AI can do 100% of the search. Real estate and aerospace keep the sign-off.
A real estate software founder in West Des Moines, a buyers agency founder in Melbourne and an AI founder in Edinburgh agree on exactly where the machine stops.
Retyping contract data? Hand it over. Searching SAP, Teamcenter and CAD files? Hand that over too.
Deciding a file is ready to close, or signing off a safety-critical design? That one has a human name on it.
Three founders, three countries, two industries. One line.
Notice what moves and what stays: the typing and the searching go, the answering-for stays.
Should a role go to a person, an AI agent, or both? We asked founders which part of a job AI actually took, and which part it didn't. Right now, every company is drawing that line, whether it admits it or not.
Retyping went to the bot. Readiness didn't.
Dane Maxwell, Founder of Paperless Pipeline (West Des Moines, United States)
“For a transaction coordinator, retyping contract data can move to AI, but deciding whether a file is ready to close still needs a person.”
“I think it is letting software save a contract value nobody checked. A wrong closing date quietly shifts every due date anchored to it, and the agent finds out when a deadline has already passed.”
Data entry is the solved part. The risk is trusting a value nobody looked at, and the cost lands downstream, not at the keyboard.
Melbourne: let AI gather it, let a human weigh it.
Ritesh Tandon, Founder of Nest or Invest - Buyers Agents (Melbourne, Australia)
“AI can completely automate data collection and preliminary property research, but human judgement remains essential for assessing risk, negotiating deals and understanding clients’ individual circumstances.”
Tandon's answer to human or agent is both. The machine does the legwork, the person answers for the outcome and for the client in front of them.
Aerospace hands AI the whole search and keeps the stamp.
Dr. Oliver King-Smith, Founder & CEO of smartR AI (Edinburgh, United Kingdom)
“When aerospace engineers spend up to 30% of project time searching fragmented data across existing systems such as SAP, Teamcenter, and CAD files, AI can automate 100% of that search. However, the final design judgment, safety sign-off, and compliance approval must remain human.”
“In defense and aerospace, customers insist on keeping humans in the loop for anything touching compliance documentation and safety-critical tolerances (such as ITAR-regulated specs).”
King-Smith's goal is augmentation, not cutting headcount. Less hunting for data, more complexity handled by the engineers already there. Keep the people, shrink the hunt.
Where they split
Dane Maxwell, West Des Moines: data entry moves to AI, and readiness "still needs a person".
Ritesh Tandon, Melbourne: AI takes data collection and preliminary research, while "human judgement remains essential".
Oliver King-Smith, Edinburgh: AI takes the search, while safety sign-off and compliance approval "must remain human".
What differs is the cost of a miss, and who answers for it: a passed deadline in a deal file, or a compliance failure on safety-critical tolerances.
What doesn't differ: nobody wants the person gone. They want the person freed up for the work that carries accountability.
My take
The pattern here is simple: AI takes the lookup, a person takes the liability. Before automating a role, find the point where a wrong answer costs real money or safety, and put a human name on it. Everything upstream of that point is fair game. If you can't name the person who signs, you're not ready to automate the step.
Your turn
What's the last task you handed to AI for good, and what's the one you took back?
Next: agents gave one team roughly five people's output per employee, one developer at least 150% more productivity, and a four-person team 60% of its research time back, yet revenue and headcount barely moved.
