For most small businesses the cheapest option is free: an auto-reply that answers the common questions and says when a person will reply, followed by the AI agent built into a platform you already use, such as Shopify Inbox's. Paid AI costs from a few cents a message to about $1 per resolved conversation, so choose by monthly after-hours volume.
Volume is the deciding number. At 20 after-hours enquiries a month, a good auto-reply and a quick first-thing sweep cost nothing and take well under an hour a month. At 400, the morning backlog becomes a part-time job, and paying per answer starts to look cheap. Most small businesses sit in between, which is where the sums below matter.
Five ways to cover the evenings, from free to per-answer
These are the realistic options, cheapest first. Which ones you can use depends on where your enquiries arrive: a Shopify agent can't answer the phone, and a phone receptionist can't reply on Instagram.
| Option | Cost | What it handles | Weak spot |
|---|---|---|---|
| 1. Auto-reply with real answers | $0 | Opening times, what you do and don't do, when a person will reply | Can't answer anything specific to the customer |
| 2. AI agent built into your platform | $0 on Shopify Inbox; other platforms vary | Order, delivery and policy questions from your store's own pages | Only as good as the pages it reads |
| 3. Per-message AI in messaging apps | Meta Business Agent: $2 per million tokens, which Meta puts at roughly 4 to 5 cents a message | WhatsApp, Instagram and Messenger questions | Charged on every message, with no free allowance |
| 4. Per-resolution helpdesk AI | Help Scout AI Answers $0.75 per resolution, on paid plans from $25 a user a month; Intercom Fin $0.99 per outcome | Email and website chat | "Resolved" can include customers who simply gave up |
| 5. AI phone receptionist | Jobber Receptionist: $29 for 30 conversations, then $0.79 each | Calls: taking details, booking, answering basics | Hand-offs to a person need careful setup |
There's also a sixth option that costs nothing and gets forgotten: answering everything at 9am. For some businesses that's genuinely fine. The rest of this tutorial is about when it stops being fine, and what the cheapest step up is.
The sums at 20, 100 and 400 after-hours enquiries a month
To compare like with like, the table uses one set of illustrative assumptions. An auto-reply with real answers settles 40% of enquiries outright. Where an AI option is used, it fully resolves 60% of enquiries and the rest wait for the morning. A person takes 4 minutes to answer each waiting enquiry, and that time is valued at $18 an hour. Meta's agent sends about four messages per conversation, roughly $0.18 a conversation, and is charged on every conversation, resolved or not. Help Scout's AI Answers needs a paid plan, so its rows include one Standard seat at $25 a month. Intercom's standalone Fin has a minimum monthly commitment, which its pricing page illustrates as 50 outcomes, so $49.50 is the floor. Jobber counts conversations, not resolutions.
| Monthly after-hours enquiries | 20 | 100 | 400 |
|---|---|---|---|
| Everything answered next morning (staff time only) | $24 | $120 | $480 |
| Free auto-reply with answers + morning sweep | $14 | $72 | $288 |
| Shopify Inbox AI agent + morning sweep | $10 | $48 | $192 |
| Meta Business Agent + morning sweep | $13 | $66 | $264 |
| Help Scout AI Answers (one Standard seat) + morning sweep | $44 | $118 | $397 |
| Intercom Fin (standalone) + morning sweep | $59 | $107 | $430 |
| Jobber Receptionist, calls (fees only) | $29 | $84 | $321 |
To see how one cell is built: at 100 enquiries, Help Scout's AI resolves 60, which costs 60 times $0.75, or $45, and the paid plan it needs adds a $25 Standard seat. The other 40 take 160 minutes of staff time, or $48. Total: $118. At 400 enquiries on Jobber, the first 30 conversations are covered by the $29, and the other 370 cost $0.79 each, $292.30, so $321.30 before anyone returns a call.
Three things jump out. First, at 20 a month the cheapest options are only a few dollars apart, and anything with a fixed monthly charge, such as a Help Scout paid seat, standalone Fin's $49.50 floor or Jobber's $29 bundle, costs more than the staff time it saves. Second, the built-in agent is cheapest at every volume if your enquiries arrive where it works, and the free auto-reply beats most of the paid options at every volume in the table. Third, the gap between options is smaller than the gap between doing something and doing nothing: at 400 a month, even the dearest AI route costs less than answering everything by hand. These are my assumptions, so rerun them with your own counts; the resolution rates in particular depend on how good your written answers are.
A charity shop's donation questions, handled overnight
In this worked example, with invented but realistic figures, a charity shop gets about 90 messages a month outside opening hours, spread across Facebook, Instagram, WhatsApp and email. The manager counted them over a fortnight and sorted them: roughly 40 about donations ("Do you take sofas?", "Can I drop bags off on Sunday?"), 20 about opening times, 15 from people wanting to volunteer, and 15 about everything else. Each morning the manager spent about 20 minutes working through them before the shop opened, around six hours a month.
Step one cost nothing. The manager wrote a donations page for the website (what the shop takes, what it can't, when and where to drop off, and that it doesn't collect), then set up away messages built from it. Meta Business Suite's inbox has free automated responses, including an away message that can be scheduled for your closed hours on Messenger and Instagram. The WhatsApp Business app has an away message with an "outside of business hours" option, which works once your hours are set in the business profile. The email got an always-on acknowledgement pointing to the same page.
The result after a month: the messages still needing a person in the morning fell from 90 to 38, because most people who asked a common question got the answer straight away. Morning time fell from six hours a month to about two and a half.
Step two was a sum, not a purchase. Would Meta's AI agent be worth adding? If it resolved 60% of the 38 remaining messages, that's about 23 fewer for the manager, saving roughly 90 minutes a month, worth about $24 at the manager's $16 an hour. But the agent is charged on every conversation it takes part in. On the messaging channels alone, around 70 conversations at about $0.18 each is $12.60 a month. The saving just beats the cost, and every wrong answer would eat the margin. The manager decided to stay on free away messages and look again if volume passed 200 a month, or when a busy season arrived.
An auto-reply that answers instead of apologising
The difference between a useless auto-reply and a useful one is whether it contains answers. The shop's old message:
Thanks for your message. We're currently closed and will get back to you as soon as possible.
The new one:
Thanks for getting in touch. We're closed now and open again at 9:30am, Monday to Saturday.
Donations: we take clothes, books, toys and small homeware, Monday to Saturday 9:30 to 4:30 at the side door. We can't take large furniture, mattresses or electrical items, and we don't collect.
Volunteering: reply VOLUNTEER and we'll send you the form in the morning.
Anything else: a person will reply by 11am.
The new message resolves the two commonest questions outright and sets a time for the rest. "As soon as possible" promises nothing; "by 11am" is a promise the shop can keep, and people stop sending the same question to a second channel. The same approach works for any business: find your three commonest after-hours questions, answer them in the auto-reply, and name the time a person will pick up the rest.
One practical note on email. Gmail's automatic reply is designed for time away: it runs from the start of one date to the end of another rather than switching on each evening, and the same sender gets it again after four days. For after-hours cover by email, an always-on acknowledgement that states your reply times works better than trying to toggle a holiday reply each night.
Platform by platform: Shopify, WhatsApp, phone and email
Shopify: test the free agent before customers do
Shopify Inbox now includes an AI agent that answers customers on its own, on the Basic plan and above for stores using new customer accounts. It draws on your store's content, and Shopify notes it can use web search as a secondary source, which is exactly why it needs testing. Take a toy shop: before switching the agent on at night, the owner asks it fifteen real questions from last month's inbox, such as "Is this suitable for a two-year-old?", "Can I return an opened jigsaw?" and "Do you gift wrap?". Any answer that isn't on the shop's own pages means a page needs writing, not that the agent should be trusted to improvise. How to build the FAQ your AI chatbot needs covers what to write.
WhatsApp: the app and the platform are priced differently
A members' club that handles court bookings and guest questions on WhatsApp has two different products to think about. The free WhatsApp Business app has the away message at no cost, and Meta's Business Agent can now answer in it for about 4 to 5 cents a message. Meta's platform terms bar general-purpose AI assistants on WhatsApp, but an agent tied to one business job, such as bookings or member questions, is allowed.
The WhatsApp Business Platform, which larger setups and most third-party chatbot tools use, is a separate product with its own pricing change on 1 October 2026: free-form service replies inside the 24-hour customer-service window become chargeable after the first 1,000 service messages per business number each month, and a payment method must be on the account. Rates vary by the recipient's market, so check Meta's pricing page rather than trusting a quoted figure. WhatsApp Business app vs API for AI replies explains which one you need.
Phone: voicemail first, receptionist second
A voicemail greeting that answers the top three questions ("We're open 9:30 to 4:30, Monday to Saturday; donations go to the side door; for anything else, leave your number and we'll call by 11") costs nothing. Businesses where a missed call is a lost job, such as trades, are where an AI receptionist earns its fee; how to set up an AI phone agent for after-hours calls walks through the setup and the hand-off rules.
Email and website chat: per-resolution helpdesks
Help Scout's AI Answers costs $0.75 per resolution as an add-on to its paid plans, which start at $25 a user a month on Standard; the free plan, for up to five users, doesn't include it. Intercom's Fin costs $0.99 per outcome and can run alongside another helpdesk without Intercom seats, subject to the minimum monthly commitment. HubSpot's Customer Agent costs about $0.50 per resolved conversation in credits, but it needs a seat on any Professional or Enterprise hub (usually Service Hub), so it's cheap only if you're already on those plans.
What "resolved" means on a per-answer bill
Per-resolution pricing sounds like paying only for success. Read the definition, because out-of-hours conversations are exactly where it bends:
- Intercom counts an "assumed resolution" when a customer goes quiet for 24 hours after Fin's last answer.
- On Zendesk a messaging conversation closes after 2 quiet hours by default (72 on email and web forms); an AI check then decides whether it counts as a verified resolution, the only kind billed since 18 May 2026, so hand-offs and unverified closes cost nothing.
- HubSpot's Customer Agent uses 50 credits, about $0.50, per resolved conversation.
- Help Scout doesn't bill a resolution if the customer escalates, searches the help docs, asks again or says they need more help.
The definitions differ, and the differences matter most at night. At 11pm, a customer who goes quiet may be asleep, satisfied, or fed up. A toy shop owner who reviews 20 of the month's 120 billed resolutions might find that five customers had given up and emailed again the next morning. Those five would have been paid for twice: once to the AI, once in staff time. Scaled up, that's about 30 of the 120 resolutions that weren't really resolutions.
The fix is a weekly sample: read ten "resolved" conversations and count how many customers came back. If more than one or two did, improve the answers those conversations needed before paying for more of them. Outcome-based AI pricing goes deeper into how each vendor counts.
When the cheap bot answers wrongly at 2am
Cheap cover goes wrong in a predictable way: the AI finds an old or ambiguous source and answers confidently from it. Had the charity shop switched on an AI agent that read its social pages, this is how it could have gone. A post from years earlier mentions a furniture collection service the shop no longer runs. At 2am the agent tells a donor, "Yes, we can collect your sofa; we'll be in touch to arrange a date." Nobody gets in touch, the donor complains publicly, and the shop spends a morning apologising.
Four habits prevent most of this:
- Remove or correct old information the AI might read: old posts, outdated pages, last year's opening hours.
- Write the "no" answers down. "We don't collect" and "we can't take mattresses" need to be stated plainly, because an AI fills silence with guesses.
- Read the overnight transcripts every morning for the first two weeks, then weekly.
- Say it's an AI and when a person will follow up. What to tell customers at the start of a chat has wording you can adapt.
The first hour of the next day, and the numbers to watch
Whatever you choose, the morning routine decides whether it stays cheap. Fifteen minutes is enough for most small businesses: answer the waiting enquiries, oldest first; read the AI's overnight conversations, if there is an AI; and fix one answer or page a day where the auto-reply or agent fell short.
Track four numbers monthly. The charity shop's first month, as an illustration:
| Measure | Before | After one month |
|---|---|---|
| After-hours messages | 90 | 94 |
| Messages needing a person next morning | 90 | 38 |
| Time until the first human reply for those | Up to a day | By 11am, as promised |
| Wrong or misleading answers sent | Not tracked | 0 |
When the second row starts rising again, or the busy season doubles the first row, rerun the volume sums with your new figures. The cheapest option at 90 messages a month is rarely the cheapest at 300.
Out-of-hours cover: follow-up questions
Should an out-of-hours bot say that it's an AI?
Yes. It's good practice anywhere, because customers forgive a bot's limits more readily when they know what they're talking to. If you sell to customers in the EU, Article 50 of the EU AI Act has required chatbots to tell people they are talking to AI since 2 August 2026. Put it in the first message, with a line on when a person will reply.
Can I just use ChatGPT to answer customers overnight?
Not directly. ChatGPT and Claude subscriptions cover you chatting with the assistant; they don't connect it to your customers, and API use is billed separately, per token. You'd need a platform that does the connecting. On WhatsApp, Meta's platform terms also bar general-purpose AI assistants, while bots tied to one business task, such as orders, bookings or support, are allowed.
Is a human answering service cheaper than AI?
It depends on your volume and how much each call needs a person. Answering services usually charge per call or per minute, so ask two services for a written quote at your real monthly volume and compare it with the per-conversation costs in the table above. A human service often wins for complex or sensitive calls and loses on simple repeated questions.
Most of my out-of-hours enquiries are phone calls. What's cheapest?
A voicemail greeting that answers the top three questions and points callers to your website costs nothing, and many callers never need more. If missed calls are costing work, an AI receptionist is the next step: Jobber Receptionist, for businesses on Jobber, costs $29 for 30 conversations and then $0.79 each. Check any receptionist's hand-off to a human before relying on it.
Further reads
- AI Chatbot vs Live Chat vs Help Desk: What a Small Team Needs — Which of chatbot, live chat or helpdesk a small team needs.
- How Much Does a WhatsApp Chatbot Cost? Fees, Tools and Setup — The full cost picture for a WhatsApp bot, fees included.
- How to Reply to Every Enquiry in Under Five Minutes With AI — Fast replies during opening hours, the other half of the job.
- How Charities Use AI Chatbots to Answer Supporter Questions — How charities set up bots for supporter questions.
- When Should an AI Chatbot Hand Over to a Human? — When and how an AI agent should pass a chat to a person.
- Best AI Chatbots for Small Shopify Stores (2026) — Options for Shopify stores that outgrow the built-in agent.
- Customer-Facing or Back-Office: Where Should AI Go First? — A side-by-side comparison of customer-facing and back-office AI as a first move, with a scoring sheet, the middle route, and two worked decisions.
- How a One-Person Travel Agency Can Reply to Enquiries Instantly — A two-speed reply system for solo travel agents: a useful answer within a minute, then an AI-drafted personal reply you check and send the same day.
- How Car Valeting and Detailing Firms Use AI to Fill the Diary — An empty-slot audit, rebooking intervals, photo-quote prompts and standby-list messages that help a valeting or detailing business keep its diary full.
- How Small Hotels Use AI to Answer Guest Messages Around the Clock — The mechanics of overnight AI replies in a small hotel: how messages flow, what the AI may say at 2am, and how to check its work each morning.
- Can an AI Chatbot Book Valuations for Estate Agents Out of Hours? — What an out-of-hours valuation bot needs: live diary access, eight qualifying questions, written diary rules and a list of things it must never say.
- AI vs Human Answering Service for HVAC Firms — When an HVAC firm should let AI answer the phone, when a person should, and why most end up with a hybrid that escalates emergencies.
- Answering Pupil and Parent Enquiries With AI at a Driving School — Build an answer sheet from your last 50 enquiries, put AI on web chat or WhatsApp, and hand test dates, complaints and progress questions to a person.
- Replying to Trade Enquiries in 60 Seconds With AI — Why most automations can't reply in 60 seconds, the three builds that can, and what a first reply to a roofing or plumbing enquiry should actually say.
- Setting Up an AI Answering Service That Never Misses an Emergency — Emergency definitions, a word-for-word screening script, a transfer chain with backups, and monthly test calls that prove your AI line catches urgent calls.
- Can AI Triage Emergency Plumbing Calls Out of Hours? — How an AI phone agent can sort night-time plumbing calls into wake-me-now, first-thing and advice, with scripted safety lines and a test plan.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Shopify help pages (Shopify Inbox AI agent); Meta pricing notes on Business Agent and the WhatsApp Business Platform; WhatsApp help pages on away messages; Meta Business Suite help on automated responses; Gmail help on automatic replies; Intercom pricing page; Help Scout, Zendesk and HubSpot pricing and billing definitions; Jobber Receptionist pricing.