How I Became the Chief Human-in-the-Loop Officer

by Susan Smith
Date Published June 24, 2025 - Last Updated June 24, 2025

I’ve built a career on change and productivity. My first professional job was with the University of Arkansas, where I started as a self-proclaimed “peon.” One day, my boss — an accomplished PhD — threw a massive three-ring binder titled Assembler Language onto my desk and asked, “How fast can you come up to speed on this?”

It had nothing to do with my role — except for the line in my job description: “Other duties as assigned.” I figured it out and got it done.

Then, we needed to network our computers and connect to national scientific libraries. So, I retrained as a network engineer. Got it done again.

“Doing it again” became my thing

Thankfully, my “I’ll do it all myself” approach eventually matured into a team-building mindset: build the process, build the team, set expectations and insert logical checkpoints to keep projects moving forward. Looking back, I realize I was practicing “human-in-the-loop” (HITL) strategies long before the term was popular.

Why HITL Matters in the AI Era

A key to successful AI projects is building truthful datasets — which can’t happen without a human in the loop. HITLs bring oversight, ethical reasoning, and common-sense judgment that machines simply can’t replicate.

Now that I’m deep into my “AI Era,” I’m still out front — doing it again.

To quote Matt Beran: “AI won’t take your job. But someone who knows how to work with AI probably will.”

That “someone” is the person who designs your data pipelines, workflows and LLM prompts — and knows where human oversight is required to ensure data integrity and security.

Old-School Skills Are Today’s Secret Weapon

The truth? My old-school change management and productivity tools still work. But you have to use them. If you automate a broken process without analyzing and redesigning it, you don’t get efficiency —you get an automatic mess.

Let’s look at what paves the way to AI success:

Rebuild (and Document) Your Processes

Processes can be your best friend or worst enemy.
Outdated workflows need to be rethought and rewritten.
Clear, well-documented processes improve data quality and transparency — critical to trustworthy AI.
You cannot automate what you don’t understand.

Normalize Your Data

If you’re not a DBA, here’s the simple version: clean up your work.

Organize your data.
Reduce redundancy.
Standardize fields and formats.

A cluttered spreadsheet won’t magically become smart just because you connect it to an LLM. A little Excel discipline goes a long way.

Create Knowledge Bases That Matter

Standardized SOPs, decision trees and documented workflows don’t just support your team — they train your AI. These become the guardrails for your models and the built-in checkpoints for your human-in-the-loop strategies.