approach in Everything You Need to Know About Conversational AI Agents – ElevenLabs Blog, which reads like a blueprint for where voice-first AI is heading.

Build Your First Agent Before You Finish This Article

You do not need another tutorial video. You need to open a tab and build your first agent before you scroll any further.

Start with a website you actually care about: your own business, a client, or a local shop you know well. Restaurants, dentists, HVAC companies, law firms, and gyms all have FAQ-heavy sites that work perfectly with ElevenLabs Agents.

Head to the ElevenLabs agents dashboard and click “New agent.” Pick “Business agent,” then “Professional services,” then “Customer support,” mirroring the plumber demo that went from zero to working agent in under 60 seconds.

When you paste the URL, the system auto-generates a system prompt, but the “main goal” field is where you quietly decide whether your agent is mediocre or money-making. Write the main goal as a single, sharp instruction that defines role, audience, and outcome.

Strong main goals usually follow a pattern: - “Create a friendly customer support agent for [BUSINESS] that answers FAQs, collects contact details, and books appointments when possible.” - “Create a sales-focused agent for [SERVICE] that qualifies leads, explains pricing clearly, and nudges callers to schedule a free consultation.” - “Create a support agent for [SOFTWARE] that troubleshoots common issues and escalates complex cases to human staff via email.”

Use concrete verbs: “answer,” “qualify,” “book,” “collect,” “explain.” Avoid vague goals like “help customers” or “talk about our services.” The clearer your main goal, the cleaner your auto-generated system prompt and the fewer hallucinations you will fight later.

After creation, add the URL again under Knowledge Base → “Add document” → “Add URL” so the agent pulls the full site content, not just the landing page. That extra step often surfaces pricing, service areas, and policies buried three clicks deep.

Now click “Preview agent,” say “hey,” and ask questions real customers ask: “Do you serve my ZIP code?”, “What are your emergency fees?”, “Can you come today?” If it holds up, hit “Copy link,” text that link to your own phone, and experience the same demo your next client will.

Frequently Asked Questions

What is ElevenLabs Agents?

ElevenLabs Agents is a no-code platform that allows users to build and deploy hyper-personalized, voice-enabled AI agents for tasks like customer support, sales, and scheduling in minutes.

Do I need coding experience to use ElevenLabs Agents?

No, you do not need any coding experience. The platform is designed to be completely no-code, allowing you to create a functional voice agent simply by providing a website URL and defining a goal.

How does the AI agent learn about a specific business?

The agent uses Retrieval-Augmented Generation (RAG). When you provide a website URL, it automatically scrapes the content, indexes it, and uses that information as a knowledge base to provide accurate, context-aware answers.

Can these AI agents be used for phone calls?

Yes, ElevenLabs' Conversational AI 2.0 platform includes full telephony support, enabling agents to handle both inbound and outbound calls through SIP trunking integration.