Your audience asks the same 12 questions. Answer them once.

Sell a course, a community, or a paid product of any kind, and within a month you can predict your inbox. Not vaguely — literally. The same dozen questions, arriving through Instagram DMs, email, and comments, every single day, from different people who each deserve a real answer and each get a slightly worse-typed version of the one you gave yesterday.

Here's the list. Check yours against it:

1. I paid but I didn't get access.
2. Where do I log in?
3. Payment failed but the money left my account.
4. What's the refund policy?
5. Is there a discount? / My coupon isn't working.
6. Does the course cover [specific topic]?
7. When are the live sessions? Are they recorded?
8. How long do I keep access?
9. Can I download the videos?
10. Do I get a certificate?
11. Can I pay another way — installments, a different method?
12. How is this different from your free content?

If you're answering these one at a time, the problem isn't the questions — every one of them is legitimate. The problem is where you're answering them. A DM is write-once, read-once. You type a careful answer, one person reads it, and it's gone. Tomorrow the same question arrives from someone new, and the work starts over. Multiply by 20–30 messages a day at two or three minutes each, and you're spending an hour a day re-typing a document that already exists in your head.

Write it once, in a place that compounds

Move each of the twelve into a help center article. Not a masterpiece — 150 to 300 words each, titled the way people actually ask ("I paid but didn't get access", not "Access Provisioning FAQ"). Your source material is your last 50 DMs: your best past answers, already written, waiting to be promoted from chat history to documentation.

Once published, each article works three shifts at once:

It answers in the chat bubble. With the AI agent on, a visitor who types "how long do I have access?" into the messenger on your site gets the answer from your article, instantly, at any hour — including the 11pm-after-payday hour when course buyers actually shop. The AI answers only from what you've published; if someone asks something outside your articles, it stays quiet and the thread waits for you, instead of improvising your refund policy on your behalf.

It answers on the open web. Your published articles form a public help site — on your own subdomain if you want it there. People do google "yourname course refund policy" before buying; now there's a page, with your words on it, instead of a Reddit thread guessing.

It answers pre-emptively. With a proactive message targeted at your checkout page, you can surface "Questions about refunds or access? Ask here" exactly where hesitation happens. Pre-sales doubt answered at the moment of doubt is the cheapest conversion work you'll ever do.

One writing session. Three surfaces. Every future asker served by work you did once.

The two questions that still need you

Look back at the list: numbers 1 and 3 — paid-but-no-access and charged-but-failed — aren't knowledge questions. They're account problems. No article fixes a payment that's stuck between a gateway and your course platform, and an AI shouldn't pretend to.

What changes is how they reach you. Instead of a panicked DM with a cropped screenshot and no email address — followed by four messages of archaeology to figure out who this person even is — the conversation arrives in your inbox as a thread, with the AI having stayed out of the way, ready for you to resolve with full context. The routine ten questions stop competing for your attention, so the two that genuinely need you get it faster. That's the actual win: not zero support, but support where your minutes go only where your judgment is required.

The launch-week version

Support volume doesn't grow smoothly for creators — it spikes. Launch week can run five to ten times your baseline, precisely when you're busiest. The set-up sequence that survives it:

A week out: write the twelve articles. One afternoon, sourced from your DM history. Publish.

Three days out: put the messenger snippet on your sales page and course site — it's two script tags, a five-minute job for you or whoever built the page.

Launch day: AI on. The twelve answer themselves while you're doing the actual launch.

Every evening of launch week: open Insights and read the knowledge gaps — the questions the AI couldn't answer, deduplicated and ranked by how many people asked. Launch weeks surface question #13 and #14 fast ("does the bundle include the old course?"). Each gap has a write-article button that starts the draft with the question as the title. Two new articles a night, and by day four your coverage fits this launch, not the last one.

The hour you get back

The arithmetic is almost embarrassing. Thirty routine messages a day, two minutes each: an hour a day, seven hours a week, a full working month per year — spent being a slightly slower, sleepier version of a document.

The replacement isn't an hour of managing a support tool. It's ten minutes a week reading the gaps list and occasionally writing a 200-word article — each one a question you'll never type again.

You started creating to make things once and have them reach many people. Your support deserves the same leverage as your content. Answer each question once, properly, in a place that remembers — and spend the recovered hour making the thing your audience actually came for.