How to Use Make.com with AI: 5 Real Workflows with Claude & ChatGPT (2026)

Watch your support inbox for new messages. Filter by label or sender domain if needed.

System prompt: You are a customer support classifier. Categorize the email as one of: billing, technical, feature-request, refund, or other. Return only the category label. Pass the email subject + body as the user message.

Use Make's Router to split the flow. Each branch handles one category: billing → finance Slack channel, technical → engineering Notion board, refund → finance team email, etc.

Optionally add a second Claude step: have it draft a suggested reply based on the category. Save the draft to Gmail or post it as a Slack message for a human to review and send.

Your team adds target keywords to a Google Sheet. Each new row triggers the scenario.

System prompt: You are an SEO content strategist. Given a target keyword, write a detailed content brief including: recommended title, meta description, target audience, key topics to cover (H2 sections), 5 related keywords, and estimated word count. Format as structured JSON. Pass the keyword from the sheet row.

Use Make's JSON module to parse the structured AI response into individual fields.

Create a new Notion page in your content calendar database, populated with the brief. Assign to a writer and set status to "Ready to Write."

Fires whenever a new contact is added to your CRM.

Use the contact's company website (from HubSpot fields) to fetch the homepage via Make's HTTP module. Extract the text content.

System prompt: Given this company's website content, write a 3-sentence summary of: what the company does, who their likely customers are, and one relevant talking point for a sales call about our product.

Write the AI summary to the HubSpot contact's notes field. Post a Slack message to the sales channel alerting the team with the enriched profile.

Fires when a new article goes live on your English WordPress site.

Use Make's Iterator to loop through your target markets: Spanish, French, German (or whichever you need).

System prompt: Translate this article to {{language}}. Do not just translate — localize it: adapt examples, idioms, and cultural references to be natural for {{country}} readers. Maintain SEO-friendly structure.

Save each translated version as a Google Doc in a shared folder for review before publishing to regional WordPress installations.

Watch your blog's RSS feed for new articles. Fires on each new post.

System prompt: Given this blog post, write 5 social media posts: 1) LinkedIn (professional, 150 words), 2) Twitter/X (punchy, under 280 chars), 3) Instagram caption (casual, with hashtag suggestions), 4) Facebook (conversational, 100 words), 5) Newsletter teaser (2 sentences that create curiosity). Return as JSON with keys: linkedin, twitter, instagram, facebook, newsletter.

Parse the 5 variants into separate Make variables.

Create one Airtable record per social post with: platform, content, scheduled date (auto-set to 3 days after publish), and status = "Draft." Your social media manager reviews and approves before posting.