20 tasks you can automate at work with AI this month
"Automate your work with AI" is easy to say and hard to start. This list is the opposite: twenty specific tasks, grouped by role, each with a suggested tool, a rough difficulty, and a note on where a human should still check the output.
How to use this list
Do not try all twenty. Read the section for your role, pick the one task that annoys you most, and build that first. One working automation teaches you more than five half-finished ones.
For each task you will see:
- What it does: the job the automation takes over.
- Tools: a reasonable starting point. Most tasks can be built in several tools. If you are unsure which automation platform to pick, read n8n vs Zapier vs Make.
- Difficulty: Easy (an afternoon, no logic), Medium (some branching or a spreadsheet lookup), or Harder (several systems, error handling, or an agent).
- Check: what still needs a human, and any privacy concern.
Time savings vary a lot by company and by how messy the input is. Treat any time estimate as a guess until you have measured it on your own work.
Before you build anything: three rules
- Check what tools you are allowed to use. Many companies have rules about which AI tools can see customer, employee or financial data. Ask IT or your manager first. It takes five minutes and avoids real trouble.
- Start with drafts, not actions. Have the automation draft the email, label the ticket or fill the spreadsheet row. Let a person approve it before anything leaves the building.
- Measure the before. Write down how long the task takes you this week. Without a baseline you cannot tell if the automation helped.
Finance
1. Invoice and receipt extraction
What it does: watches an inbox for invoices or receipts, pulls out supplier, date, amount and currency, and writes a row to a spreadsheet.
Tools: n8n or Make, with an AI model (Claude or ChatGPT) to read the attachment, plus Google Sheets or Excel.
Difficulty: Medium.
Check: AI can misread numbers, especially on scanned documents. Have a person confirm every amount before it goes into the books, and flag anything over a set value.
2. Expense policy pre-check
What it does: compares submitted expenses against your written policy and flags items that look out of policy, with a short reason.
Tools: Claude or ChatGPT with your policy document, run from Zapier or Make when a new expense is submitted.
Difficulty: Medium.
Check: the AI flags, a person decides. Never let it reject a claim on its own.
3. Month-end variance commentary
What it does: takes this month's and last month's figures and drafts plain-English notes on the biggest changes for your management report.
Tools: Claude or ChatGPT with an exported spreadsheet. Can be scheduled in n8n.
Difficulty: Easy to start, Medium to schedule.
Check: the AI does not know why numbers moved. It can describe the change; you supply the reason.
4. Overdue invoice reminders
What it does: checks your receivables list each morning and drafts polite reminder emails for invoices past due, in your house style.
Tools: Zapier or Make on a schedule, with an AI step to write the email.
Difficulty: Easy to Medium.
Check: review drafts before sending, especially for key clients or anyone already in a dispute.
Operations
5. Inbox triage
What it does: reads each new email in a shared inbox, labels it by type, drafts replies to routine questions, and flags the few that need a person.
Tools: n8n with Claude, or Zapier with an AI step. This is one of the example builds in the AI for Automation track.
Difficulty: Medium.
Check: keep drafts as drafts at first. Review the labels for a week before trusting them.
6. Monday status report
What it does: collects last week's numbers from your tools, writes a short summary, and posts it to your team before Monday starts.
Tools: n8n or Make on a schedule, an AI model to write the summary, and Slack, Teams or email to post it.
Difficulty: Medium.
Check: confirm the numbers pull correctly for a few weeks. A summary built on a broken data source is confidently wrong.
7. Meeting notes to actions
What it does: turns meeting notes or a transcript into decisions, owners, deadlines and draft follow-up emails, in your company's format.
Tools: Claude or ChatGPT with a saved prompt or template. Can be triggered when notes land in a folder.
Difficulty: Easy.
Check: the person who ran the meeting should confirm owners and dates. Be careful with confidential meetings and which tools see the transcript.
8. Supplier or vendor document review
What it does: reads incoming quotes or contracts and pulls out key terms (price, dates, renewal, notice period) into a comparison table.
Tools: Claude or ChatGPT for the reading, a spreadsheet for the output.
Difficulty: Easy to Medium.
Check: this is a summary, not legal review. Anything important still goes to the right person.
HR
9. Job description drafts
What it does: turns a hiring manager's rough notes into a structured job description in your company's template.
Tools: Claude or ChatGPT with your template and a couple of past examples.
Difficulty: Easy.
Check: review for accuracy, inclusive language and any requirements that are not really required.
10. Candidate application summaries
What it does: summarizes each application against the role's stated requirements so the hiring manager can read faster.
Tools: an AI model connected to your applicant inbox or tracking system through n8n or Zapier.
Difficulty: Medium.
Check: this is sensitive. Check local rules and company policy on using AI in hiring. Use it to summarize, never to reject. A person reads every application that matters.
11. Onboarding checklist and welcome emails
What it does: when a new hire is added, creates their onboarding tasks, sends scheduled welcome emails, and reminds managers of first-week steps.
Tools: Zapier or Make. An AI step can personalize the welcome message.
Difficulty: Easy to Medium.
Check: confirm what personal data moves between systems and that each system is approved for it.
12. Policy question answering
What it does: answers common staff questions (leave, expenses, travel) from your handbook, with a link to the right section.
Tools: Claude or ChatGPT with the handbook loaded, or an n8n workflow that answers questions posted in a chat channel.
Difficulty: Medium.
Check: AI can state wrong answers confidently. Always include the source section and tell staff to confirm anything important with HR.
Sales
13. Prospect research before calls
What it does: given a company name, gathers public information and drafts a one-page brief: what they do, recent news, likely priorities.
Tools: Claude or ChatGPT with web access where available, or an n8n workflow that runs before each calendar event.
Difficulty: Easy to Medium.
Check: verify any fact you plan to mention on the call. AI research can mix up companies with similar names.
14. Lead research and first-message drafts
What it does: takes a list of target companies, researches each, and drafts a first message you would actually send.
Tools: n8n or Make looping over a spreadsheet, with an AI step for research and drafting.
Difficulty: Medium to Harder.
Check: read every message before it goes out. Follow the email and privacy rules that apply in your market.
15. CRM updates from call notes
What it does: turns your notes or a call transcript into structured CRM fields: stage, next step, date, key objections.
Tools: Zapier or n8n connected to your CRM, with an AI step to extract fields.
Difficulty: Medium.
Check: confirm the fields before they save, at least at first. Bad CRM data spreads quickly.
16. Proposal first drafts
What it does: combines your proposal template, the client's brief and your call notes into a first draft.
Tools: Claude or ChatGPT with your template and two past proposals as examples.
Difficulty: Easy.
Check: pricing, scope and commitments must be checked by a person every time.
Marketing
17. Content repurposing
What it does: turns one long piece (an article, a webinar transcript) into several short posts, an email and a summary.
Tools: Claude or ChatGPT with a style guide. Can run in Make or n8n when a new piece is published.
Difficulty: Easy.
Check: edit for voice and accuracy. Do not publish anything you have not read.
18. Social and review monitoring
What it does: collects new mentions or reviews, sorts them by sentiment and topic, and drafts replies for the routine ones.
Tools: n8n or Make with an AI step, posting a digest to Slack or email.
Difficulty: Medium.
Check: a person approves every public reply, especially to complaints.
19. Campaign performance summaries
What it does: pulls weekly numbers from your ad and email tools and writes a short summary of what changed.
Tools: n8n or Make on a schedule, an AI model for the summary.
Difficulty: Medium.
Check: double-check the data pull. AI can describe numbers but cannot tell you if the tracking was broken.
20. Brief and copy variations
What it does: turns a short creative brief into several headline and copy variations for testing.
Tools: Claude or ChatGPT with your brand guidelines.
Difficulty: Easy.
Check: review for claims you cannot back up and anything that breaks advertising rules in your market.
All twenty at a glance
| # | Task | Role | Starting tools | Difficulty |
|---|---|---|---|---|
| 1 | Invoice and receipt extraction | Finance | n8n or Make + AI + sheet | Medium |
| 2 | Expense policy pre-check | Finance | Zapier or Make + AI | Medium |
| 3 | Variance commentary | Finance | Claude or ChatGPT | Easy |
| 4 | Overdue invoice reminders | Finance | Zapier or Make + AI | Easy |
| 5 | Inbox triage | Ops | n8n + Claude | Medium |
| 6 | Monday status report | Ops | n8n or Make + AI | Medium |
| 7 | Meeting notes to actions | Ops | Claude or ChatGPT | Easy |
| 8 | Vendor document review | Ops | Claude or ChatGPT | Easy |
| 9 | Job description drafts | HR | Claude or ChatGPT | Easy |
| 10 | Application summaries | HR | n8n or Zapier + AI | Medium |
| 11 | Onboarding emails | HR | Zapier or Make | Easy |
| 12 | Policy question answering | HR | Claude or ChatGPT, n8n | Medium |
| 13 | Prospect research | Sales | Claude or ChatGPT, n8n | Easy |
| 14 | Lead research and drafts | Sales | n8n or Make + AI | Harder |
| 15 | CRM updates from notes | Sales | Zapier or n8n + AI | Medium |
| 16 | Proposal first drafts | Sales | Claude or ChatGPT | Easy |
| 17 | Content repurposing | Marketing | Claude or ChatGPT | Easy |
| 18 | Mention and review monitoring | Marketing | n8n or Make + AI | Medium |
| 19 | Campaign summaries | Marketing | n8n or Make + AI | Medium |
| 20 | Copy variations | Marketing | Claude or ChatGPT | Easy |
A simple four-week plan
If you want to actually get one of these working this month, here is a pace that fits around a full-time job.
- Week 1: pick one task. Time it. Write down every step you take, in order, and what information you need at each step.
- Week 2: build the simplest version that works end to end, even if it only handles the easy cases. Keep it in draft mode.
- Week 3: handle the messy cases. Add checks, a fallback for errors, and a way for a person to approve the output.
- Week 4: run it on real work for a week. Measure time spent again. Show it to your manager.
That is roughly the arc of the AI for Automation track: audit your week, build your first workflows, add agents and data, then ship it in your job. If you want a hands-on starting point, try your first n8n automation.
FAQ
Do I need to know how to code?
No. Every task on this list can be started with no-code tools. Some get easier later if you are comfortable with a little logic, like "if this, then that," but you can learn that as you go.
How much time will these save?
It depends on how often you do the task, how messy the inputs are, and how much checking the output needs. Measure the task before and after on your own work rather than relying on anyone's estimate, including ours.
Is it safe to put company data into AI tools?
It depends on the tool, your plan, and your company's rules. Ask IT or your manager which tools are approved for which kinds of data before you build. When in doubt, start with tasks that use non-sensitive information.
Which task should I start with?
The one you do most often, with the most predictable steps, and the lowest cost if something goes wrong. Drafting tasks (meeting notes, reports, first drafts) are usually the safest first build.
Can my employer pay for training on this?
Often, yes, if you frame it around a specific task. See how to get your employer to pay for an AI course, including an email template.
Build one of these with an instructor
The AI for Automation track runs 8 live sessions over 4 weeks. You pick a real task from your job and ship a working automation by session 8, with n8n, Zapier, Make and Claude Cowork.
Apply for January 2027Book a free 60-min session