By Michael-Patrick Moroney

A Goldman Sachs marketing analyst began in 2023 drafting competitive briefs with her in-house "GS AI Assistant." It was a beginning with outlines and summaries. She got bolder. She made it play devil's advocate, tell her where her logic was missing.

She did better work. Coworkers copied her process. She got promoted.

Around the room, another analyst approached it differently. She loaded the entire assignment into the chatbot, didn't even edit it, and fired it off. Her article read just fine - but it didn't take off. Nobody went with her.

Same tools. Truly different results.

Why Experience Still Matters

Veteran staff are not only leveraging AI to be productive. They understand where the human nuance resides. They understand who requires a heads-up on a significant decision, when the tone of an email makes all the difference, and how to navigate that awkward silence in a meeting.

"AI is great at generating options," says MIT Sloan professor Danielle Li, "but someone still needs to know which ones to select - and when to distrust them."

McKinsey echoed that in a 2024 workplace report: "If you insert AI into an imperfect process, it just gets you imperfect results faster."

What CarGurus Did Right

In 2024, CarGurus launched a companywide test called "AI Forward." It wasn't top-down. Product, engineering, marketing, and legal teams explored tools like ChatGPT and Midjourney together, without the expectation of having to deliver.

"Emotional support overrode the tools themselves," explained Sarah Rich, data scientist leading the effort. "We didn't want to just have people use AI. We wanted people to trust their own judgment when they used it."

They discovered the greatest users weren't necessarily tech savvy. They were people who understood how their business actually worked.

*Understanding the game is different from knowing the code.*

What the Research Indicates

One meta-study of 106 experiments concluded that humans working with AI, especially on tasks involving creativity, did better than both alone. In one case, bird identification accuracy increased from 81 to 90 percent with human-AI combination.

But a second study, with 3,500 people, showed that over-dependence might actually backfire. When humans relied on AI to produce first drafts, the drafts were higher quality. But then, if they wrote alone afterward, performance declined. Their instincts withered. Judgment decreased.

AI excels when we are engaged.

*Jenny Holzer LED installation: “Abuse of power comes as no surprise.”*

Same Writers, Same Tools, Differing Results

Jordan is a new employee at the firm. She pastes the memo question into ChatGPT, reviews the results, and sends it.

Alex has been on the job for about ten years. She employs AI to explore sets of angles. She rephrases the summary, rapidly scans against last quarter's reports, adds a context paragraph she knows leadership will ask for, and calls to add in legal before the deck goes out.

Jordan finishes in ten minutes. Decisions are made by Alex's memo.

*Quiet leadership. Deep context. Knowing which numbers matter.*

The System That's Working

Here are some tips for reaching the top of research and practice:

Use AI for routine work, not for decision work.
Keep your hands on anything that involves judgment, tone, or trust.
Edit and stand behind the final product - it still carries your name.
Be transparent about what works. Cultures that talk about AI perform better with it.
Build stamina. The initial shine will fade. The constant edge over time.

Final Thought

AI is not beating thoughtful work. It's stressing it.

Those who are thriving today are not automating. They're staying connected. They know how their business functions. They're asking questions more. They're not users - they're interpreters.

That's the difference between output and influence.

And that is who thrives with AI.