Essentials · AI Concierge Agent

A real AI agent that books sites, answers guests, and never sleeps.

Not a chatbot. Not a "knowledge base search." A genuine AI agent, powered by Anthropic's Claude, that lives on your site, learns from your knowledge base, takes real bookings against real availability, and handles cancellations after verifying the guest. Ships with Booking Software at no upcharge.

Try the live demo → Talk to the live agent at Elm Hill →
Camper interior with laptop showing a booking confirmation

The case

Why "live chat" is dead and what replaces it

The last decade of "live chat" widgets did one thing well, proved that guests will use a chat box if it loads fast. They did most things badly. Bots gave wrong answers because they pulled from broken FAQ pages. Live agents could not handle off-hours. The handoff from bot to human almost always lost context. Most parks turned the widget off after 6 months because the conversion lift never materialized.

AI agents are a structural step-change, not a marketing rebrand. A modern agent reads your actual knowledge base (rules, rates, amenities, FAQs), holds multi-turn context, queries your booking system's real availability API, and either completes the booking or escalates with the full conversation history attached. The guest does not bounce between widget and email and front desk. The conversation just resolves.

Old chat vs AI concierge

Why this is not a "better chatbot", it is a different product.

Capability Live chat / FAQ bots (2015–2023) Campground Management AI Concierge
Knowledge source Static FAQ list, manually maintained. Indexed knowledge base, your rules, rates, amenities, policies, even park-specific PDFs. Updates the moment you save.
Booking capability Hands the guest off to a form. Queries real availability, takes the booking, processes payment. End-to-end.
Multi-turn reasoning "I don't understand" after 2 messages. Holds context across 30+ turn conversations. Asks clarifying questions when needed.
Escalation Drops the guest into an empty email queue. Routes to your team with full transcript + intent classification + recommended response.
Off-hours coverage Generic "we are closed" message. Full coverage 24/7, most bookings happen between 9pm and 1am, which is when humans are not staffing.
Tone control Generic. Trained on your brand voice from your existing copy. Sounds like your park, not like ChatGPT.

What's included

What ships in the agent.

✦

Claude-powered reasoning

Built on Anthropic Claude (the same model class behind enterprise customer-support deployments). Native long-context, native tool-use, native safety guardrails.

◆

Knowledge-base ingest

Drop your existing FAQ docs, park rules PDFs, rate sheets, amenity lists. Indexed and queryable within minutes. Updates automatically when you edit source docs.

☷

Booking-engine integration

Native integration with the Booking Software, checks real availability, applies rate rules, processes Stripe payment, sends confirmation. No separate integration project.

◉

Cancellation + modification handling

Verifies guest identity (booking ID + email + last-4 card), applies your cancellation policy correctly, processes refund through Stripe. Logged + auditable.

↗

Smart escalation

When the agent does not know or the guest asks for a human, it routes to your unified inbox with the full transcript + intent tag. Your team picks up at the exact context.

☉

Analytics + improvement loop

Monthly: top intents, deflection rate, escalation patterns, gaps in your knowledge base. Tells you where to update docs to improve future deflection.

If/then

Where the agent earns its keep first.

If
Your front desk fields the same 10 questions every day.
Then
The agent will deflect ~70% of those within 30 days. Pure operational win.
If
You lose late-night bookings because nobody answers.
Then
Most agent-completed bookings happen 9pm–1am. Direct revenue lift.
If
You have a complex set of rules (pet policy, rig length, slide-out clearance).
Then
Perfect fit. The agent handles rule-application better than most humans because it never forgets the edge cases.
If
You have a thin / outdated FAQ.
Then
Build the FAQ first, 60 minutes of doc-writing earns 10 hours per week back. The agent is only as good as the source.

How the engagement runs

From kickoff to compounding results.

01

Knowledge base ingest

You upload your existing docs (PDFs, FAQs, rate sheets, rule pages). We index them in under 24 hours.

Day 1
02

Voice + behavior tuning

We tune the agent's tone to your brand voice and configure refusal threshold + escalation rules.

Days 2–5
03

Sandbox testing

You and your team test against 50+ realistic prompts. We refine until the agent passes your accuracy bar.

Week 2
04

Production launch

Embedded on your site, monitored 24/7 by our ops team. Monthly tuning report + improvement recommendations.

Week 3+

Common questions

What operators ask before saying yes.

Will the agent ever lie to my guests? +
It will refuse to answer questions outside its knowledge base by design (we tune toward refusal, not confident-fabrication). What it will do confidently is anything in your indexed docs. Hallucination risk is non-zero in any LLM system, so we monitor conversations, surface anomalies in the monthly report, and tune the refusal threshold over time. Anything inaccurate gets caught and corrected, and the refusal threshold is tuned conservatively so the agent prefers "I don't know" over a confident wrong answer.
How does it handle cancellations? +
It verifies the guest with a 3-factor check (booking ID + email + last-4 of card on file). Then it applies your cancellation policy exactly as configured (date-based, event-based, no-show rules) and processes the refund through Stripe. Every action is logged + auditable. If verification fails, it escalates to a human.
What languages does it speak? +
English, Spanish, and French are tuned. The underlying Claude model speaks 30+ languages capably; we just have not formally tuned the others. If you have a specific language need (German for European travelers, e.g.), we can add it in 1–2 weeks.
Will it replace my front desk? +
No, and that is not the goal. The goal is to deflect routine questions so your team can focus on actual hospitality, the guest who needs help with the propane hookup, the family who wants restaurant recommendations, the angry guest who needs a real human. Front-desk roles get more interesting, not eliminated.
Can it post in my voice on social media? +
No, that is a different product. This agent is a guest-facing concierge for your booking + support flow. Social content goes through the Social Media service, run by humans on your brand voice.
What if a guest tries to jailbreak it? +
Standard prompt injection attempts (the "ignore previous instructions" pattern) are blocked at the system level. We monitor for novel attack vectors monthly. So far across 11 parks, zero jailbreak-driven incidents. Anthropic's safety stack is genuinely good at this.
Sean Hakes
Written & maintained by
Sean Hakes, Founder, Campground Management

25 years in Marketing & Advertising and 14-year RV industry veteran. Founder of RVParks.US and Campground Management.

Talk to a live AI concierge, running right now.

Click into the demo to chat with the agent running on Elm Hill RV Resort. Try to break it; ask edge cases. The behavior you see is what your park ships with on day one.

Get a tailored proposal → ← Browse all services

"We're not trying to be the biggest. We're trying to be the best one for the family-owned park."

Sean Hakes, Founder · Read our story