47,000 support tickets a month, closed without a person touching any of them. That’s the number that stopped my scroll today, and the post behind it is worth a few minutes of your time.
I came across it from a LinkedIn creator who sold their last company. Going through that sale, the author noticed how much of ops was just the same steps over and over. Their focus now is Zapier’s new Next Gen Zaps, and the claim is a bold one: onboarding doesn’t need a new hire anymore. The author also thinks most founders hire for it far too early.
📊 The Numbers Behind the Claim
The author points to three teams already running this kind of automation on Zapier:
- JustEat: restaurant partners onboard 60% faster
- Palo Alto Networks: $150,000 a year saved, with demo accounts set up for 3,000 employees
- Mercari: 47,000 tickets a month resolved without a human
Here’s how I read these numbers. JustEat’s 60% is about time-to-value. A partner who’s live sooner starts bringing in revenue sooner, so faster onboarding pays off on the revenue side.
The Palo Alto figure is about cost. In plenty of markets, $150,000 a year is close to a full-time ops salary, maybe more. Creating 3,000 demo accounts by hand is exactly the kind of work that ends up as somebody’s whole job.
Mercari’s number is the one that shows scale. 47,000 tickets a month comes to more than 1,500 a day. No small team can absorb that volume by adding people one at a time.
🔁 The Hiring Trap Most Founders Fall Into
The author describes a pattern I’ve seen at a lot of startups:
A customer signs.
The founder scrambles.
Then someone gets hired just to keep up.
When you grow this way, headcount rises in a straight line with customers. Every new account means more manual work, and eventually more salaries. The author’s argument is that a big chunk of that work is repeatable steps, and repeatable steps can be automated.
⚙️ How Next Gen Zaps Actually Work
You describe a process in plain language inside Claude, ChatGPT or Cursor. It then deploys to Zapier and runs on its own, even with your laptop closed. No more babysitting a script on your own machine.
According to the creator, here’s what makes it different:
- AI only where judgment matters: the onboarding plan and the at-risk check
- Everything else runs as fixed steps
- Zapier suggests locking in more as patterns repeat, so it gets cheaper
- When your stack changes, an agent spots the break and drafts a fix
- It can pause for a human to approve
- Admin app controls per team are rolling out
The first two points are the clever part, in my view. A lot of teams go too far the other way and run every step through an AI model, which makes the whole thing slower, pricier and harder to predict. Here the AI only handles the two decisions that need judgment, and the rest runs on rails. As patterns repeat, more steps get locked in as fixed logic, so your cost per customer keeps dropping.
💡 How to Apply This to Your Own Onboarding
Even if you’re not ready to change tools, the approach works anywhere. Here’s a simple way to start:
- Write down every step that happens between “contract signed” and “customer is live”
- Mark the steps that need real judgment, like building a custom plan or spotting an account that’s struggling
- Treat everything else as fixed, rule-based automation
- Add a human approval checkpoint before anything goes out to the customer
- Every few weeks, look for repeating patterns you can lock in as fixed steps
If you try describing the process to an AI assistant, keep it concrete. Something like: “When a new customer signs, create their workspace, send the welcome email, draft a 30-day onboarding plan based on their goals, and flag them as at-risk if they haven’t logged in after 7 days. Ask me to approve the plan before sending.”
Onboarding scales with customers, not headcount.
That one line from the author sums up the strategy. Before you post your next ops job, check how much of that role is really just repeated steps.
The full post on LinkedIn has more context from the author, so give it a read.