A 24/7 AI receptionist trained on your business.

Yes, occasionally — but we minimize this in three ways: (1) The system prompt explicitly tells the bot to refuse to answer questions outside its knowledge and to offer human handoff instead. (2) The knowledge base is grounded in your actual content, not the model's general training data, so it can't make up facts about your business. (3) During the 30-day post-launch tuning we monitor real conversations and patch any wrong answers as they come up. For high-stakes questions (pricing, legal, medical) the bot is tuned to defer to a human.

We test extensively for adversarial inputs during stress testing. Standard prompt-injection attempts ("ignore your previous instructions and…") are blocked by the system prompt structure. The bot also has guardrails to refuse to give advice on competitors, refuse to discount prices on its own initiative, and refuse to discuss controversial topics. No bot is 100% bulletproof but ours fails gracefully — at worst it refuses to engage and offers human handoff.

No — we charge a one-time setup fee and that's it. The actual model API usage is billed directly to YOUR account with the AI provider (Anthropic for Claude, OpenAI for GPT-4). For most small business chatbot traffic this runs $5–$50/month depending on conversation volume — way cheaper than competitor SaaS chatbot platforms that charge $50–$300/month.

Two layers of training: (1) System prompt — a long, detailed instruction set that tells the model who your business is, what you sell, your hours, common customer questions, your brand voice, and what NOT to say. (2) Knowledge base — your actual content (service pages, FAQs, GBP posts, internal docs) is made available to the model on every conversation via retrieval-augmented generation, so it cites real facts from your site rather than guessing.

Yes — if you have a calendar system (Google Calendar, Cal.com, Calendly, or our custom booking system). The chatbot can ask the user for their preferred time, check your real availability, and book directly. For complex booking (multi-step, requires staff matching, etc.) the chatbot collects the request and hands off to a human.

When a conversation hits an escalation trigger (user says "speak to a human," angry sentiment detected, complex unresolved question, etc.) the bot acknowledges and offers two paths: (1) Email handoff — user provides contact info, we email you the full transcript and they get an "We'll respond shortly" confirmation. (2) Live handoff — only if you've set up a real-time chat tool (we can integrate with Crisp, Tidio, etc.). Most clients use the email handoff approach.

Conversation transcripts are stored in your CRM (Suncoast CRM if you're a client; HubSpot, Pipedrive, etc. if you have one) AND emailed to you per conversation if you opt in. We don't hold a copy on our servers beyond what's needed for the next message. Privacy-conscious clients can opt for ephemeral mode (no transcript retention at all).

They can — every chatbot we build offers an "I'd rather talk to a person" option from the first message and at any point during the conversation. The bot complements your human team, doesn't replace them. Many clients see the chatbot handle 60–70% of conversations completely and the remaining 30–40% become well-qualified human handoffs.

Yes — the dashboard lets you edit the FAQ list, add new common-question answers, and update hours/pricing without touching code. Major changes (brand voice, new services, system prompt rewrites) are typically done by us as part of an ongoing maintenance plan or as one-off updates.

Platform chatbots (Wix Chat, Shopify Inbox) are mostly traditional decision-tree systems with limited LLM integration. Our setup uses a true LLM (Claude or GPT-4) with custom training on your real content — much more conversational, much better at handling unexpected questions. Cost-wise: platform chatbots are typically $50–$300/month forever; ours is one-time setup + ~$5–$50/mo direct API usage to the model provider.