What Should You Never Automate? What We Won’t Build
A good AI consultant can tell you what they would refuse to automate. Here is our list, and the reasoning behind it.
Automation should remove repetitive work, not remove judgment. We keep a human approval step on consequential actions, and we will not build systems that make people decisions, move money unreviewed, impersonate or deceive, or send mass messages to people who never asked. Asking any provider what they would refuse is one of the fastest ways to judge them.
Our rule of thumb
Rules should automate what is certain. AI can handle what is ambiguous, with a person checking the result. Humans own consequences. If a mistake would be costly or hard to undo, a person approves the action. That is how every system we build is designed, and it is why the pricing page says plainly that consequential actions keep a human approval step.
What we will not automate
Decisions about people
Hiring, firing, discipline, pay, credit and benefits belong to people who can be held accountable. A system can organize applications or prepare a summary. It should not make the call.
Moving money without approval
Automation can prepare an invoice, flag an anomaly or queue a payment. A person approves before funds move.
Professional advice sent straight to your customers
Medical, legal and financial guidance needs a qualified human reviewing it. A draft is fine. An unreviewed answer is not.
Fake reviews, invented testimonials or impersonation
If it would deceive a customer about who is speaking or what happened, we will not build it, however it is framed.
Bulk messages to people who did not ask
Writing to your own customers and contacts is ordinary business. Mass outreach to strangers is spam, and it damages the sender.
Anything irreversible that nobody reviews
If you could not undo it and no person sees it first, it needs a checkpoint added or it does not get built.
What we automate instead
The pattern is to automate the preparation and keep the decision. A recorded meeting becomes a drafted follow-up that waits for approval. An invoice is read and filed, with exceptions routed to a person. The work arrives ready; your people make the call.
Questions to ask any provider
- What would you refuse to automate, and why?
- Where does a person approve before something goes out?
- Who owns the system, and can I take it elsewhere?
- What happens to my data, and which tools see it?
If you are not sure which of your tasks are safe to automate, start with the free AI Opportunity Snapshot. It scores judgment and risk alongside time saved, and it tells you when a human should stay in the loop.
Takeaways
- Automate repetition, not judgment.
- Keep a human approval on anything consequential or hard to undo.
- A provider who cannot name what they would refuse is a provider to question.
Where would this work in your business?
Take the two-minute AI Opportunity Snapshot and we will point to where automation is most likely to pay off for you.