I was in a call last month with a CEO who wanted to automate everything. Customer onboarding, support tickets, sales follow ups, even the quarterly business reviews. His exact words were "If we can automate it, we should."
I've spent 12 years building WordPress sites, wiring up CRMs like Keap and HubSpot, and lately connecting AI tools to just about everything. I love automation. But that CEO question made me uncomfortable, and it took me a few days to figure out why.
The things I regret automating
Three years ago I built a fully automated onboarding sequence for a client in GoHighLevel. New customers got 14 emails over 30 days. Each email had a video, a checklist, and a CTA. Open rates were decent at around 28%. Click rates were about 4%.
The problem showed up six months later. Customer lifetime value dropped by 31%. Churn in the first 90 days went up. When we interviewed the customers who left, most of them said they never felt like they actually connected with anyone at the company.
We rebuilt it. Now the first email comes from a real person, the account manager records a personal Loom video at day 3, and there's a mandatory 15 minute call at day 10. The rest can stay automated. Churn dropped back down. It takes more time, but the math works.
Where the line actually is
I used to think the line was about complexity. Automate the simple stuff, keep humans for the complex decisions. That's wrong.
The real line is about relationship value. If the interaction builds trust or changes how a customer sees your company, a human needs to be involved. If it's purely transactional and the customer expects speed over personality, automate it.
Here's what I actually automate without hesitation:
- Data entry and CRM updates when a form is submitted or a purchase happens
- Sending invoices, receipts, and payment reminders
- Tagging and routing support tickets based on keywords before a human sees them
- Generating first draft reports from analytics data that someone reviews before sending
- Scheduling and calendar coordination
I don't automate the first conversation with a new lead anymore, even though I did for years. I tried using AI to handle initial sales conversations with a HubSpot chatbot integration last year. It could answer 80% of questions correctly. But the 20% it got wrong made us look careless, and we lost deals because of it. Now AI qualifies and collects information, then a human takes over before any pricing or strategy discussion happens.
What I tell clients now
When a client asks what they should automate, I ask them to list every customer interaction they have. Then I ask which ones they'd be embarrassed to admit were automated if the customer found out.
That embarrassment test works better than any framework I've found. If you'd feel awkward saying "that was a bot" when it matters, don't automate it.
The CEO from that call ended up automating about 60% of what he originally wanted. The other 40% stayed human. His support team was worried they'd lose their jobs, but instead they stopped doing data entry and started having better conversations. Customer satisfaction scores went up 22 points in four months.
Automation should make your humans more human, not replace the humanity. When it does the opposite, you've automated the wrong thing.
