AI Academy/Automation recipes

Recruiting Automation Recipes: Four Workflows That Run on a Schedule

Four step by step recruiting automations that run on their own once built: a Monday list of companies that are hiring, a daily new applicant digest, follow ups for quiet candidates and a weekly CRM clean up.

4recipes

ResourceNo code

The short answer

Each recipe below is a small job you set up once and then leave to run: every Monday, every morning or every week. Each one gives you two ways to build it: a scheduled task inside ChatGPT, Claude or Grok, or an automation tool such as Zapier, Make or n8n.

Every recipe has a human checkpoint. The automation gathers, sorts and drafts. You decide who gets contacted and what gets sent.

Monday list of hiring companies
Daily new applicant digest
Follow up quiet candidates
Weekly CRM clean up

Which route should you use?

Route	Best for	What it costs to start
Scheduled task in ChatGPT	Research and summaries on a timer	Tasks are on every plan, with 3 active tasks on Free and Go and 5 on Plus, according to OpenAI’s tasks guide
Scheduled task in Claude	Research plus actions in connected tools	Pro, Max, Team and Enterprise plans, according to Claude’s scheduling guide
Automations in Grok	Research on a timer, or a job that starts when a matching email arrives	Scheduled automations on every plan; email triggers need SuperGrok, per xAI’s Automations post
Zapier	Moving data between apps when something happens	Free plan with 100 tasks a month and two step Zaps, per Zapier pricing
Make	Longer workflows with branches	Free plan with 1,000 credits a month and two active scenarios, per Make pricing
n8n	Teams with someone technical	Free if you host it yourself; the cloud version is paid after a trial, per n8n pricing

Plans and limits change often, so check the vendor page before you build. For the full picture of how scheduled tasks work, read Module 8, Lesson 4. For the basics of automation, start with Module 5, Lesson 4.

Recipe 1: A Monday list of companies that are hiring

Runs weeklyAgency and business developmentAllow an afternoon to set up

What you get: every Monday morning, a short table of companies in your niche that showed a hiring signal in the last seven days, with the signal, a link to the source and a suggested first line for outreach.

What you need: your ideal client profile (the playbook builder writes one for you) and a ChatGPT, Claude or Grok plan with scheduled tasks.

Route A: a scheduled task in ChatGPT, Claude or Grok

Open a new chat and paste the prompt below with your details filled in.
Check the first result by hand. Open every source link and delete any company that does not fit.
Tighten the prompt based on what was wrong, for example by adding industries to skip.
Ask the assistant to run it every Monday at 7am. In ChatGPT you can say that in the chat; in Claude, use the schedule option for the task; in Grok, create it at grok.com/automations.

Search the web for companies that match this profile: [ideal client profile: industry, size, location]. Find ones that in the last 7 days posted jobs for [roles you fill], announced funding, opened a new office or hired a new leader for [function]. Return a table with: company, signal, date, source link, and a one line opening I could use in outreach. Only include companies where you found a source link. Skip these companies: [off limits list]. Maximum 15 rows.

Route B: an automation tool

Create saved job alerts on the boards your targets use, delivered by email or RSS.
In Zapier or Make, trigger on each new alert, send the text to an AI step that extracts company name, role and location, and add a row to a Google Sheet.
Add a weekly step that emails you the new rows every Monday.

Human checkpoint: you choose who to contact. Never let this recipe send outreach on its own; a wrong company or stale signal is embarrassing at scale.

Watch out for: made up companies or links. Keep the rule that every row needs a source link, and open it before you use it.

Recipe 2: A daily digest of new applicants

Runs dailyCorporate and agencyAllow an afternoon to set up

What you get: one email or message each morning listing yesterday’s applicants per open job, with the must haves each one clearly meets, instead of a notification for every application.

What you need: an ATS with a daily summary setting, a Zapier or Make app, or an AI connector. Check your ATS first; many already offer a daily digest in notification settings, and that is the simplest route.

Route A: an automation tool

Trigger: new candidate or new application in your ATS.
Action: add a row to a Google Sheet with job, source and the application answers. Leave out contact details you do not need.
Every morning at 8am, a second automation sends the new rows to an AI step with the prompt below, then emails you the result. This needs more than two steps, so on Zapier it needs a paid plan; Make’s free plan can run it.

Here are yesterday's new applications: [rows]. Group them by job. For each applicant, list which of these must haves the application clearly shows: [must haves per job]. Do not rank, score or reject anyone. End with the count per job and any application that looks incomplete or duplicated.

Route B: Claude with a connected ATS

If your ATS has a connector, such as the Ashby connector in Claude’s directory, a daily scheduled task can read new applications directly and send you the same summary. Start with read only access. Module 8, Lesson 5 covers connections.

Human checkpoint: a person reviews every applicant. The digest saves you from clicking through notifications; it does not screen anyone out.

Watch out for: laws on automated screening. If the summary starts to act like a ranking, the rules in Module 4, Lesson 3 may apply.

Recipe 3: A follow up for candidates who have gone quiet

Runs daily or weeklyAny deskAllow a few hours to set up

What you get: a list of candidates who have not replied in a set number of days, each with a drafted follow up waiting in your drafts folder for you to check and send.

What you need: a CRM or ATS with a “last contacted” date that is kept up to date, and either an outreach tool or an email connector.

Route A: your outreach tool

If you already use a sequencer from the Outreach category, build the follow ups as sequence steps that stop when someone replies. This is the most reliable route because the tool tracks replies for you.

Route B: a scheduled task that drafts, never sends

Export or connect your pipeline so the task can see name, role, stage and last contacted date.
Schedule the prompt below for Tuesday and Friday mornings.
If your email is connected, ask for the follow ups to be saved as drafts only.

From this pipeline: [export or connected CRM], find candidates in the stages [Contacted, Replied, Screened] whose last contact was more than [7] days ago. For each one, draft a follow up under 60 words that adds something new about the role rather than just checking in. Save them as drafts; do not send anything. Give me a list of who you drafted for and why.

Human checkpoint: you read and send each message. People who have gone quiet often have a reason, and a tone deaf automatic nudge costs you the relationship.

Watch out for: anyone who opted out. Include your do not contact rules in the prompt and in your CRM.

Recipe 4: A weekly CRM clean up

Runs weeklyAny deskAllow an afternoon to set up

What you get: every Friday, a short fix list: duplicate records, records missing required fields, candidates stuck in one stage too long and jobs with no activity.

What you need: a list of the fields every record must have (the playbook builder writes one) and a weekly CSV export, or a CRM connector such as HubSpot for Claude.

Route A: a weekly export

Save a CRM view with the columns you care about and export it as CSV each Friday. Many CRMs can email a scheduled export.
Upload it to ChatGPT, Claude or Grok with the prompt below.
Fix the records yourself, or hand the list to whoever owns the data.

This is this week's CRM export: [attach CSV]. Required fields are: [list]. Find: likely duplicates (same email, or same name and company), records missing any required field, candidates in the same stage for more than [21] days, and jobs with no activity in [14] days. Return four short tables with the record ID and what is wrong. Do not change or delete anything.

Route B: a connected CRM

With a connector, schedule the same prompt weekly and let it read the CRM directly. Keep the connection read only for this job. Merging and deleting records should stay manual until you trust the results for several weeks.

Human checkpoint: a person approves every merge and deletion.

Watch out for: privacy. A full CRM export is personal data, so use a business AI plan or remove names and contact columns first.

How do you keep an automation from going wrong?

Run it by hand for two weeks before you schedule it. Keep every recipe read only or draft only at first. Put a limit on anything that spends credits. And write down what each automation does and who owns it, because the person who built it will not always be the one who finds it broken.

Quick answers

Do I need to code?

No. Every recipe here works with plain English prompts and point and click automation tools. n8n is the only option that suits someone technical.

Which is cheaper, an AI scheduled task or Zapier?

If you already pay for ChatGPT, Claude or Grok, a scheduled task costs nothing extra. Zapier and Make are better at moving data between apps the moment something happens, and their free plans cover light use.

Can an automation send emails to candidates for me?

It can, but these recipes deliberately stop at drafts. Sending at scale without review is how recruiters end up in spam folders and on the wrong side of candidates.

Where can I compare automation tools?

The Workflow automation category lists the tools with prices, and the free MCP readiness checker shows which of your tools an AI assistant can reach.