It's 2am. A customer in London has a billing issue with your product. Your team is in San Francisco, asleep. The customer sends an email. Eight hours later, when your team wakes up, they respond. By then, the customer has already contacted their bank for a chargeback.

Stories like this make founders think they need 24/7 support. And for some businesses, they do. But for most small to mid-size teams, full 24/7 human coverage is prohibitively expensive. The real question is: what kind of 24/7 coverage do you actually need?

The Cost of 24/7 Human Support

To cover 24 hours with humans, you need a minimum of three shifts. Each shift needs at least one person (for basic coverage) or two (for any redundancy).

At one person per shift, three shifts: 3 full-time agents. At $50,000 per year each (salary, benefits, equipment), that's $150,000/year. For a small team, that might be more than the entire current support budget.

At two people per shift: $300,000/year. That's before you account for weekends, holidays, sick days, and the difficulty of hiring for overnight shifts (overnight workers typically command a 10 to 15% pay premium).

For companies with fewer than 2,000 customers, this math rarely works. The volume between midnight and 6am doesn't justify a dedicated agent. You might get 5 to 15 messages during those hours. Paying someone $25/hour to handle 5 messages is $37.50 per message in labor cost.

What Customers Actually Expect

Research on customer expectations varies by channel:

Live chat: customers expect an immediate response. If you offer live chat, you're promising someone is there. Offering chat during business hours only (with a clear "we're offline" message) is fine. Offering 24/7 chat and then having nobody there is worse than not offering it at all.

Email: customers expect a response within 4 to 8 hours during business hours. Off-hours emails with a 12-hour response time are acceptable for most non-urgent issues.

In-app messaging: similar to chat. If it looks real-time, customers expect a fast response.

Social media: varies. Twitter/X users expect a response within 1 to 2 hours. Instagram DMs are more forgiving.

The key insight: customers don't need a human at 2am for most issues. They need acknowledgment. They need to know their message was received and will be handled. And for a subset of issues (urgent, high-impact), they need actual resolution.

The AI + Human Hybrid

This is where AI changes the economics completely.

During business hours (say, 9am to 6pm your time): human agents handle complex issues. AI handles the simple stuff, plus classification and routing.

During off-hours (6pm to 9am, weekends, holidays): AI handles everything it can. Simple queries get instant responses. Complex queries get classified, prioritized, and queued for human response when the team is back. Urgent issues get pushed to an on-call person's phone.

The customer at 2am in London gets an instant response to their billing question: "I can see the charge on your account. Here's what happened: [explanation]. If you'd like a refund, click here to process it. If you need additional help, our team will review this first thing in the morning."

For simple issues, that response is final. Problem solved at 2am with no human involvement. For complex issues, the customer gets acknowledgment and a timeline, which is dramatically better than silence.

Cost of AI Off-Hours Coverage

AI doesn't have shifts. It handles messages at 2am the same as 2pm. The cost per message is the same: $0.20 for classification, $0.30 for classification plus automated response.

If you get 100 off-hours messages per week (a reasonable number for a company with 1,000+ customers across time zones), AI coverage costs about $80 to $120/week. That's $350 to $500/month.

Compare that to one overnight agent: $3,000 to $4,000/month.

The AI handles 60 to 70% of off-hours messages completely (the simple ones). The other 30 to 40% get classified, prioritized, and queued. Your team processes the queue first thing in the morning.

When You Actually Need Human 24/7

Some businesses genuinely need humans available around the clock. The common thread: urgent, high-impact situations where a delayed response has serious consequences.

Emergency services and healthcare. A patient portal needs someone available for urgent medical questions. AI can triage, but a human needs to be reachable.

Financial services. If a customer's account is compromised at 3am, they need a human who can freeze the account immediately.

Infrastructure and hosting. If your customer's website goes down, they're losing money every minute. AI can acknowledge and escalate, but someone needs to be working on the fix.

Enterprise SLA obligations. If your contract promises 24/7 support with a 1-hour response time, you need humans available.

For everyone else (e-commerce, SaaS, consumer apps, services), AI off-hours coverage with an on-call escalation path for emergencies is sufficient and 10x cheaper than full human coverage.

Setting Up On-Call for Emergencies

Even with AI covering off-hours, you need a safety net for urgent issues.

Define what constitutes an emergency. Not "customer is upset." Something like: "service outage affecting multiple customers," "security breach," "customer reports unauthorized account access," or "legal/compliance emergency."

Set up a rotation. One person per week carries the on-call phone. They don't sit at their desk. They just keep their phone nearby. If AI classifies something as an emergency off-hours, it sends a push notification to the on-call person.

Keep the emergency volume low. If your on-call person is getting woken up every night, either your "emergency" definition is too broad or you have a product quality problem to fix.

Compensate on-call fairly. Most companies pay a flat on-call stipend ($200 to $500/week) plus additional compensation for each off-hours incident handled. Burning out your team with unpaid on-call is a fast way to lose your best people.

The Practical Setup

Here's what most small teams should build:

Business hours (9am to 6pm): full team online, AI handles simple queries, humans handle everything else. Target response time: under 15 minutes.

Extended hours (6pm to 10pm): one agent online (or rotate among team), AI handles the simple stuff. Target response time: under 30 minutes.

Off-hours (10pm to 9am, weekends): AI only, with on-call escalation for emergencies. Target response time for AI: instant. For human follow-up on queued tickets: within 2 hours of business hours opening.

This gives you practical 24/7 coverage at a fraction of the cost of full human staffing. Customers in other time zones get instant AI responses for simple issues and morning responses for complex ones. Emergency situations reach a human within minutes.

Supp handles the off-hours layer. Classification at $0.20 per message, automated responses for simple queries at $0.30, push notifications for emergencies, and everything queued with full context for your team's morning review. Setup takes about 15 minutes.