12 business automations worth building this week

Most teams lose hours every week to work a machine should be doing: copying numbers between dashboards, chasing unpaid invoices, rewriting one video into ten posts. None of it is hard. It just never gets prioritized because it is nobody's job.

This is the shortlist I'd build first — the top three automations in each of four areas (paid ads, operations, SEO, and content), picked for one thing: hours saved per week against effort to ship. Every one can be standing up in days with tools that exist right now. For each you get the stack, the flow, the AI prompt where one is involved, and the place to keep a human in the loop.

1. Daily cross-channel ad report. Pull yesterday's spend, ROAS, CPA, and CTR from Meta and Google Ads and drop a clean summary in Slack before you open your laptop. Build this one first if you run ads: it is read-only, so it proves your API plumbing with zero risk to spend.

Saves ~4–6 hrs/week · Beginner · Make → Meta Marketing API + Google Ads API → Google Sheets → Claude → Slack

1. A scheduler fires at 7 AM.
2. Call the Meta Marketing API /insights endpoint for spend, actions, CTR, and purchase ROAS at campaign level, date_preset=yesterday.
3. Query the Google Ads API with a GAQL searchStream for cost, conversions, CTR, and average CPC DURING YESTERDAY.
4. Append both to a Google Sheet, one row per campaign per day.
5. Pass the combined data to Claude, then post the summary to Slack.

You are a paid-media analyst. Given yesterday's campaign metrics and the 7-day averages from the sheet, write a 5-bullet Slack summary: total spend, blended ROAS, best campaign, worst campaign, and one thing to watch. Flag in red any campaign where CPA rose more than 30% versus its 7-day average, or spend over $X with zero conversions. Be terse.

Watch out: Meta only returns purchase_roas if a purchase event actually fired, so handle the empty case. Google Ads cost comes in micros, so divide by a million. Both APIs rate-limit per account, so stagger calls if you manage several.

2. Auto-pause underperforming ads. Check active ads on a schedule and pause any that are burning spend with no conversions, then notify you so you can override. This protects budget while you sleep.

Saves ~2–3 hrs/week plus wasted spend · Intermediate · n8n → Meta + Google Ads APIs → Google Sheets (your rules) → Slack

1. n8n runs every four hours.
2. Fetch active ads with today's spend and conversions.
3. Read your rules from a sheet, e.g. pause if spend over $50 and zero conversions today, or CPA above twice target.
4. If an ad breaks a rule, set its status to PAUSED via the platform API.
5. Slack the ad name, the spend, and an undo option with the ad ID.

Watch out: this spends and saves real money, so run it in notify-only mode for a week first (alert, don't act) until you trust the rules. Gate on a minimum spend, not just elapsed time, or you'll kill an ad at 9 AM that would have converted by 5 PM.

3. Generate ad creative from a brief. Take one offer brief and spin up 10–20 copy variations, optionally paired with generated images, dropped into a sheet for review. This feeds the algorithm the creative volume it wants without you writing fifteen versions by hand.

Saves ~5–8 hrs/week · Intermediate · Make → Airtable → Claude → image API → Airtable approval → Meta API as paused drafts

You are a direct-response copywriter. From this brief, write N distinct Meta ad variations. Vary the angle across pain-point, social proof, curiosity, urgency, and benefit-led. Each: primary text under 125 words, headline under 40 characters, one CTA. Return strict JSON. No emojis unless the brand tone says otherwise.

Watch out: Meta rejects banned claims around health, money, and “you,” so add a compliance check. Generated images get artifacts, which is why human review stays mandatory before any spend. Keep variations genuinely different or you'll split budget across near-duplicates.

4. New client onboarding orchestrator. When a deal closes, spin up everything a new client needs (folder, project board, contract, kickoff email, Slack channel) with no manual setup.

Saves ~4–6 hrs/week · Intermediate · Make → CRM webhook → Google Drive → ClickUp/Asana → PandaDoc → Gmail → Slack

1. A deal moves to Closed Won and fires a webhook.
2. Copy a Drive template folder and rename it for the client.
3. Create a project from a task template, assign an owner, set dates relative to today.
4. Generate the contract from a template, prefilled, and send to sign.
5. Send a branded kickoff email, create the client Slack channel, invite the team.

Watch out: double-check your template IDs or you'll clone the wrong board. If terms vary per deal, gate the contract send behind a human check. Slack channel creation rate-limits at volume, so queue the calls.

5. Monday ops dashboard. Every Monday, pull the numbers that matter (revenue, pipeline, active projects, overdue tasks, support backlog) into one digest, so nobody spends the morning building a status deck. Build this one first if you are in ops: it is read-only and the connections it sets up get reused everywhere else.

Saves ~3–5 hrs/week · Beginner · Make → Stripe + HubSpot + ClickUp + Intercom → Google Sheets → Claude → Notion/Slack

You are a chief of staff. Given this week's KPIs and last week's, write a 6-bullet exec summary covering revenue and MRR trend, pipeline health, delivery status, and support load. Flag in red anything declining more than 15% week over week, or any overdue count above the threshold. End with the single most important thing to address this week.

Watch out: Stripe MRR isn't one tidy endpoint, so pull from subscription data or a tool like ChartMogul. Normalize timezones so “last week” means the same thing in every tool.

6. Invoice and AR follow-up autopilot. Watch for unpaid invoices and send escalating, polite reminders on their own, then flag a human only when something goes truly overdue. The real return here is faster cash.

Saves ~2–4 hrs/week plus quicker collection · Intermediate · Make → Stripe/QuickBooks/Xero → Google Sheets → Gmail → Slack

1. Daily cron at 9 AM queries for invoices past their due date.
2. Branch on days overdue: a friendly nudge at three days, a firmer note at seven with the owner cc'd, a Slack alert to finance at fourteen.
3. Send the stage-appropriate email with a payment link.
4. Log every touch so nobody double-chases.
5. When the paid webhook fires, stop the sequence and post a confirmation.

Write a payment reminder for invoice #X ($amount, N days overdue) to the client. Scale the tone with lateness: day-3 friendly, day-7 firm but polite, day-14 direct. Always thank them, include the payment link, and offer to help if there's an issue. Under 90 words.

Watch out: tone is brand risk, so never let the day-14 email reach a key account without a human glance, and route VIPs to manual. Use the paid webhook to kill sequences instantly, or you'll dun someone who already paid.

7. Programmatic page generation at scale. Turn a structured dataset (locations, use-cases, integrations, X vs Y) into hundreds of unique, indexable landing pages. The trick is that each page has to be genuinely useful, not thin.

A big one-time build, then near-zero per page · Advanced · Airtable → Make → Claude → CMS API → GSC + IndexNow

Write a unique landing page for {service} in {city}. Use these data points: {stats, local facts, FAQs}. Sections: H1, an 80-word intro, three benefit blocks, a localized paragraph built from the data, and four FAQs. It must be specific to this city and service - no boilerplate that would read identically for another row. Natural keyword use, no stuffing.

Watch out: thin, duplicate content gets deindexed, so enforce a minimum unique-data threshold and skip rows with sparse data. Roll out in batches and watch indexation in Search Console before scaling to thousands. And build the internal links, or every page ends up orphaned.

8. Keyword to content brief generator. Take a target keyword, analyze who is ranking, and output a writer-ready brief: intent, outline, entities to cover, questions to answer, word-count target. Start here for SEO — it is fast, low-risk, and feeds your whole content engine.

Saves ~5–8 hrs/week · Intermediate · Airtable → Make → DataForSEO/Ahrefs/Semrush → Claude → Google Docs/Notion

You are an SEO strategist. Given the keyword, its volume and difficulty, the top-10 titles and headings, and the People Also Ask questions, produce a brief: search intent in one line, recommended title and H1, an H2/H3 outline that covers what top results cover plus a gap they miss, entities to include, PAA questions to answer, and a target word count based on competitors.

Watch out: SERP and keyword APIs bill per call, so cache results and don't re-pull the same keyword. Heading scrapers choke on JS-rendered sites, so keep a fallback. The brief is a starting point, so keep an editor to catch intent the model misreads.

9. Automated internal linking engine. Crawl your site, find pages that should link to each other based on topical relevance, and either surface the links for review or insert them. Internal linking is the most under-done on-page lever there is.

Saves ~3–6 hrs/week · Advanced · Screaming Frog → Supabase pgvector or Pinecone → Make → Claude → CMS API or a review sheet

1. Weekly crawl, or trigger on a new post.
2. Embed every page into a vector database.
3. For each page, run a similarity search for related pages it doesn't already link to.
4. Claude proposes natural anchor text and the sentence to place it in.
5. Auto-insert for low-risk pages, or queue suggestions in a sheet for approval.

Watch out: don't over-link, since too many internal links dilute equity and read as spam, so cap them per page. Avoid exact-match anchor over-optimization. Default to review mode before inserting into live content.

10. One long-form piece, repurposed everywhere. Take one pillar piece, say a YouTube video, podcast, or blog post, and spin it into a week of platform-native posts: tweets, LinkedIn posts, captions, a newsletter blurb, short-form scripts. This is the highest-ROI content automation there is, because most creators already make the pillar content and waste most of it.

Saves ~6–10 hrs/week · Intermediate · YouTube/RSS → Make → transcript → Claude → Airtable → Buffer/Blotato/Metricool

From this transcript, create platform-native content. Twitter: five standalone tweets plus one thread, hook-first, no hashtags. LinkedIn: two posts with line breaks and one takeaway each. Instagram and TikTok: three video hooks for the first three seconds, with captions. Newsletter: a 100-word summary plus the link. Match this voice: {voice notes}. Each piece must stand alone.

Watch out: platform-native means a tweet isn't a LinkedIn post, so don't cross-post identical text. Keep a human edit step, because repurposing drifts from your voice. Auto-captions get messy, so clean them before feeding the model.

11. Multi-platform scheduler from one calendar. Fill one calendar in Airtable or Notion and let it publish each post to the right platforms at the right times, with platform-specific formatting and media.

Saves ~4–7 hrs/week · Intermediate · Airtable/Notion → Make → Blotato/Buffer or native APIs → Slack

1. The scheduler checks the calendar every 15 minutes for approved posts that are due.
2. Read the copy, media, target platforms, and time.
3. Format per platform — trim or expand text, map aspect ratios, adjust hashtags.
4. Publish through Blotato or Buffer, or the native API for each network.
5. Write the live post URL back to the row, confirm in Slack, and alert on failure so it can re-queue.

Watch out: every platform has approval hoops — Instagram and TikTok publishing need a business account and app review, and the X API has tier costs. Match each network's media requirements. Always write back success or failure, because a silent failure means you think you posted and didn't.

12. Trend and idea sourcing engine. Scan where your niche talks (Reddit, X, YouTube, Google Trends, competitor posts) and deliver a ranked list of ideas and angles every morning, so you never start from a blank page.

Saves ~3–5 hrs/week · Intermediate · Make → Reddit + X + YouTube APIs + Google Trends → Claude → Airtable → Slack

Here are trending items in {niche} from Reddit, X, YouTube, and Google Trends, with engagement metrics. Pick the five with the most content potential for {brand}. For each, give a punchy hook, the best format, why it's timely, and the source link. Skip anything off-brand or already saturated by us. Rank by opportunity.

Watch out: API access is the friction here, since Reddit and X both have paid tiers now, so budget for it or fall back to RSS. Filter hard for relevance or you'll drown in noise, and act within a day or two, because trends move fast.

If you only build one, make it the read-only reporting automation for whatever you spend the most time on: the daily ad report, the Monday ops dashboard, or the content idea digest. They carry no risk because they don't change anything, they pay off the first morning you don't build a deck by hand, and they force you to wire up the API connections that every other automation here reuses.

After that, the order is simple: get visibility first, then automate the safe actions (scheduling, reminders, repurposing), and save the money-moving and publishing steps for last, with a human checkpoint on anything that touches spend, customers, or your live site.

These twelve add up to north of 50 hours a week if you run all four areas, and the bigger wins don't even show up in that number: cash collected faster, ranking drops caught in hours instead of weeks, dropped action items that no longer slip. Pick the one that maps to your most painful weekly chore and build it this week. The plumbing you set up will carry the next three.