A prospect lands on your site at 11:47pm on a Saturday. They've been thinking about this purchase for three weeks. Tonight, for whatever reason, they're ready. They have two tabs open — yours and your competitor's.
You are in bed. Your competitor has an AI concierge.
This post is about what happened next — and why most "chatbot" implementations make the situation worse, not better.
The story
One of our clients runs a specialist B2B service with an average contract value in the mid-five figures. Long sales cycles. Most deals need a 30-minute call before anyone commits.
A few months after we launched their site with an AI concierge built into the hero, this happened at 2:14am:
Prospect: "Do you work with businesses under $5M?"
AI: "Yes, about half our clients are in that range. What are you trying to solve?"
Prospect: (long paragraph about their specific problem)
AI: (specific, relevant answer referencing their actual situation)
— 11 messages later —
AI: "Based on what you've described, I think a 30-minute call with Jake would be a good next step. Want me to put a hold on his calendar?"
Prospect: "Yes."
By 9am Monday, the meeting was confirmed, the proposal was drafted, and the deal — ultimately worth $48k — was effectively closed.
If the site had just been a site, the prospect would have bounced. If the site had had a generic chatbot with "Hi! How can I help?", the prospect would also have bounced. Both outcomes were close to certain.
Why most chatbots are worse than nothing
The problem isn't the technology. The problem is that most implementations are deployed the way a small business owner installs a speed bump — you bolt it in, you tell everyone it's there, you stop thinking about it. Nobody asks what it's actually doing.
Three specific failures, in order of how much damage they do:
-
Generic training
A chatbot trained on "general web content" answers every question with a shrug and a "please contact support." It's worse than a FAQ page because it wastes the user's time to tell them the same thing the FAQ would have said instantly. Result: lower trust than having nothing.
-
No qualification logic
If every conversation ends with "book a call," you've built a form with extra steps. Good concierges know which conversations to close, which to qualify, and which to escalate. They're not there to just "be helpful" — they're there to move the right conversations forward.
-
No handoff
A prospect who invested 10 minutes chatting, only to get punted to a "check your email for a confirmation," feels punished for engaging. Real concierges remember the conversation, summarise it for the human, and hand it off so the call starts from where the chat ended, not from scratch.
Go to any site with a chatbot and type: "I think your main service is too expensive for my business." If the reply isn't specific, curious, and confident — that chatbot is a liability, not an asset.
What a good one does
A real concierge does four things that matter, in this order:
- Knows the business cold — not just the website content, but the pricing edges, the typical client, the common objections, the wrong-fit signals.
- Holds the conversation — asks follow-up questions, references what was said earlier, maintains tone across ten or twenty messages without sounding like a form.
- Qualifies honestly — says "we're probably not the right fit for you" when that's true. This is counterintuitive. It also dramatically increases trust with the prospects who are a fit.
- Hands off cleanly — writes the internal summary, books the meeting, drafts the proposal, flags the deal worth interrupting a human for.
The quiet economics
Most service businesses get between 30% and 50% of their traffic outside of business hours. If even half of those prospects are serious, and even a fraction of those need real-time engagement to commit, the cost of not having a good concierge is a continuous, invisible leak.
The lead you lost last Tuesday didn't send you an email about it. They just typed your competitor's URL and moved on.
A website is a salesperson. The only useful question is whether that salesperson is awake.
Should you have one?
Short answer: if your average customer value is over about $2,000 and you lose more than 5 leads a month outside business hours, almost certainly yes. Under those numbers, the math gets tight and a well-built contact form may do fine.
If you want us to run the numbers for you, book the session. We'll plug in your real traffic, after-hours ratio, and ACV — and tell you either "do this now" or "don't waste the money." No sales pressure either way.